Faculty of Engineering Dept. ELEC Pleinlaan 2 1050 Brussels, Belgium

Department ELEC : Fundamental Electricity and Instrumentation ANNUAL REPORT 2012

Editor: Ann Pintelon, Vrije Universiteit Brussel, Dept. Elec, Faculty of Engineering

Table of contents

0. Table of contents 1.

Introduction to the department ELEC............................................................................ 7 1.1

Introduction to the long term strategy of the department ELEC.................................. 7

1.2

Research TEAMS of the department ELEC: Overview of the 4 teams ........................... 8

1.2.1

Team A: Automatic Measurement Systems, Telecommunications and Laboratory of

Underwater Acoustics. ............................................................................................... 8 1.2.2

Team B: System Identification and Parameter Estimation of Linear and Non-Linear

Systems .................................................................................................................. 9 1.2.3

Team C: Applied Signal Processing for Engineering (ASPE) .................................... 9

1.2.4

Team D: Medical Measurements and Signal Analysis (M2ESA) ................................ 9

1.3

THE ELEC futsal team ....................................................................................... 10

1.4

Staff of ELEC (Status 01/01/2013)...................................................................... 11

1.4.1

General Director .......................................................................................... 11

1.4.2

Team A: Automatic Measurement systems, Telecommunications and Laboratory of

Underwater Acoustics .............................................................................................. 11 1.4.3

Team B: System Identification and Parameter Estimation of Linear and Nonlinear

Systems ................................................................................................................ 12 1.4.4

Team C: Applied Signal Processing for Engineering (ASPE) .................................. 15

1.4.5

Team D: Medical Measurements and Signal Analysis (M2ESA) .............................. 17

1.4.6

Technical and Administrative Staff .................................................................. 18

1.4.7

Professors Emeriti ........................................................................................ 19

1.4.8

In memoriam prof. em. Dr. Ronald Van Loon (1940-2012) .................................. 20

1.4.9

List of phone numbers and e-mail addresses (Status 01/01/2013)........................ 21

1.4.10

Industrial Partnership ................................................................................ 22

1.4.11

National or international contacts ................................................................ 22

1.4.11.1

Visiting professors/researchers ............................................................ 22

1.4.11.2

Scientific missions ............................................................................. 26

1

Annual report ELEC 2012

1.5

Organisation chart of department ELEC (1/1/2013) ............................................... 33

1.6

Functional organisation of the dept. ELEC ............................................................ 34

1.7

List of the most important measurement equipment .............................................. 35

1.7.1

Signal Generators ........................................................................................ 35

1.7.2

Spectrum Analysers, Impedance Analysers, Network Analysers ............................ 35

1.7.3

Digitizers .................................................................................................... 36

1.7.4

Miscellaneous .............................................................................................. 36

1.7.5

Underwater Acoustics ................................................................................... 37

1.8

Financial support 2012 ..................................................................................... 38

1.9

Awards .......................................................................................................... 39

1.9.1

Grade of Fellow (IEEE) .................................................................................. 39

1.9.2

Grade of Senior member (IEEE)...................................................................... 39

1.9.3

Awards from IEEE Instrumentation and Measurement Society (US)....................... 40

1.9.4

Award from IEEE Control Systems Society ........................................................ 41

1.9.5

Grade of Fellow (IFAC) .................................................................................. 41

1.9.6

Belgian Francqui Chair ULB ............................................................................ 41

1.9.7

Awards granted by the VUB, on the proposition of the department ELEC ................ 42

1.9.8

Distinguished Service Award from IMEKO ......................................................... 42

1.9.9

Joseph F. Keithley Award in Instrumentation and Measurement: IEEE Field Award ... 42

1.9.10

Doctor Honoris Causa ................................................................................ 43

1.9.11

Member of the “Royal Flemish Academy Of Belgium For Science And The Arts” ... 44

1.9.12

Paper/presentation awards (since 2008) ....................................................... 44

1.9.13

Master thesis awards................................................................................. 44

1.10

International conferences/workshops organised by the dept. ELEC ........................... 45

1.10.1 2.

16th IFAC Symposium on System Identification – SYSID 2012 .......................... 45

Short Description of the Research Projects/ Team......................................................... 48 2.1

Team a: automatic measurement systems, Telecommunications and laboratory of

underwater acoustics .................................................................................................. 48 2.1.1

Introduction to Team A ................................................................................. 48

2

Table of contents

2.1.2

Short Description of the Research Projects of Team A ......................................... 49

2.1.2.1

Modeling of the channel transfer function and the crosstalk for specific historical

connectivity practices in DSL copper networks and assessment of the effect on the achievable data rates .......................................................................................... 49 2.2

Team B: system identification and parameter estimation ........................................ 51

2.2.1

Introduction to Team B ................................................................................. 51

2.2.2

Short Description of the Research Projects of Team B ......................................... 56

2.2.2.2

Design of optimal inputs for nonlinear dynamic systems .............................. 58

2.2.2.3

User-friendly identification of massive MIMO systems ................................ 58

2.2.2.4

Measurement and estimation of Linear Parameter-Varying systems ............. 60

2.2.2.5

A frequency-domain formulation of Least-Squares support vector machines to

deal with coloured noise. ...................................................................................... 60 2.2.2.6

Nonlinear dynamic modeling with Neural Networks ..................................... 62

2.2.2.7

Identification

and

comparison

of

several

nonlinear

models

for

the

glucoregulatory system ........................................................................................ 64 2.2.2.8

Detection and Quantification of the Influence of Time Variation in Frequency

Response Function Measurements Using Arbitrary Excitations, .................................. 65 2.2.2.9

Brain Tissue Differentiation and Characterization using Piezoelectric Actuators

driven by multisine excitation .............................................................................. 68 2.2.2.10

Fast

algorithms

and

software

for

weighted

mosaic

Hankel

low-rank

approximation ................................................................................................... 69 2.2.2.11

Iterative update of the pole locations in a Wiener-Schetzen model ........... 72

2.2.2.12

Improved, user-friendly identification method for the nonlinear LFR block-

oriented model ................................................................................................... 73 2.2.2.13

Identification of a nonlinear LFR block-structure with two static nonlinearities . ...................................................................................................... 74

2.2.2.14 2.3

Characterisation and modelling of lithium-ion batteries for electric vehicles. 76

Team C: Applied Signal Processing for Engineering (ASPE)...................................... 77

2.3.1

Introduction to Applied Signal Processing for Engineering (ASPE) ......................... 77

2.3.2

Short Description of the Research Projects of Team C ......................................... 79

2.3.2.1

A Framework for Microwave Design Oriented Modeling ............................... 79

3

Annual report ELEC 2012

2.3.2.2

Study and development of analysis and design techniques in a model-based

framework ......................................................................................................... 79 2.3.2.3

Development and improvement of a forced oscillating technique (FOT) method

for non-invasive long impedance measurements and the use of (nonlinear) modeling for the monitoring of lung diseases.

2.4

.......................................................................... 80

2.3.2.4

Heat Transport in Borehole Heat Exchangers: New modeling approach ........... 81

2.3.2.5

An identification algorithm for parallel Wiener-Hammerstein systems ............. 83

2.3.2.6

Error bound on polynomial approximation of steep functions ........................ 84

Team D: Medical Measurements and Signal Analysis (M2ESA) .................................. 86

2.4.1

Introduction to project M2ESA ........................................................................ 86

2.4.2

Introduction to project MEMON ....................................................................... 87

2.4.3

Short Description of the Research Projects of TEAM D......................................... 88

2.4.3.1

A simple non-parametric pre-processing technique to correct for non-stationary

effects in measured data ...................................................................................... 88 2.4.3.2

Fractional order time series models for extracting the haemodynamic response

from functional Magnetic Resonance Imaging data ................................................... 90 2.4.3.3

Fractional models for modeling complex linear systems under poor frequency

resolution measurements ..................................................................................... 91 2.4.3.4

Discriminant Analysis for an automatic signal detection technique useful in

cognitive radios................................................................................................... 93 2.4.3.5

Characterization of the probability density functions of measured signals. ....... 94

2.4.3.6

A robust signal detection method for fMRI data under correct Rice conditions .. 95

2.4.3.7

Odd Random Phase Electrochemical Impedance Spectroscopy (ORP-EIS) for

Glucose Sensing .................................................................................................. 97 3.

Education .............................................................................................................. 99 3.1

The introduction of the bachelor-master structure ................................................. 99

3.2

BRUFACE (www.bruface.eu/EN/) ........................................................................ 99

3.3

Courses lectured in the faculty of engineering ..................................................... 100

3.3.1 3.4

Lectures and practical courses ...................................................................... 100

Minor “Measuring, Modelling, and Simulation of Dynamic Systems” offered within the

2nd Master Electronics and Information Technology........................................................ 102

4

Table of contents

3.5

Designing systems from concepts: the ping-pong tower project ............................. 103

3.5.1

Situating the project ................................................................................... 104

3.5.2

The project ............................................................................................... 105

3.5.3

From concept to working system................................................................... 105

3.5.3.1

Step 1: Understanding the problem........................................................ 105

3.5.3.2

Step 2: Strategic thinking..................................................................... 106

3.5.3.3

Step 3: Creating the different blocks ...................................................... 107

3.5.3.4

Step 4: Going back to the system level ................................................... 108

3.5.3.5

Step 5: Presentation............................................................................ 109

3.5.3.6

Conclusion ......................................................................................... 109

3.6

Courses lectured in the faculty of Science and Bio-Engineering .............................. 109

3.7

Doctoral Training programme .......................................................................... 109

3.8

National and international courses .................................................................... 110

3.8.1

National courses (since 2003): ..................................................................... 110

3.8.1.1

Identificatie van systemen (Identification of Systems) ............................... 110

3.8.1.2

Courses lectured at the Katholieke Universiteit Leuven (KUL)...................... 110

3.8.1.3

Open course program-IMEC academy: DSP concept explained with well-chosen

exercises. 111 3.8.2

4.

International courses (since 2003): ............................................................... 112

3.8.2.1

Characterisation of Multiport Systems through 3-port LSNA Measurements ... 112

3.8.2.2

The use of multisines ........................................................................... 112

3.8.2.3

GIS training in SEAFDEC, Thailand ......................................................... 112

3.8.2.4

Measuring, Modeling, and Designing in a Nonlinear Environment ................. 112

3.8.2.5

VUB - doctoral school on Identification of Nonlinear Dynamic Systems ......... 112

Bibliography......................................................................................................... 114 4.1

Books .......................................................................................................... 114

4.2

Journal papers (2012) .................................................................................... 115

4.3

Conference papers (2012) ............................................................................... 122

4.4

Abstracts (2012) ........................................................................................... 133

5

Annual report ELEC 2012

5.

4.5

Workshops (2012) ......................................................................................... 135

4.6

Seminar presentations organised by the dept. ELEC (2012) .................................. 139

4.7

Patents ........................................................................................................ 141

4.8

Doctoral dissertations ..................................................................................... 142

4.9

Thesis tot het behalen van het aggregaat van het hoger onderwijs ......................... 152

Location of the university (VUB) and the dept. ELEC ................................................... 153 5.1

Arrival by car: ............................................................................................... 153

5.2

From the Brussels National airport at Zaventem:................................................. 153

5.3

From Brussels South Airport (Charleroi)............................................................. 154

5.4

Arrival by train: ............................................................................................. 154

5.5

Arrival by subway (€ 2,00/jump-ticket): ............................................................ 155

6

Introduction to the department ELEC

1. Introduction to the department ELEC 1.1

INTRODUCTION TO THE LONG TERM STRATEGY OF THE DEPARTMENT ELEC

‘ELEC’ stands for “Fundamental Electricity and Instrumentation” (in Dutch: “Algemene Elektriciteit en Instrumentatie”) and the name corresponds to the educational and research tasks and objectives of the department. The main research activity of the department is the development of new measurement techniques using advanced signal processing methods, embedded in an identification framework. When we make a measurement, we have to make a number of decisions: firstly a model for the considered part of reality is proposed (e.g. for a resistance measurement Ohm’s law can be selected, describing the relation between the voltage across the resistor and the current through it); next a number of measurements is made (e.g. a number of current and voltage measurements); finally the quantities of interest are extracted from these measurements by matching the model to the data. Often an intuitive approach is used. However, in the presence of measurement errors this can lead to a very poor and even dangerous behaviour: the user wouldn't remark that something is going seriously wrong. This is the major motivation for the development of the identification theory. It offers a systematic approach to ‘optimally’ fit mathematical models to experimental data, eliminating stochastic distortions as much as possible. As such it can be considered as the modern formulation of the measurement problem, and for that reason the identification approach is the “fil rouge” in most of the activities of the department. Each measurement (or identification session) consists of a series of basic steps: Collect information about the system; Select a (non) parametric model structure to represent the system; Select the model parameters to fit the model as well as possible to the measurements (this requires a “goodness of fit” criterion); Validate the selected model.

Most of the research activities of the department are related to one of these problems, but this does not narrow our focus. At this moment we deal with a very wide scope of application fields: Systems covering the frequency range from a few mHz up to 50 GHz,

7

Annual report ELEC 2012

Linear systems and non-linear systems, Lumped systems and distributed systems. We applied the measurement and modelling techniques to the identification of electrical machines (frequency range 0.01 Hz till 1 kHz, linear models, 2 inputs and 2 outputs), mechanical vibrating systems (frequency range below 5 kHz, linear or non-linear, up to 2 inputs/2 outputs), electronic circuits and filters (frequency range up to 5 MHz, linear and non-linear models, single input/single output or multiple input/multiple output), underwater acoustics (frequency range up to a few MHz, 1 input and 2 outputs), distributed systems (telecommunication lines, up to a few hundreds MHz), microwave applications (frequency range up to 50 GHz, non-linear, 6-port measurements). Since a few years we apply those methods also to the analysis of biological samples used as records of global climate change. For some of these applications the efforts are focused on the development of new measurement instruments (measurement of telecommunication lines, non-linear microwave analyser), for others we focused completely on the development of new data processing and modelling techniques, or even worked on the underlying fundamental theoretical aspects. To cover this wide application range, we make use of an extensive measurement park. Most of it nowadays consists of VXI-based data-acquisition systems, although we have also some classical instruments like network and spectrum analysers. All these instruments are computer controlled in a MatlabTM environment.

1.2

RESEARCH TEAMS OF THE DEPARTMENT ELEC: OVERVIEW OF THE 4 TEAMS

1.2.1

Team A: Automatic Measurement Systems, Telecommunications and Laboratory of Underwater Acoustics. Identification of distributed systems Incorporation of knowledge engineering in complex measurement problems Application of information theory in data telecommunication by wire problems (xDSL) Modelling and Identification of transmission lines and wireless channels (LOS and NLOS) Wireless local loop Signal processing techniques related to measurement in Earth Science problems Development of PC and PXI based instrumentation Underwater acoustic studies Environmental G.I.S. development

8

Introduction to the department ELEC

4G communication Navigation techniques, including but not limited by, applications for Location Based Services Positioning Techniques using the cellular network (focus on GSM) Proactive Location-based Services (LBS) using multiple positioning technologies.

1.2.2

Team B: System Identification and Parameter Estimation of Linear and NonLinear Systems Study and development of basic concepts on identification theory Identification of linear and nonlinear concentrated and distributed systems Experiment design Time and frequency domain system identification Development and distribution of a system identification toolbox Identification of time-varying systems

1.2.3

Team C: Applied Signal Processing for Engineering (ASPE) Instrumentation setup contributions. Instrumentation calibration contributions. Modelling high frequency nonlinear systems

1.2.4

Team D: Medical Measurements and Signal Analysis (M2ESA) Development of customized measurement setups Pre/post - processing of medical measurement data User-friendly guidelines to accurate and practical signal analysis Customized linear and nonlinear modeling techniques for medical and high-frequency applications

9

Annual report ELEC 2012

Address:

Vrije Universiteit Brussel, Department ELEC, Pleinlaan 2, Building K, 6th floor, B-1050 Brussels, Belgium

Secretariat: phone: +32 (0)2 629 29 47 or 27 67,

e - mail:

Telefax: +32 (0)2 629 28 50

[email protected] [email protected]

www-environment: http://wwwir.vub.ac.be/elec/

1.3

THE ELEC FUTSAL TEAM

The mixed girls/boys ELEC futsal team was founded in September 2010, sponsored by the ELEC department. After some intensive recruiting and scouting we got a very motivated team. The goal of the first season was to establish a good team spirit with the support of our international supporter clan. Most of the team were fairly new to the sport, or hadn't played for decades. However, through persistence and some fine teamwork we managed to get a draw and a win in the second season. Somewhere in the middle of the competition we started to train more regularly. Indeed, practice makes perfect... The third season is now halfway done, and we already managed to score two wins! We expect even more victories in the near future now that we have established some good team tactics, and our talents start to develop. It must be noted that the winning of matches is not the priority. Just like we do during the day, we try to work together. The emphasis is on team work, fun and openness. Everyone is welcome to join, especially in the (now weekly) training sessions.

10

Introduction to the department ELEC

1.4

STAFF OF ELEC (STATUS 01/01/2013)

1.4.1

General Director

Johan Schoukens (head of the department) received both the degree of master in electrical engineering in 1980 and the degree of doctor in engineering (PhD) in 1985 from the Vrije Universiteit Brussel (VUB), Brussels, Belgium. From 1981 to 2000, Dr. Schoukens was a researcher of the Belgian National Fund for Scientific Research (FWO-Vlaanderen) at the Electrical Engineering (ELEC) Department of the VUB where he is currently a full-time professor in electrical engineering. His main research interests include system identification, signal processing, and measurement techniques. Dr. Schoukens has been a Fellow of IEEE since 1997. He was the recipient of the 2002 Andrew R. Chi Best Paper Award of the IEEE Transactions on Instrumentation and Measurement, the 2002 Society Distinguished Service Award from the IEEE Instrumentation and Measurement Society, and the 2007 Belgian Francqui Chair at the Université Libre de Bruxelles (Belgium). Since 2010, he is a member of Royal Flemish Academy of Belgium for Sciences and the Arts. In 2011 he received a Doctor Honoris Causa degree from the Budapest University of Technology and Economics (Hungary). Phone: +32 (0)2 629 29 44 e-mail: [email protected]

1.4.2

Team A: Automatic Measurement systems, Telecommunications and Laboratory of Underwater Acoustics

Leo Van Biesen (coordinator of team A) was born in Elsene, Belgium, on August 31, 1955. He received the degree of Electro-Mechanical Engineer from the Vrije Universiteit Brussel (VUB), Brussels in 1978, and the Doctoral degree (PhD) from the same university in 1983. Currently he is a full senior professor. He teaches courses on fundamental electricity, electrical measurement techniques, signal theory, computer-controlled measurement systems, telecommunication, underwater acoustics and Geographical Information Systems for sustainable development of environments. His current interests are signal theory, modern spectral estimators, time domain reflectometry, wireless local loops, xDSL technologies, underwater acoustics, and expert systems for intelligent instrumentation. He has been chairman of IMEKO TC-7 from 1994-2000 and President Elect of IMEKO for the period 2000-2003 and the liaison Officer between the IEEE and IMEKO. Prof. Dr. Ir. Leo Van Biesen has been president of IMEKO until September 2006. He is also member of the board of FITCE Belgium and of USRSI Belgium. Phone: +32 (0)2 629 29 43 e-mail: [email protected]

Taka Yoshizawa was born in Nagoya, Japan, on 22nd April, 1961.

He has total of 24 years of industry experience, much of it in 2G/3G cellular infrastructure product design and development at Motorola in the US, where he gained experience in most of the engineering work involved with the lifecycle of mobile infrastructure system. After moving to Belgium in 2006, he joined Technicolor (former Thomson) in Edegem as a System Architect for the femtocell project. Since 2011, he is a Senior System and Standard Engineer at Ubiquisys Ltd in the UK. Since 2007, he has been actively contributing to 3GPP, Broadband Forum, and Femto Forum for femtocell related standardization activities. In 2009, he won the inaugural Femto Forum Award for the recognition for his contribution to femtocell standardization. He has filed ~10 patents (most of them still pending) in both US and Belgium for telecom related technologies. He has a BS degree of Information and Computer Science from Georgia Institute of Technology and an MS Telecommunication from Southern Methodist Univ. in 1992 and 2002, respectively. He is currently pursuing Ph.D. under the guidance of Prof. Van Biesen. His research interest is next generation mobile technology, such as Relay Node, Heterogeneous Network, and Femtocell. Email: [email protected]

11

Annual report ELEC 2012

1.4.3

Team B: System Identification and Parameter Estimation of Linear and Nonlinear Systems

Johan Schoukens (head of the department) received both the degree of master in electrical engineering in 1980 and the degree of doctor in engineering (PhD) in 1985 from the Vrije Universiteit Brussel (VUB), Brussels, Belgium. From 1981 to 2000, Dr. Schoukens was a researcher of the Belgian National Fund for Scientific Research (FWO-Vlaanderen) at the Electrical Engineering (ELEC) Department of the VUB where he is currently a full-time professor in electrical engineering. His main research interests include system identification, signal processing, and measurement techniques. Dr. Schoukens has been a Fellow of IEEE since 1997. He was the recipient of the 2002 Andrew R. Chi Best Paper Award of the IEEE Transactions on Instrumentation and Measurement, the 2002 Society Distinguished Service Award from the IEEE Instrumentation and Measurement Society, and the 2007 Belgian Francqui Chair at the Université Libre de Bruxelles (Belgium). Since 2010, he is a member of Royal Flemish Academy of Belgium for Sciences and the Arts. In 2011 he received a Doctor Honoris Causa degree from the Budapest University of Technology and Economics (Hungary). Phone: +32 (0)2 629 29 44 e-mail: [email protected]

Péter Zoltán Csurcsia was born in Budapest, Hungary, on 17.05. 1985.

He obtained his Bachelor of Engineering diploma (BEng in EE, summa cum laude) and his Technical Teacher diploma (Ed in EE, summa cum laude) from Budapest Tech in 2007 and 2008. Parallel with Electrical Engineering he studied technical informatics at Budapest Tech between 2004 and 2008. From 2008 he was a student at the Budapest University of Technology and Economics (BUTE) and at Vienna University of Technology. He graduated in MSc in Embedded Systems and in Applied Informatics (MSc, summa cum laude) in 2010. Now, he is a doctoral student at the BUTE (his advisor Prof. Dr. István Kollár) and at the Vrije Universiteat Brussel (with Prof. Dr. ir. Johan Schoukens). He worked as an IT Teacher from 2006-2010 and as a Program/Web designer. His research interests cover the topics of system identification, digital signal processing, software and and internet technologies. Phone: +32 (0)2 629 29 46 e-mail: [email protected] Alexander De Cock was born in Brussels (Jette), Belgium, on the 29th of April, 1989. He received the degree of Master of Science: Engineering Sciences: Electronics and Information Technology in September 2012 from the Vrije Universiteit Brussel (Belgium). While his thesis was focused on the application of l1-regularisation in the context of frequency domain identification, his main interest is now optimal experiment design for the identification of structured nonlinear system.

Phone:+32 (0)2 629 36 65 e-mail: [email protected] Egon Geerardyn was born on April 15th, 1988 in Jette (Brussels), Belgium. He graduated as an Electrical Engineer in Electronics and Information Theory (profile Measurements, Modelling and Simulations) in 2011 at the Vrije Universiteit Brussel. In October 2011 he joined the department ELEC as a PhD student. His main interests comprises system identification of linear and nonlinear systems, workflow automation and Linux.

Phone: +32 (0)2 629 28 44 e-mail: [email protected]

Jan Goos was born in Geel (Belgium) in 1986. He graduated from the Katholieke Universiteit Leuven as an engineer in Computer sciences (Artificial Intelligence) in 2009 and in Mathematical Engineering in 2011. In October 2011 he joined the department of ELEC as a PhD student. His main interests are the measurement and modeling of Linear Parameter Varying (LPV) systems, but he also loves non-linear dynamics.

Phone: +32 (0)2 629 29 46 e-mail: [email protected]

12

Introduction to the department ELEC

Mariya Ishteva was born on April 15, 1980, in Sofia, Bulgaria. She received a BSc degree in Computer Science in 2002 from Sofia University, Bulgaria and Msc degree in Mathematics in 2005 from University of Karlsruhe (TH), Germany. Mariya defended her PhD thesis in December 2009 at Katholieke Universiteit Leuven, Belgium. Afterwards she was a postdoc at Universite catholique de Louvain, Belgium in 2010 and at Georgia Institute of Technology, USA in 2011 and 2012. Since January 2013 Mariya is a postdoc at department ELEC, working on structured low-rank approximations. Her main research interests are in the fields of (multi)linear algebra, machine learning, data mining and optimization.

Phone: +32 (0)2 629 36 65 e-mail: [email protected] John Lataire was born in Brussels, Belgium, in 1983. He graduated as an Electrical Engineer in Electronics and Information Processing in July 2006. He received the degree of doctor in Engineering Sciences (Doctor in de Ingenieurswetenschappen) on March 24, 2011. Both degrees were obtained at the Vrije Universiteit Brussel, Belgium. From October 2007 till October 2011 he has been on a Ph.D. fellowship from the Research Foundation - Flanders (FWO). Since August 2006 he is working as a researcher at the department ELEC-VUB, Brussels, Belgium. His main interests are the measurement and identification of slowly time-varying, weakly nonlinear dynamic systems, formulated in the frequency domain. Phone: +32 (0)2 629 29 42 e-mail: [email protected] Ebrahim Louarroudi was born in Antwerp (Mortsel), Belgium, in 1985. He received the M.S. degree in electromechanical engineering from the Vrije Universiteit Brussel, Brussels, Belgium, in July 2009, where he has been working toward the Ph.D. degree in the Department of Fundamental Electricity and Instrumentation (ELEC) since October 2009. His main interests are the measurement and frequency-domain identification of weakly nonlinear periodically time-varying systems excited by periodic and arbitrary signals.

Phone: +32 (0)2 629 28.44 e-mail: [email protected] Anna Marconato was born in Trento, Italy, on April 8th, 1980. She received the B.Sc. in Mathematics and the M.Sc. in Telecommunication Engineering from the University of Trento, Italy, in 2002 and 2005, respectively. In 2009 she was awarded a joint PhD degree from the University of Trento, Italy, and the Vrije Universiteit Brussel, Belgium. From September 2009 she has been a postdoctoral researcher at Department ELEC, Vrije Universiteit Brussel, Belgium. Her main research interests are in the fields of nonlinear system identification and machine learning.

Phone: +32 (0)2 629 28.44 e-mail: [email protected] Ivan Markovsky obtained MS degree in Control and Systems Engineering from the Technical University of Sofia in July 1998 and Ph.D. degree in Electrical Engineering from the Katholieke Universiteit Leuven in February 2005. From January 2007 to September 2012 he was a lecturer at the School of Electronics and Computer Science of the University of Southampton. Since October 2012 he is with the department ELEC of the Vrije Universiteit Brussel. His current research interests are structured low-rank approximation, system identification, and data-driven control.

Phone: +32 (0)2 629 29 79 e-mail: [email protected]

13

Annual report ELEC 2012

David Oliva Uribe was born on May 24th 1975, Mexico City, Mexico. His research interests are in the field of system identification techniques for piezoelectric transducers, in particular the characterization of biological tissues using tactile sensors in medical applications. He graduated with Honours in Electronic and Communication Engineering in 1997 and obtained a Master in Sciences with Specialization in Manufacturing Systems in 2000, both from Tecnológico de Monterrey in Mexico City. From April 2007 to December 2010 he worked as Team Leader of the Research Group of Medical Engineering and Mechatronic Systems at the Institute of Dynamics and Vibrations Research from the Leibniz University of Hannover. Since January 2011, he joined the Department ELEC, Vrije Universiteit Brussel, where he is working toward a joint PhD degree. Phone: +32 (0)2 629 28.44 e-mail: [email protected] Rik Pintelon was born in Gent, Belgium, on December 4, 1959. He received a master’s degree in electrical engineering in 1982, a doctorate (PhD) in engineering in 1988, and the qualification to teach at university level (geaggregeerde voor het hoger onderwijs) in 1994 from the Vrije Universiteit Brussel (VUB), Brussels, Belgium. From 1982 to 1984 and 1986 to 2000, Dr. Pintelon was a researcher with the Belgian National Fund for Scientific Research (FWO-Vlaanderen) at the Electrical Engineering (ELEC) Department of the VUB. From 1984 to 1986 he did his military service overseas in Tunesia at the Institut National Agronomique de Tunis. From 1991 to 2000 he was a part-time lecturer at the department ELEC of the VUB, where he is currently a full-time professor in electrical engineering. His main research interests include system identification, signal processing, and measurement techniques. Dr. Pintelon is the coauthor of 4 books and the coauthor of about 200 articles in refereed international journals. He has been a Fellow of IEEE since 1998. Dr. Pintelon is the recipient of the 2012 IEEE Joseph F. Keithley Award in Instrumentation and Measurement (IEEE Technical Field Award). Phone: +32 (0)2 629 29 44 e-mail: [email protected] Koen Tiels was born in Halle (Belgium) in 1987. He graduated as an Electrotechnical-Mechanical Engineer in July 2010 at the Vrije Universiteit Brussel. In September 2010 he joined the ELEC department as a PhD student. His main interests are in the field of nonlinear block structured system identification.

Phone: +32 (0)2 629 36 65 e-mail: [email protected] Konstantin Usevich was born on 21 April 1986 in St. Petersburg, Russia (at that time Leningrad, USSR). He received a Master’s degree in Applied Mathematics and Computer Science in 2007 from the Department of Statistical Modelling, Faculty of Mathematics and Mechanics, Saint-Petersburg State University, Russia. In February 2011 he was awarded a PhD degree (kandidat fiz.-mat. nauk) in Applied Mathematics and Computer Science by the same department. Konstantin was employed as a Postdoctoral Research Fellow by the School of Electronics and Computer Science, University of Southampton, UK, in September 2011. In October 2012 he joined the Department of Fundamental Electricity and Instrumentation, Vrije Universiteit Brussel, Belgium, as a Postdoctoral Research Fellow. Konstantin’s research interests include low-rank approximation of structured matrices and its applications to time series analysis, image processing, system identification and computer algebra. Phone: +32 (0)2 629 29 79 Fax: [email protected] Laurent Vanbeylen was born in Elsene, Belgium, on January 6, 1983. He graduated as an Electrical Engineer in Electronics and Information Processing in 2005 at the Vrije Universiteit Brussel. Laurent pursued a PhD since August of the same year, at the department ELEC, and obtained the degree in 2011 with his thesis entitled "Nonlinear dynamic systems: blind identification of block-oriented models, and instability under random inputs". Presently, he is active at the ELEC department as a doctor-assistant. His main interests remain in the field of nonlinear dynamic systems, with the emphasis on the identification of nonlinear feedback models.

Phone: +32 (0)2 629 29 79 e-mail: [email protected]

14

Introduction to the department ELEC

Anne Van Mulders was born in Jette on September 19, 1984. In July 2007, she received the degree of Mechanical Engineering from the Free University of Brussels (VUB). She joined the department of ELEC in September 2007 and obtained a PhD in 2012 with her thesis entitled "Tackling two drawbacks of polynomial nonlinear state-space models". Her main research interests are in the field of nonlinear system identification.

Phone: +32 (0)2 629 29 79 e-mail: [email protected]

Dhammika Widanage was born on November 3rd 1983, Kuwait City, Kuwait. Dhammika Widanage’s research interests are in nonlinear system identification, signal processing and control, with applications in lithium-ion battery modelling, mechatronic and electric vehicles. He is a postdoctoral researcher for two departments, the Department of Electrotechnical Engineering and Energy Technology and the Department of Fundamental Electricity and Instrumentation at the Vrije Universiteit Brussel. He was awarded a Ph.D. in 2008 for research done at the Stochastic and Complex Systems Laboratory Group from the University of Warwick, UK and graduated with a FirstClass Honours in Electronic and Communication Engineering (BEng - Bachelor of Engineering) in 2004 from the University of Warwick, UK.

Phone: +32 (0)2 629 29 46 e-mail: [email protected]

1.4.4

Team C: Applied Signal Processing for Engineering (ASPE)

Gerd Vandersteen (coordinator of team C) was born in Belgium in 1968 and received the degree in electrical engineering from the Vrije Universiteit Brussel (VUB), Brussels, Belgium, in 1991. In 1997, he received his PhD in electrical engineering, entitled “Identification of Linear and Nonlinear Systems in an Errors-in-Variables Least Squares and Total Least Squares Framework”, from the Vrije Universiteit Brussel/ ELEC. During his postdoc, he worked at the micro-electronics research centre IMEC as Principal Scientist in the Wireless Group with the focus on modeling, measurement and simulation of electronic circuits in state-of-the-art silicon technologies. This research was in the context of a collaboration with the Vrije Universiteit Brussels. From 2008 on, he is working as Prof. at the Vrije Universiteit Brussels/ELEC within the context of measuring, modeling and analysis of complex linear and nonlinear system. Within this context, the set of systems under consideration is extended from micro-electronic circuits towards to all kinds of electro-mechanical systems.

Phone: +32 (0)2 629 29 44 e-mail: [email protected] Matthias Caenepeel was born in Brasschaat, Belgium on August 2, 1989. He received the degree of Electrical Engineer in Electronics and Information Processing in July 2012 from the Vrije Universiteit Brussel, Brussels, Belgium. In September 2012 he joined the department ELEC as a PhD student. His main interests are RF design and system modelling.

Phone: +32 (0)2 629 28 44 e-mail: [email protected]

15

Annual report ELEC 2012

Adam Cooman was born in Jette (Belgium) in 1989. He graduated as an Electrical Engineer in Electronics and Information Processing in July 2012 at Vrije Universiteit Brussel (VUB). In August 2012 he joined the department ELEC as a PhD student. His main interests are the design of analog/RF circuits and nonlinear modelling.

Phone: +32 (0)2 629 28 44 e-mail: [email protected] Hannes Maes was born in Hasselt (Belgium) on May 17, 1990. He received the degree of Electrical Engineer in Electronics and Information Processing in July 2012 from the Vrije Universiteit Brussel, Brussels, Belgium. In September 2012 he joined the department ELEC as a PhD student. His main interests are system modeling for biomedical applications and non linear modeling.

Phone:+32 (0)2 629 36 65 e-mail: [email protected] Griet Monteyne was born in Vilvoorde, Belgium, on January 19, 1985. She graduated as an Electrotechnical-Mechanical Engineer (Burgerlijk Ingenieur) in July 2007 at the Vrije Universiteit Brussel (VUB), Brussels, Belgium. In October 2007 she joined the ELEC department of the VUB as a PhD student. Until October 2009 her research was in collaboration with the department ANS (Advanced Nuclear Systems) of the SCK-CEN (Belgian Nuclear Research Centre) in Mol, Belgium. Currently her work is related to geothermal heat pumps. The object is to model the thermal dynamic behavior of the surrounding geology by use of noise analysis techniques.

Phone: +32 (0)2 629 29 49 e-mail: [email protected] Yves Rolain (1961, Belgium) received the Electrical Engineering (Burgerlijk Ingenieur) degree in July 1984, the degree of computer sciences in 1986, and the PhD degree in applied sciences in 1993, all from the Vrije Universiteit Brussel (VUB), Brussels, Belgium. He is currently a research professor at the VUB in the department ELEC. His main interests are microwave measurements and modelling, applied digital signal processing and parameter estimation / system identification.

Phone: +32 (0)2 629 29 44 e-mail: [email protected] Maarten Schoukens was born in Jette (Belgium) in 1987. He graduated as an Electrical Engineer in Electronics and Information Processing (profile Measurement and Modelling of Dynamic Systems) in July 2010 at the Vrije Universiteit Brussel. In September 2010 he joined the department ELEC as a PhD student. His main interests are in the field of nonlinear block oriented system identification of multiport microwave systems.

Phone: +32 (0)2 629 28.44 e-mail: [email protected]

16

Introduction to the department ELEC

Diana Ugryumova was born in Kiev, Ukraine, on October 26th, 1984. She received the degree in Applied Mathematics (chair of Mathematical Theory of Systems and Control) at the University of Twente in the Netherlands in March 2010. Diana joined the department of ELEC in May 2010 as a PhD student. Her current research is about identification of distillation columns. The aim of this research is to enhance the performance of a distillation column through a better modeling and control strategy.

Phone: +32 (0)2 629 29 49 e-mail: [email protected]

1.4.5

Team D: Medical Measurements and Signal Analysis (M2ESA)

Wendy Van Moer (coordinator Team D), (12/07/1974), received the Engineer and Ph. D. degrees in Engineering from the Vrije Universiteit Brussel (VUB), Brussels, Belgium, in 1997 and 2001, respectively. She is currently a part time lecturer with the Electrical Measurement Department (ELEC), VUB and a post-doctoral researcher of the Research Foundation-Flanders (FWO). Her main research interests are nonlinear measurement and modeling techniques for medical and highfrequency applications. She was the recipient of the 2006 Outstanding Young Engineer Award from the IEEE Instrumentation and Measurement. Since 2007 she has been an associate editor for the IEEE Transactions on Instrumentation and Measurement and in 2010 she became an associate editor of the IEEE Transactions on Micro wave Theory and Techniques. Phone: +32 (0)2 629 28 68 e-mail: [email protected] Kurt Barbé received the M.Sc. degree in mathematics (option Statistics) and the Ph.D. degree in electrical engineering from the Vrije Universiteit Brussel (VUB), Brussels, Belgium, in 2005 and 2009, respectively. He is currently a Postdoctoral Research Fellow with the Flemish Research Foundation (FWO). At the VUB, he is with the Department of Fundamental Electricity and Instrumentation (ELEC) and a member of the research team Medical Measurements and Signal Analysis (M2ESA). His research interests are system identification, time series analysis, digital and analogue signal processing, statistical methods, finite and short sample measurement analysis and biomedical applications. Dr. Barbé has been an Associate Editor for the IEEE Transactions on Instrumentation and Measurement since 2010 and he is the recipient of the ‘Outstanding Young Engineer Award 2011’ of the IEEE Instrumentation and Measurement society. Phone: +32 (0)2 629 29 46 e-mail: [email protected] Lee Gonzales Fuentes was born in Arequipa, Perú in March 31, 1985. She received the Bachelor degree in Electronic Engineering with a major in Telecommunications from the Universidad Nacional de San Agustín (UNSA), Arequipa, Perú in November 2008 and the Master degree in Electronics/Telecommunications from the University of Gävle (HIG), Gävle, Sweden in January 2012. Lee joined the department of ELEC in March 2012 as a Ph.D. student in the domain of pre- and postprocessing of measured data. Her main research interests are measurements and modeling techniques for high-frequency applications, wireless network design, telemetry and digital signal processing. In May 2012, Lee received the 2012 Graduate Fellowship Award from the IEEE Instrumentation and Measurement Society. Phone: +32 (0)2 629 29 53 e-mail: [email protected]

17

Annual report ELEC 2012

Lieve Lauwers was born in Jette, Belgium, on March 15, 1982. She received the degree of Electrical Engineering (option Photonics) in 2005 from the Vrije Universiteit Brussel (VUB), Brussels, Belgium. In August 2005 she joined the department ELEC of the VUB as a PhD student. In May 2011, she received the PhD degree in Electrical Engineering. Currently, she is working as a post-doctoral researcher of the FWO (Research Foundation Flanders). Her main interests are in the field of modeling techniques for biomedical applications.

Phone: +32 (0)2 629 29 53 e-mail: [email protected] Oscar J. Olarte Rodríguez was born in Colombia, March 31, 1980. In June 2004 he received the degree of Electronic Engineer from the Universidad Industrial de Santander (UIS). In September 2007, he obtained in the same University his Master degree in Electronic Engineer. In November he joined department ELEC at the Vrije Universiteit Brussel (VUB) for his PhD. His main research interests are in the field of signal processing and time series.

Phone: +32 (0)2 629 29 79 e-mail: [email protected]

1.4.6

Technical and Administrative Staff

Wim Delcourte Wim Delcourte was born in 1962 (Belgium). He graduated as an Industrial Engineer (Industrieel Ingenieur) in 1987. Since May 1989 he is with the department ELEC of the Vrije Universiteit Brussel (VUB full time tenure). Actualy he is responsible for the design of electronic measurement instruments for research and education, for the repair and maintenance of complex measurement instruments and computers (20%). He also takes care of the management of the computerrooms of the faculty (80%).

Phone: +32 (0)2 629 29 52/29 14 e-mail: [email protected] Bea Huygen was born in Uccle (Belgium) on May 30, 1970. She graduated as beauty consultant and runned a beauty center for 12 years. The period 2008-2010 she worked as a receptionist in the medical sector. In april 2011 she joined the department ELEC, as a part-time secretary. She is mainly responsible for the general secretariat and the financial administration of the department.

Phone: +32-(0)2-629.29.47 e-mail: [email protected]

18

Introduction to the department ELEC

Johan Pattyn was born in 1974 (Belgium). After 2 years of engineering studies he joined the navy to become a radar technician. Since december 2009 he is with the department ELEC of the Vrije Universiteit Brussel (full time tenure). He is responsible for Rapid PCB Prototyping and also is involved in the maintenance and repair of the instruments and circuits for the students lab's.

Phone: +32 (0)2 629 29 52 e-mail: [email protected]

Ann Pintelon was born in October 1963 (Belgium). In 1985 she received the Ms degree in Physical Education (VUB). Since 1990 she has been with the department ELEC at the Vrije Universiteit Brussel on a GOA contract (Measurement, Modelling and Identification of Dynamic Systems), between 2008 and 2011 she has been working on the contract “Methusalem”: Centre for Data Based Modelling and Model Quality Assessment (4/5 tenure), and since March 2011 a 90% tenure at the university. She is mainly responsible for the scientific reports of the department ELEC (annual report, web-pages ELEC, …); the management of the infocentre (library, publication list, …); the administrative organisation of conferences, workshops and doctoral school; hosting of visiting researchers.

Phone: +32 (0)2 629 27 67 e-mail: [email protected]

Sven Reyniers was born in 1973 (Belgium). He has several years of ICT - experience in multinationals, including several international missions (France, Germany). Since 2006, he is working at the Vrije Universiteit Brussel (full time tenure) at the department ELEC. As system and network administrator he is responsible for the health and availability of the department servers and network.

Phone: ++32 (0)2 629 29 52/29 45 e-mail: [email protected]

1.4.7

Professors Emeriti

Alain Barel was born in Roeselare, Belgium, on July 27, 1946. He received the degree in Electrical Engineering from the Université Libre de Bruxelles, Belgium, in 1969, the Postgraduate degree in telecommunications from Rijks Universiteit Gent (State University of Gent), Belgium, in 1974, and the doctor of applied science from the Vrije Universiteit Brussel in 1976. He worked as assistant and Lecturer at the Vrije Universiteit Brussel (VUB). From 2006 until September 2011 he has been 10% active at the department as Professor emeritus and teaches microwaves, and until September 2012 been active as a voluntary researcher at the dept. ELEC.

e-mail: [email protected]

19

Annual report ELEC 2012

Michel Gevers was born in Antwerp, Belgium, in 1945. He obtained an Electrical Engineering degree from the Université Catholique de Louvain, Belgium, in 1968, and a Ph.D. degree from Stanford University, California, in 1972. He holds a Honorary Degree from the Vrije Universiteit Brussel and the University of Linköping, Sweden, and a few other titles. He has been President of the European Union Control Association (EUCA) from 1997 to 1999, and Vice President of the IEEE Control Systems Society in 2000 and 2001. Between 1990 and 2010 he has been the coordinator of the Belgian Interuniversity Network DYSCO (Dynamical Systems, Control, and Optimization) funded by the Federal Ministry of Science. His research interests are in system identification and its interconnection with robust control design, optimal experiment design, data-based control design, optimal control and filtering, and realization theory. He has published about 250 papers and conference papers, and two books: "Adaptive Optimal Control - The Thinking Man's GPC", by R.R. Bitmead, M. Gevers and V. Wertz (Prentice Hall, 1990), and "Parametrizations in Control, Estimation and Filtering Problems: Accuracy Aspects", by M. Gevers and G. Li (Springer-Verlag, 1993). Until September 2012 he was 20% active at the department ELEC as Professor emeritus, and now still active as a voluntary researcher at the dept. ELEC. e-mail: [email protected]

1.4.8

In memoriam prof. em. Dr. Ronald Van Loon (1940-2012) Ronny Van Loon. Born in Antwerp, 1940, he obtained a degree in Physics at the ULB, and a PhD in Science at the VUB. He joined the VUB in 1970 as assistant, then Professor at the Faculty of Applied Science. In parallel with the teaching activities, he had research and logistic activities as a medical physicist at the university hospital AZ-VUB: from 1982 on in the Radiotherapy department focusing on EC granted projects in Hyperthermia, from 1989 on in the Radiology department involved in dosimetry and quality assurance. He directed the subgroup QUARAD (“Quality in Radiology”), a team involved in dosimetry, radiation protection and quality improvement in radiology. This group is acting in Belgium as reference centre for the Quality Assurance of the technical aspects of breast cancer screening. Prof. Van Loon was the Past-President of the Belgian Hospital Physicist Association, and member of the Board of the “Federal Agency of Nuclear Control” and the “Belgian Society of Radioprotection”. He was delegate of the VUB in the VLIR-cooperation and Development Cell, and coordinates a cooperation project in Hanoi (Vietnam) and in VLIR International Training programme on Medical Physics.

Until September 2005, he was still 10% active at the department ELEC as professor emeritus. He also contribiuted to the IAEA's (International Atomic Energy Agency) teaching programs on radiation protection in medical applications of ionizing radiation. He was scientific advisor at the “Belgian Museum of Radiology”, Brussels, and he was active as a voluntary researcher at the dept. ELEC. In October 2008, he received the title of “Doctor Honoris Causa” from the Hanoi University of Technology. On 20 December 2012, prof. dr. em. Ronald Van Loon suddenly passed away. We will remember him as a warm friendly and very wise colleague, who remained actively involved at the faculty of Engineering and the university. He always defended his opinion and his philosophical beliefs. He was loyal to the basic philosophy of the Vrije Universiteit Brussel: the principle of “free inquiry”, based on a text of Henri Poincaré, and the principle that the institution must be managed according to the model of democracy.

20

Introduction to the department ELEC

1.4.9

List of phone numbers and e-mail addresses (Status 01/01/2013)

NAME

PHONE

E-MAIL

K. BARBÉ M. CAENEPEEL A. COOMAN P. CSURCSIA A. DE COCK W. DELCOURTE E. GEERARDYN L. GONZALES FUENTES J. GOOS B. HUYGEN M. ISHTEVA J. LATAIRE L. LAUWERS E. LOUARROUDI H. MAES A. MARCONATO I. MARKOVSKY G. MONTEYNE O. OLARTE RODRÍGUEZ D. OLIVA URIBE J. PATTYN A. PINTELON R. PINTELON S. REYNIERS Y. ROLAIN J. SCHOUKENS M. SCHOUKENS K. TIELS D. UGRYUMOVA K. USEVICH L. VANBEYLEN L. VAN BIESEN G. VANDERSTEEN W. VAN MOER A. VAN MULDERS D. WIDANAGE E. ZHANG

+32(0)2 629 29 46 32(0)2 629 28.44 32(0)2 629 28.44 +32(0)2 629 29 46 +32(0)2 629 36 65 +32(0)2 629 29 52/29 14 +32(0)2 629 28 44 +32(0)2 629 29 53 +32(0)2 629 29 46 +32-(0)2-629.29.47 +32(0)2 629 36 65 +32(0)2 629 29 42 +32(0)2 629 29 53 +32(0)2 629 28.44 +32(0)2 629 36 65 +32(0)2 629 28.44 +32(0)2 629 29 42 +32(0)2 629 29 49 +32(0)2 629 29 79 +32(0)2 629 28.44 +32(0)2 629 29 52 +32(0)2 629 27 67 +32(0)2 629 29 44 +32(0)2 629 29 52 +32(0)2 629 29 44 +32(0)2 629 29 44 +32(0)2 629 28.44 +32(0)2 629 36 65 +32(0)2 629 29 49 +32(0)2 629 29 79 +32(0)2 629 29 79 +32(0)2 629 29 43 +32(0)2 629 29 44 +32(0)2 629 28 68 +32(0)2 629 29 79 +32(0)2 629 29 46 +32(0)2 629 28 44

[email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected]

21

Annual report ELEC 2012

1.4.10

Industrial Partnership Dr. ir. Alain GEENS, BelV ir. Frank LOUAGE, MSc Zobeida Cisneros Barros; Address Systems N.V. Dr. ir. Luc PEIRLINCKX, Phonetics Topographics, Belgium Lic. Marc PERSOONS, Tresco Navigation Systems ir. Serge TEMMERMAN, SEBA service NV (measuring instruments for telecom. cables) Dr. Frank UYTDENHOUWEN, Banama-Telecom Dr. ir. Marc VANDEN BOSSCHE, Dr. ir. Frans Verbeyst; NMDG Engineering bvba

1.4.11

National or international contacts

1.4.11.1 Visiting professors/researchers Ibrahim ALJAMAAN, University of Calgary, Canada, 19/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Khaled ALJANAIDEH, University of Michigan, USA, 19/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Amjad Hisham ABU-RMILEH, Universidad de Girona, Spain: 04/01/2012 – 30/03/2012 and 01/05/2012 – 30/06/2012: research in the frame of PhD “Frequency domain identification and measurement techniques for the development of artificial pancreas” Lee Barford, Agilent Measurement Research Laboratory, USA: 01/03/2012, presentation at ELEC seminar “Multi- and Many-Core Processing in Embedded and Signal Processing Systems” Niclas BJÖRSELL, University of Gävle, Center for RF Measurement Technology, Sweden: 15/10/2012 –…. Guestprofessor. Frank BOEREN, Eindhoven University of Technology, The Netherlands, 20/04/2012 – 16/05/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Joost BOLDER, Eindhoven University of Technology, The Netherlands, 20/04/2012 – 16/05/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Marco CAMPI, University of Brescia, Dept. of Electronics for Automation, Italy: 07/03/2012 – 09/03/2012. Micro symposium on system identification

22

Introduction to the department ELEC

Paolo Carbone, University of Perugia, Italy: 15/10/2012 – 15/12/2012. Presentation at ELEC seminar “Positioning, Synchronizing and other Amenities” Sebastian Yuri Cavalcanti CATUNDA, Universidade Federal do Rio Grande do Norte, Natal, Brazil, 30/11/12: identification for waste water treatment Tadeusz DOBROWIECKI, Budapest University of Technology and Economics, Department of Measurement and Information Systems, Budapest, Hungary:16/03/2012 – 20/03/2012. Carl Niklas EVERITT, Kungliga Tekniska Högskola, Sweden, 20/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Tomás García Sánchez, Polytechnic University of Catalonia, Spain, 19/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Kristóf GÁTI, Budapest University of Technology and Economics, Hungary, 20/04/2012 – 16/05/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Keith GODFREY, University of Warwick, School of Engineering: 29/10/2012 – 31/10/2012: cotutelle agreement: meeting with Hin Kwan Wong and Johan Schoukens Luka HAFNER, University of Ljubljana, Faculty of Mechanical Engineering, Slovenia: 01/09/2012 – 17/12/2012. IAESTE the International Association for the Exchange of Students for Technical Experience Jairo Andrés HERNÁNDEZ NARANJO, University of Ghent, Belgium, 21/05/2012 – 15/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Håkan HJALMARSSON, Royal Institute of Technology, Dept. Signals, Sensors, Systems (control group), Sweden: 07/03/2012 – 09/03/2012. Micro symposium on system identification Mariya ISHTEVA, Georgia Institute of Technology, USA: 14/10/2012 – 19-10/2012. Presentation at ELEC seminar “Recent matrix and tensor decompositions in data mining and machine learning” Rafael Karrer, Technische Universität Wien, Austria: 24/08/2012 – 05/10/2012. IAESTE the International Association for the Exchange of Students for Technical Experience. Sandor KOLUMBAN, Budapest University of Technology and Economics, Department of Measurement and Information Systems, Budapest, Hungary: 17/09/2012 – 16/10/2012: research in the frame of PhD

23

Annual report ELEC 2012

Róbert Ku era, Slovenská technická univerzita v Bratislave, Slovakia: 03/09/2012 – 21/12/2012. IAESTE the International Association for the Exchange of Students for Technical Experience Per LANDIN, University of Gävle, Center for RF Measurement Technology, Sweden: 01/03/2012 – 31/03/2012. PhD student, Cotutelle agreement. Lennart LJUNG, Linköping University, Sweden, 07/03/2012 – 09/03/2012. Micro symposium on system identification Zimo LU, Tianjin university, China, 20/04/2012 – 16/05/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Mikaya LUMORI, University of San Diego, Electrical Engineering, San Diego, CA, USA, 08/06/2012 – 24/08/2012 Sabbatical on System Identification. Raúl MACÍAS MACÍAS, Polytechnic University of Catalonia, Spain, 19/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Charles NADER, University of Gävle, Centre for RF Measurement Technology, Sweden: 04/01/2012 – 31/03/2012. PhD student, Cotutelle agreement. Brett NINNESS, University of Newcastle, Australia: 26/11/2012 – 27/11/2012. Presentation at ELEC seminar “Control, Identification and Cheap Computing” Tomas NORDSTRÖM, The Telecommunications Research Center Vienna, TU Wien, Austria, 18/03/2012-19/03/2012 Sadegh RAHROVANI, Chalmers University of Technology, Sweden, 20/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Manoel FLORES DA CUNHA, Universidade Federal do Rio Grande do Sul, Brazil, 20/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Ghosh RAMKRISHNA, Åbo Akademi University, Finland, 19/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Daniel E. RIVERA, Arizona State University, Tempe, USA: 18/10/2012 – 19/10/2012 Benjamin SANCHEZ, Universidad Politechnica de Catalunya (UPC), Barcelona, Spain: 13/02/2012 – 17/02/2012 and 17/12/2012 – 21/12/2012, Research on System Identification

24

Introduction to the department ELEC

Thomas SCHÖN, Linkoping university, Sweden: 22/05/2012 – 23/05/2012 and 05/06/2012 – 10/06/2012: presentation at ELEC seminar “Computational inference in dynamical system” Kushagra SINGHAL, IIT Kanpur, India: 10/05/2012 – 10/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Yujiao SONG, Chalmers University of Technology, Sweden, 20/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Christian SPROCK, University of Paderborn, Germany: 21/05/2012 – 15/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Ralph STROOP, Klinik für Neurochirurgie, Ruhr-Universität Bochum, Germany: 22/04/2012 – 25/04/2012 and 17/09/2012 – 19/09/2012: Piezoelectric Tactile Tissue Differentiation Sensor System: measurements in the frame of joint PhD University of Hannover – Vrije Universiteit Brussel; David Uribe Patricio VALENZUELA, Kungliga Tekniska Högskola, Sweden, 20/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems Mattijs VAN BERKEL, Technische Universiteit Eindhoven, The Netherlands: 30/11/2012. Research on Microwaves Paul VAN DEN HOF, Delft University of Technology, Signals, Systems and Control Group, The Netherlands: 07/03/2012 – 09/03/2012. Micro symposium on system identification Lajos VARGA, Budapest University of Technology and Economics, Hungary: 19/05/2012 – 16/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems David WESTWICK, Department of Electrical and Computer Engineering, University of Calgary, Canada: 26/06/2012-10/07/2012. Identification of nonlinear systems using block oriented models Hin Kwan WONG, University of Warwick, School of Engineering, UK: 12/03/2012 – 05/04/2012 and 26/09/2012 – 08/11/2012. PhD student, Cotutelle agreement Jin YAN, University of Michigan, USA, 18/05/2012 – 14/06/2012. Doctoral School VUB – Dept. ELEC: Identification of Nonlinear Dynamic Systems

25

Annual report ELEC 2012

1.4.11.2 Scientific missions Kurt BARBÉ 13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Efficient use of short data records for FRF modeling by using fractional poles”

18/05/2012

19/05/2012

MeMeA 2012 IEEE Symposium on Medical Measurements and Applications, Budapest, Hungary, May 18-19, 2012. Plenary speaker “An innovative oscillometric blood pressure measurement: getting rid of the traditional envelope”

16/07/2012

18/07/2012

Invited lecturer Summer School on “Distributed Data Acquisition System”, University of Calabria, Consenza, Italy.

05/11/2012

23/12/2012

Visiting professor at the Dipartimento di Elettronica, Informatica e Sistemistica of the University of Calabria, Consenza, Italy (Prof. Domenico Grimalidi)

19/09/2012

21/09/2012

SMACD 2012 International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, Seville, Spain, 19-21 September 2012. Presentation of paper “Determining the Dominant Nonlinear Contributions in a multistage Op-amp in a Feedback Configuration”

08/10/2012

08/10/2012

Presentation of poster at the DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, Belgium, October 8, 2012 “Determining the Dominant Nonlinear Contributions in a multistage Op-amp in a Feedback Configuration”

Adam COOMAN

Péter Zoltán CSURCSIA 13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Identification of Time-varying Systems using a Two-dimensional B-spline Algorithm”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012.

Egon GEERARDYN 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “How to obtain a broad band FRF with constant uncertainty?”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Quasilogarithmic Multisine Excitations for Broad Frequency Band Measurements”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012.

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Quasi-logarithmic Multisines for Broad Frequency Band Measurements”

1/10/2012

30/11/2012

Visiting PhD Researcher at the Technische Universiteit Eindhoven, CST, Control Systems Technology, collaboration with Assistant Prof. Tom Oomen. Topic: Leveraging the LPM for H infinity robust control.

1/11/2012

2/11/2012

DCT 24 hour meeting 2012, Dynamics and Control Technology, 1-2 November 2012, Deurne, The Netherlands. Presentation of poster “Quasi-logarithmic Multisines for Broad Frequency Band Measurements”

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “The Transient Impulse Response Modeling Method and the Local Polynomial Method for Nonparametric System Identification”

Michel GEVERS 11/07/2012

Lee GONZALES FUENTES 13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Cognitive Radios: Discriminant Analysis Finds the Free Space”

26

Introduction to the department ELEC

Jan GOOS 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “State Space Identification for Linear ParameterVarying Systems”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012.

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Order selection of LPV State Space models using a subspace method in the frequency domain”

Sandor KOLUMBAN 23/09/2012

26/09/2012

Participating at ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands.

03/2012

07/2012

Visiting Post Doc Researcher at the Technische Universiteit Delft, 3mE, Dutch Centre for Systems and Control, collaboration with Assistant Prof. Roland Tóth. Topic: Applying machine learning techniques to system identification, formulated in the frequency domain.

27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Detecting the time variation in an assumed linear, time invariant measurement”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Non-Parametric Best Linear Time Invariant Approximation of a Linear Time-Varying System”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012.

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Dealing with Correlated Errors in Least-Squares Support Vector Machine Estimators.”

08/10/2012

08/10/2012

Presentation of poster at the DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, Belgium, October 8, 2012 “Accuracy-complexity trade-off of some nonlinear identification methods: benchmark case studies”

John LATAIRE

Lieve LAUWERS 13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “A robust signal detection method for fMRI data under correct Rice conditions”

13/05/2012

16/05/2012

Participating at I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012.

Ebrahim LOUARROUDI 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Nonparametric Identification of Linear Periodically Time-Varying Systems Using Arbitrary Inputs”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Nonparametric Estimation of the Instantaneous Transfer Function of Linear Periodically Time-Varying Systems Excited by Arbitrary Signals”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012.

Anna MARCONATO 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Nonlinear block-oriented identification for insulinglucose models”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Separate Initialization of Dynamics and Nonlinearities in Nonlinear State-Space Models”

27

Annual report ELEC 2012

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Identification of the Silverbox Benchmark Using Nonlinear State-Space Models”

20/09/2012

20/09/2012

KU Leuven Seminars on Optimization in Engineering, KULeuven, Belgium. Presentation “Combining system identification and statistical learning in nonlinear data-driven modeling”

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Accuracy-complexity trade-off of some nonlinear identification methods: benchmark case studies”

08/10/2012

08/10/2012

Participating at the DYSCO IAP study day, Château-Ferme de Profondval, Court-StEtienne, Belgium, October 8, 2012

Griet MONTEYNE 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Robust or fast local polynomial method: How to choose?”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Transient Suppression in Non-Parametric Frequency Response Function Estimates of Heat Diffusion Phenomena”

21/05/2012

20/06/2012

Chalmers University of Technology Building Services Engineering, Department of Energy and Environment, Chalmers University of Technology, Gothenburg: Measurement campaign at Chalmers University of Technology. This University has a unique setup for measuring the thermal response of boreholes. With these measurement the developed theoretical results are verified in practice.

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Extension of Local Polynomial Method for Periodic Excitations”

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Frequency Domain Modeling of Heat Transport around Borehole Heat Exchangers”

08/10/2012

08/10/2012

Participating at the DYSCO IAP study day, Château-Ferme de Profondval, Court-StEtienne, Belgium, October 8, 2012

08/02/2012

11/02/2012

5th International Conference on Advanced Technologies & Treatments for Diabetes (ATTD), Barcelona, February 8th and 11th, 2012. Poster Presentation : “Combining Broad-Band Multisines Excitations and Dielectric Spectroscopy for Non-Invasive Glucose Measurements”

27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Dielectric Spectroscopy for Non-invasive Glucose Measurements”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Using the Best Linear Approximation as a First Step to a New Non-Invasive Glucose Measurement”

08/07/2012

12/07/2012

Universidad Politécnica de Valencia (UPV), Electrical Engineering Department. In this scientific mission was studied the feasibility to develop an electronic odor system to detect glucose. The setup, composed by of 32 metal semiconductors sensors operating at different temperatures, was developed by the UPV (Spain) and the University of Gavle (Sweden).

27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Chairman session System Identification

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “The Best Linear Approximation of Nonlinear Systems Operating in Feedback”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Errors-In-Variables Identification of Linear Dynamic Systems Using Periodic Excitations” and member of the local organizing committee

08/10/2012

08/10/2012

Participating at the DYSCO IAP study day, Château-Ferme de Profondval, Court-StEtienne, Belgium, October 8, 2012

Oscar OLARTE

Rik PINTELON

28

Introduction to the department ELEC

Yves ROLAIN 17/06/2012

22/06/2012

79th ARFTG Microwave Measurement Conference, Non-Linear Measurement Systems, Convention Center, Montréal, Canada, June 22, 2012. Presentation of paper “Synchronizing modulated NVNA measurements on a dense spectral grid”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Member of the local organizing committee

29/10/2012

01/11/2012

Workshop organized at the 42nd European Microwave Conference (W09 (EuMC/EuMIC), 29 October - 1 November 2012, Amsterdam, The Netherlands. Presentation at workshop “Why professional LSNA, NVNA, scope and VSA users care about accurate phase information”

Johan SCHOUKENS 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Chairman session System Identification and member of organizing committee.

26/04/2012

26/04/2012

Visit Bo Wahlberg, KTH-Stockholm. Opponent Per Hägg, Lic. Thesis

02/05/2012

02/05/2012

Technical University Delft. Lecture on Biomedical Workshop NeuroSIPE (Frans van der Helm, Alfred Schouten). Lecture on System Identification in the presence of nonlinear distortions.

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Study of the maximal interpolation errors of the local polynomial method for frequency response function measurements”

01/06/2012

04/06/2012

Visit to the Technical University Budapest (Prof. T. Dobrowiecki, Prof. I. Kollar,Prof. I. Vajc), discussion joined PhD Sandor Columban

14/06/2012

15/06/2012

Uppsala University. Lecture on the occasion of the retirement of prof. Torsten Soderstrom: “Nonlinear system identification in an Errors-in-Variables-Setting: a good idea?”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “User Choices for Nonparametric Preprocessing in System Identification” and general chair of the local organizing committee.

19/11/2012

19/11/2012

Visit to Technical University Eindhoven. Discussion with Tom Oomen and Eegon Geerardyn

Maarten SCHOUKENS 20/02/2012

04/05/2012

Research stay at Linköping University, Division of Automatic Control, Department of Electrical Engineering Host: prof. Martin Enqvist

15/03/2012

15/03/2012

Seminar at Division Automatic Control, Department of Electrical Engineering, Linköping University, Sweden. Presentation: “Identification of Wiener-Hammerstein systems using the best linear approximation”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Parallel Wiener identification starting from linearized models”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Identification of Hammerstein-Wiener Systems”

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Combining the best linear approximation and dimension reduction to identify the linear blocks of parallel Wiener systems”

08/10/2012

08/10/2012

Presentation of poster at the DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, Belgium, October 8, 2012 “Combining the best linear approximation and dimension reduction to identify the linear blocks of parallel Wiener systems”

29

Annual report ELEC 2012

Koen TIELS 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Parameter reduction of SISO Wiener-Schetzen models”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Parameter Reduction of MISO Wiener-Schetzen Models Using the Best Linear Approximation”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Reducing the Number of Parameters in a Wiener-Schetzen Model”

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Iterative update of the pole locations in a Wiener-Schetzen model”

Diana UGRYUMOVA 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Identification and modeling of distillation columns from transient response data”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Identification and Modeling of Distillation Columns From Transient Response Data”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Polynomial approximation errors for integrator frequency response functions”

27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Piezoelectric Tactile Tissue Differentiation Sensor System: Concepts and Measurement Challenges”

11/07/2012

13/07/2012

Participating at the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012.

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Tissue Differentiation Sensor System For Brain Tumour Resection Based On Multisine Excitation ”

David URIBE

Konstantin USEVICH 02/10/2012

02/10/2012

Operational Research and Statistics Seminars 2012-2013, Cardiff School of Mathematics. Lecture: “Mosaic Hankel structured low-rank approximation with variable projection.”

Leo VAN BIESEN 08/02/2012

10/02/2012

Braunschweig, Germany: attending IMEKO Advisory Board and Technical Board on invitation of Prof. Manfred Peters, vice-president of the Physikalisch Technische Bundesanstalt.

16/05/2012

20/05/2012

International Conv. 2012, Salons de l’Aveyron, Paris, France. Attending as Belgium representative.

18/08/2012

23/08/2012

Mission on behalf of VLIR as promoter of “Eigen Initiatief project Vrije Universiteit Brussel with Institut Supérieur Des Techniques Appliquées (ista)”, Kinshsa, Congo. Visit at Minister for Education and ISTA

06/09/2012

16/09/2012

Attending IMEKO Advisory Board and Technical Board Busan, South Korea and participation with 2 papers at the IMEKO XX World Congress

01/12/2012

04/12/2012

Meeting with President and Secretary-General of IMEKO at Imeko Secretariat, Budapest, Hungary

30

Introduction to the department ELEC

Wendy VAN MOER 27/03/2012

29/03/2012

Participating at the 31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 27-29, 2012

01/05/2012

30/06/2012

HiG, Gavle, Zweden as visiting Professor: “Measurement techniques for characterizing the nonlinear behavior of microwave systems: a comparison”

13/05/2012

16/05/2012

Participating at I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012.

18/05/2012

19/05/2012

MeMeA 2012 IEEE Symposium on Medical Measurements and Applications, Budapest, Hungary, May 18-19, 2012. Presentation of paper “Saving lives by integrating cognitive radios into ambulances”

10/10/2012

13/10/2012

Attending the IEEE Instrumentation and Measurement Society ADCOM meeting

1/12/2012

07/12/2012

Participating at Globecom 2012, conference, Anaheim, USA

Anne VAN MULDERS 13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “Robust optimization method for the identification of nonlinear state-space models.”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Identification of a block-structured model with localised nonlinearity”

29/08/2012

30/08/2012

MHE Workshop 2012 , OPTEC Workshop on Moving Horizon Estimation and System Identification, Leuven, August 29-30, 2012. Presentation of paper “Imposing structure on identified nonlinear state-space models”

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Identification of a nonlinear LFR block-structure with two static nonlinearities”

08/10/2012

08/10/2012

Presentation of poster at the DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, Belgium, October 8, 2012 “Accuracy-complexity trade-off of some nonlinear identification methods: benchmark case studies”

Laurent VANBEYLEN 27/03/2012

29/03/2012

31st Benelux Meeting on Systems and Control, Nijmegen, The Netherlands, March 2729, 2012. Presentation of paper “Identification of nonlinear systems via a powerful block-oriented nonlinear model”

13/05/2012

16/05/2012

I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012. Presentation of paper “From two frequency response measurements to the powerful nonlinear LFR model”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Initial Estimates for the LFR Nonlinear Model Structure Via the Best Linear Approximation” and Chairman session “Maximum-Likelihood Estimation”.

29/08/2012

30/08/2012

Participating at MHE Workshop 2012 , OPTEC Workshop on Moving Horizon Estimation and System Identification, Leuven, August 29-30, 2012.

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “Improved, user-friendly initialization method for the identification of the nonlinear LFR block-oriented model”

Gerd VANDERSTEEN 11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Member of the local organizing committee

06/07/2011

08/07/2011

École Polytechnique Fédérale de Lausanne (EPFL) – Switserland. Measurements on Tokamak at EPFL, Lausanne, in cooperation with TU Eindhoven

Dhammika WIDANAGE 11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Design and Application of Signals for Nonlinear System Identification”

31

Annual report ELEC 2012

Hin Kwan WONG 13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “The Use of Binary Sequences in Determining the Best Linear Approximation of Nonlinear Systems”

11/07/2012

13/07/2012

16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012. Presentation of paper “Parametric Identification of Elastic Modulus of Polymeric Material in Laminated Glasses”

23/09/2012

26/09/2012

ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands. Presentation of poster “On the use of non-white input for identification of errors-in-variables dynamical system”

11/07/2012

Erliang ZHANG

32

Introduction to the department ELEC

1.5

ORGANISATION CHART OF DEPARTMENT ELEC (1/1/2013)

33

Annual report ELEC 2012

1.6

FUNCTIONAL ORGANISATION OF THE DEPT. ELEC

34

Introduction to the department ELEC

1.7

LIST OF THE MOST IMPORTANT MEASUREMENT EQUIPMENT

1.7.1

Signal Generators HP E1445A VXI Arbitrary Waveform Generator, fmax < 40 MHz (3x) HP E1340A VXI Arbitrary Waveform Generator, fmax < 42 MHz (3x) Synthesizer/Function Generator, Agilent 33120A (2x) NI 5411 AWG Noise Source, HP 346B, 10 MHz-18 GHz Agilent, 4142B, DC power supply HP E1434A VXI Arbitrary Waveform Generator, 4 channel source, fsmax: 65 KHz HP, Signal Generator, HP 83650B, 45 MHz - 40 GHz Tektronix, AWG710, 4 GHz Tektronix AWG 7052, 5GHz Arbitrary Waveform Generator Rhode & Schwarz, Vector Signal Generator, SMIQ06B, 300 kHz - 6.46 GHz Agilent 33250A, NI 5411, 2x AWG 33220A Agilent 81101A, Pulse generator, 50MHz

1.7.2

Spectrum Analysers, Impedance Analysers, Network Analysers 2 channel Dynamic Signal Analyser, HP 3562, 100 kHz (2x) Impedance Analyser, HP 4192A, 5 Hz - 13 MHz Vector Impedance Meter, HP 4193 A, 0.4 - 110 MHz Spectrum Analyser, R&S FSU, 20 Hz- 67 GHz µwave Network analyser, E8364B, 10 MHz - 50 GHz µwave Network analyser, N5242A, 10 MHz – 26.5 GHz 4 port Noise Gain Analyser, Eaton 2075 B, 10 MHz - 1800 MHz Network Analyser, HP 8753 C, 300 kHz - 6 GHz Spectrum Analyser, HP 8565 E, 9 kHz - 50 GHz PNA Network Analyser, Agilent, 5 0MHz - 50 GHz

35

Annual report ELEC 2012

Impedance Analyzer, Agilent E4991A, 10MHz - 3GHz Anritsu BTS Master MT8222A, High Performance, Handheld Base Station Analyzer

1.7.3

Digitizers 4 channel digitizer, Nicolet 490, 200 MHz, 8/12 bit 4 channel Digital Sampling oscilloscope, HP 54120T, 20 GHz, 11 bit 1 channel, HP E1430A VXI ADC 10 MHz, 16 bit (10x) 1 channel, HP E1437A VXI ADC 20 MHz, 16 bit (4x) 2 channel, HP E1429B VXI ADC 20 MHZ, 12 bit (2x) 8 channel, HP E1433A VXI ADC 196 KHz 2 NI 5911 flexres digitizer TDS 3032 digital phosphor oscilloscope, Agilent D5060321 300MHz, TDS 2001C 50 MHz 6 NI Elvis II

1.7.4

Miscellaneous Dual programmable filter, Difa PDF 3700, 100 kHz Dual adjustable filter, Wavetek, 100kHz Logic state analyser HP 1645A µwave power meter, HP436A, 10 MHz- 18 GHz 4 VXI racks +4 MXI controllers + Digital cards (2x Agilent E4841A + 1 Agilent E4805A) HP E1450, VXI timing module HP E1446A, VXI power module generator Wafer Probe Station Polytec Optical Fibre Vibrometer Velocity range (Doppler interferometer): 1, 5, 25, 125, 1000 mm/s/V Displacement range (Fringe Counter): 2, 8, 20, 80, 320, 1280, 5120

m/V

2 PXI mainframes + MXI controller + embedded controller Custom-built measurement setup for making geo-referenced GSM network measurements

36

Introduction to the department ELEC

4 hTC P3600 smart phones (equipped with 2G, 3G, Bluetooth, WiFi and GPS) 2 JRC DGPS 200

1.7.5

Underwater Acoustics Raytheon V860 echosounder B&K hydrophones, amplifiers etc. Panametrics transducers (500 kHz, 1MHz) D-GP5 Beacon Receiver KODEN (KBR-90) 1 watertank + positioning system Anritsu BTS Master MT8222A, High Performance, Handheld Base Station Analyzer

37

Annual report ELEC 2012

1.8

FINANCIAL SUPPORT 2012 Sponsor (project leader)

Duration project

Activity

Approx. amount (in €)

ADSI302 (K. Barbé) BAS12 (J. Schoukens) DWTC 282 (R. Pintelon) EU427 (J. Schoukens)

2011-2013

Central PhD office

2012

Core Funding dept. ELEC

2012-2017

Dynamical systems, control and optimization (DYSCO)

2013-2018

EU Advanced ERC-grant: Data Driven Structured Modelling of Nonlinear Dynamic Systems

FWOAL561 (G. Vandersteen)

2010-2013

Black box modeling of boreholes for MPC of ground-coupled heat pumps

160000

FWOAL599 (R. Pintelon)

2011-2014

Parametric and non-parametric techniques for modeling complicated time-variant dynamic systems

120000

FWOAL648 (J. Schoukens)

2012-2015

Modelling of structured nonlinear dynamic systems using parametric and nonparametric methods

280000

FWOAL622 (J. Schoukens, G. Vandersteen) FWOAL673 (Y. Rolain)

2013-2016

Detection and validation of variable viscoe elastic effects in the respiration system.

140000

2013-2016

Parametrized macro models for linear and nonlinear microwave and RF.

120000

FWOTM445 (W. Van Moer)

2007-2013

Identfication of multiport nonlinear microwave systems, bench fee post-doc. Wendy Van Moer

20000

FWOTM562 (K. Barbé)

2010-2013

System Identification of Finite Records, bench fee post-doc Kurt Barbé

12000

FWOTM586 (L. Lauwers)

2011-2014

Bench fee post doc “Identification of dynamic systems in FMRI signals, disturbed by RICE divided noise”

FWOTM610 (Y. Rolain)

2011-2014

Bench fee PhD student Maarten Schoukens “Identification of block-oriented models with parallel structures for nonlinear systems”

GIFT128 (M. Van Moer) GIFT 137 (K. Barbé)

2012

Non-US University Gift Agreement

2012-2013

Density estimators for the disturbing noise in sampling oscilloscopes, IEEE Instrumenation and Measurement Society

15000

HERC3 (W. Van Moer)

2008-2013

Multidisciplinary centre for measurements using advanced technologies of the Brussels University Association.

73418

IWT419 (J. Schoukens) Legal Expertise (L. Van Biesen) License Identification Toolbox (J. Schoukens) NDA29 (L. Van Biesen) NDA65 (A. Barel) NDA240 (G. Vandersteen) NDA367 (W. Van Moer) NDA584 (L. Van Biesen) NDA639 (W. Van Moer) OZR2169 (J. Schoukens)

2009-2013

LeCoPro Learning Control for Production Machines

Since 1995

Expert to the court

Confidential

Since 1994

Identification Toolbox

Confidential

Since 2004

Cellular positioning

Confidential

Since 2005

Confidential

2008-2018

Mutual non-disclosure agreement (NDA) :project for a RDS TMC reciever box NDA – Secrecy Agreement

Since 2009

NDA – Secrecy Agreement

Confidential

2011-2018

NDA – Non Disclosure Agreement

Confidential

2011-2014

NDA – Confidential Disclosure Agreement

Confidential

Since 2010

Bench fee Joint PhD VUB-Eindhoven University of Technology, Rijlaarsdam D.J.

OZR2200 (J. Schoukens) OZR2402 (J. Schoukens)

Since 2010

Bench fee Joint PhD VUB- Universiteit Hannover, Uribe David

Since 2012

Bench fee Joint PhD VUB-Budapest University of Technology and Economics, Sandor Kolumban

38

6500 98353 500000 2500000

4000 3720

22000

82168

Confidential

2000 4000 4000

Introduction to the department ELEC

Sponsor (project leader)

Duration project

Activity

OZRMETH1 (J. Schoukens)

2007-2013

Center for Data Based Modelling and Model Quality Assessment

SRP19 (G. Vandersteen)

2012-2017

SRP (Zwaartepunt): Center for model-based system improvement - From Computer-Aided Engineering to ModelAided Engineering

VLIR219 (L. Van Biesen)

2009-2014

Cooperation between l'Institut Supérieur des techniques Appliquées (ISTA) and Université Technologique de Kinshasa (UNITEK) in information and communication technology (ICT)

WDGO894 (L. Van Biesen) WDGO995 Kurt Barbé

2011-2012

Research Telecommunication

2012-2013

Service-agreement between the research institute 'VUB' and the 'Bel V'

WDGO1003 (W. Van Moer)

2012-2014

Research Agreement Cognitive Radio for Nuclear Power Plants

WDV12 GAMAX kft (J. Schoukens)

Since 2000

Graphical user interface for the frequency domain system identification toolset

1.9

AWARDS

1.9.1

Grade of Fellow (IEEE)

Approx. amount (in €) 3850000 450000

309700

187240 Confidential 106177 Confidential 2/3 ELEC 1/3 patent fund

The Institute of Electrical and Electronic Engineers, Inc. elected the grade of fellow to: Michel Gevers: for contributions to the understanding and identification of linear multivariable systems (1990) Johan Schoukens: for contributions to frequency domain system identification and the integration of measurement, signal processing and estimation theory (1997) Rik Pintelon: for fundamental research in frequency domain system identification and its applications in instrumentation, control and signal processing (1998) Yves Rolain: for contributions to measurement and modeling of nonlinear microwave devices (2005)

1.9.2

Grade of Senior member (IEEE) Gerd Vandersteen: In recognition of professional standing (2007) Wendy Van Moer: In recognition of professional standing (2007)

39

Annual report ELEC 2012

1.9.3

Awards from IEEE Instrumentation and Measurement Society (US) Johan Schoukens: Andrew R. Chi Best Paper Award Award “Identi-fication of Volterra Kernels Using Interpolation” by J.G. Nemeth, J. Schoukens and Istvan Kollar, IEEE Transactions on I&M, Vol. 51, No. 4, pp. 770-775 (2002) Yves Rolain received the Recipient of the 2004 IEEE Instrumentation and Measurement Society award “For Contributions to Nonlinear Circuit technology”. Johan Schoukens: IEEE Society Distinguished Service Award For technical and professional leadership of the IEEE Instrumentation and Measurement Society as Technical Program CoChair of IMTC/96 and author of conference papers on an annual basis, Associate Editor of the IEEE Transactions on Instrumentation and Measurement and member of the Society of Administrative Committee Wendy Van Moer received the “2006 Outstanding Young Engineer Award” from the IEEE Instrumentation and Measurement Society for outstanding contributions to nonlinear circuit theory. Rik Pintelon received the “2010 TIM Outstanding Associate Editor Recognition” for the meticulous, objective, professional and timely manner by which responsibilities while overseeing the review processes of numerous papers in 2010 are conducted. Wendy Van Moer received the “TIM Outstanding Associate Editor Recognition” in 2010 and 2011 for important contributions to TIM: for the meticulous, objective, professional and timely manner by which responsibilities while overseeing the review processes of numerous papers in 2010 are conducted. Rik Pintelon On behalf of the IEEE Transactions on Instrumentation and Measurement (TIM) administrative committee and the Instrumentation and Measurement Society (IMS) Publications Committee sincerely thanks Rik Pintelon for His meticulous, objective, professional and timely manner by which he conducted his responsibilities while overseeing the review processes of numerous papers in 2011. Therefore, we acknowledge your important contributions to TIM by recognizing Rik Pintelon as a 2011 TIM Outstanding Associate Editor. Kurt Barbé received the “2011 Outstanding Young Engineer Award” from the IEEE Instrumentation and Measurement Society for the innovative application of statistical techniques and signal analysis in biomedical measurements. Lee Gonzales Fuentes received the IEEE Graduate Fellow” from the IEEE Instrumentation and Measurement Society for her project “Kernel density estimators for the disturbing noise in sampling oscilloscopes”.

40

Introduction to the department ELEC

Kurt Barbé received the “TIM Outstanding Associate Editor Recognition’ of the IEEE Transactions on Instrumentation and Measurement in 2012 Kurt Barbé received the ‘Best Reviewer Award’ of the IEEE Transactions on Instrumentation and Measurement in 2009 and 2012

1.9.4

Award from IEEE Control Systems Society

Michel Gevers: Distinguished Member of the IEEE Control Systems Society in recognition of exceptional service to the Society and the profession (1997)

1.9.5

Grade of Fellow (IFAC)

Michel Gevers: For fundamental contributions to system identification and its connection to control (2006)

1.9.6

Belgian Francqui Chair ULB

Prof. Em. Michel Gevers (1994-1995): “Automatic, dynamic and analysis of systems” Prof. Dr. ir. Johan Schoukens (2006-2007): “Identification of linear systems in the presence of nonlinear distortions: a frequency domain approach”. Linear models are at the basis of many engineering activities. The aim of this course is to identify these models from experimental data. In real life, nonlinear distortions violate the ideal linear framework. Two solutions are discussed to extend the classic linear modelling approach. First the linear framework will be extended to include these distortions using best linear approximations and nonlinear noise sources. Alternatively, the nonlinear distortions will be explicitly modelled. Lectures (see pdf-files at http://wwwtw.vub.ac.be/elec/ELECcourse.htm): Inaugural: System Identification from data to model Lesson 1: Frequency Response Function Measurements Lesson 2 : Impact of Nonlinear Distortions on the Linear Framework Lesson 3: System Identification (pdf-file) Lesson 4 : Identification of Linear Systems Lesson 5: Identification of Nonlinear Systems

41

Annual report ELEC 2012

1.9.7

Awards granted by the VUB, on the proposition of the department ELEC Title of Doctor Honoris Causa to Prof. P. Eykhoff (Technische Universiteit Eindhoven) on April 4, 1990 (VUB, Brussels) Medal of Excellence to William Hewlett and David Packard on March 3, 1995 (VUB, Brussels) Medal of Excellence to Joseph F. Keithley on June 4, 1996 (Gothic Town Hall of Brussels) Title of Doctor Honoris Causa to Prof. M. Gevers (Université Catholique de Louvain CESAME) on November 28, 2001 (VUB, Brussels)

1.9.8

Distinguished Service Award from IMEKO

The International Measurement Confederation extends to Prof. Leo Van Biesen this Distinguished Service Award: As recognition and appreciation for his valuable contribution to the international exchange of scientific and technical information relating to developments in measuring techniques, instrument design and manufacture and in the application of instrumentation in scientific research and in industry. For his continuous support in IMEKO as member of several TCs, delegate of the Belgian Member Organization to the General Council, President Elect and Chairman of the Technical Board from 2000 to 2003, President of the Confederation from 2003 to 2006 and Past President and Chairman of the Advisory Board from 2006 to 2009.

1.9.9

Joseph F. Keithley Award in Instrumentation and Measurement: IEEE Field Award

Prof. Rik Pintelon received the 2012 Joseph F. Keitley Award in Instrumentation and Measurement, for outstanding contributions in electrical measurements

Rik Pintelon has played a pioneering role in introducing system identification to the instrumentation and measurement field as a modern approach to solving measurement problems. System

42

Introduction to the department ELEC

identification involves using statistical methods to build mathematical models of dynamical systems using measured data. Dr. Pintelon's innovative methods have found important use in a diverse range of areas, including measurement and modeling of metal corrosion and deposition, electric machines, inner-ear dynamics, and analysis of civil engineering structures. Dr. Pintelon also developed a frequency domain approach to system identification and pushed for its adoption within the control systems community. In 1991, he and his colleagues were successful in developing the Frequency Domain System Identification (FDIDENT) Toolbox for the pupular MATLAB program, which exposed his work to a large audience. Dr. Pintelon also published a highly cited book on system identification in 2001 (System Identification: A Frequency Domain Approach, IEEE Press), with a second edition that appeared in spring 2012.

1.9.10

Doctor Honoris Causa

Prof. Em. Michel Gevers received the title of “Doctor Honoris Causa” from the Vrije Univeristeit Brussel in November 2001 and from the Linköping University (Sweden) in 2010. Prof. Em. Ronny Van Loon received the title of “Doctor Honoris Causa” from the Hanoi University of Technology, in October 2008, for his personal contributions to the VLIR HUT IUC program in particular and the development of Hanoi University of Technology in general over the past 10 years. Thanks to his tremendous efforts as a key promoter since the establishment in 1998, the VLIR IUC programs with HUT has vigorously developed and reaped fruitful achievements, significantly contributing to the expansion of international network and international academic exchange at Hanoi University of Technology. Prof. Dr. ir. Johan Schoukens received the title of “Doctor Honoris Causa” from the Budapest University of Technology (Hungary) in May 2011.

43

Annual report ELEC 2012

1.9.11

Member of the “Royal Flemish Academy Of Belgium For Science And The Arts”

Prof. Dr. ir. Johan Schoukens has been elected in December 2009 as member of the “Royal Flemish Academy Of Belgium For Science And The Arts” for the section “Technical Sciences”.

1.9.12

Paper/presentation awards (since 2008)

Carine Neus received at the Symposium on Communications and Vehicular Technology in the Benelux (2008) the “Best Paper Award” for the paper “Challenges for Loop Identification and Capacity Estimation of DSL with Single Ended Line Testing”. Carine Neus received from IMEKO the “Best Paper Award” for the paper “Feasibility and problems of DSL loop topology identification via single-ended line tests” presented at the 16th IMEKO TC4 International Symposium and 13th International Workshop on ADC Modelling and Testing, Florence, Italy (September 2008) Mussa Bshara and Leo Van Biesen received the “Top Six Achievement Award “Winning Paper” for the paper “Potential Effects of Power Line Communication on xDSL Inside the Home Environment” presented at the VIII Semetro. 8th International Seminar on Electrical Metrology João Pessoas, Paraíba, Brazil June 17 - 19, 2009 John Lataire and Rik Pintelon received the “2008 Award Winner-Measurement Science” for the article “Noise level estimation in weakly nonlinear slowly time-varying systems” published in Measurement Science and Technology, Vol. 19, No. 10 (2008) Mussa Bshara and Leo Van Biesen received the “Best Paper Award” for the paper “Fingerprintingbased Localization in WiMAX networks depending on SCORE measurements”, presented at the Fifth Advanced International Conference on Telecommunications, AICT 2009, Venice/Mestre, Italy, May 24-28, 2009 Yves Rolain received the “Automated RF techniques group best paper award” from IEEE in 2010 John Lataire received the “Best Junior Presentation Award 2010” at 29th Benelux Meeting on Systems and Control in Heeze, The Netherlands. He received the DISC trophy for the presentation of the paper “Frequency Domain Least Squares Estimator of Time-Varying”.

1.9.13

Master thesis awards

Diane De Coster received in October 2011 from FWO the “Barco High Tech Awards for Master thesis”, for her master thesis entitled “Ontwerp en realisatie van een geminiaturiseerde elektronische lock-in detectiemodule voor het meten van biomoleculen in fotonische 'lab-on-a-chip' systemen”.

44

Introduction to the department ELEC

Maarten Schoukens received in March 2011 from IMEC the “IMEC-award for the best Master thesis at the faculty of Engineering, at the Vrije Universiteit Brussel” for his master thesis entitled “Ontwerp en realisatie van een compensatie voor niet-lineaire RF vermogenversterkers”. Egon Geerardyn received in March 2012 from IMEC the “IMEC-award for the best Master thesis at the faculty of Engineering, at the Vrije Universiteit Brussel” for his master thesis entitled “A simulation Method for Pin-pointing the Dominant Nonlinear Contributors in CMOS Circuits”.

1.10

INTERNATIONAL CONFERENCES/WORKSHOPS ORGANISED BY THE DEPT. ELEC

International conferences International Instrumentation and Measurement Technology Conference (IMTC), Brussels (Belgium), 4-6 June, 1996 16th International IFAC Symposium on System Identification (IFAC-SYSID), Brussels (Belgium), 11-13 July, 2012

International workshops 22nd Benelux meeting on Systems and Control, Vossemeren (Lommel), 19-21 March, 2003 26th Benelux meeting on Systems and Control, Vossemeren (Lommel), 13-15 March, 2007 30th Benelux meeting on Systems and Control, Vossemeren (Lommel), 15-17 March, 2011

1.10.1

16th IFAC Symposium on System Identification – SYSID 2012

The 16th IFAC Symposium on System Identification (SYSID 2012) was held in Brussels, Belgium, July 11-13, 2012. This symposium covered major aspects of system identification, experimental modeling and signal processing, ranging from theoretical and methodological developments to practical applications. The symposium was held at the SQUARE Brussels Meeting Center, right in the heart of Brussels. The location of the meeting center at the “Mont des Arts” district is a short stroll from all of the historic city’s major attractions, such as the Grand-Place to point out one of them, with over 13.000 hotel rooms just around the corner, plus the pick of the city’s restaurants, bars and shops. The Central Station, which is located in front of the convention center, has direct and fast train connections with Brussels National Airport, as well as high-speed train connections to London, Paris, Amsterdam, Cologne.

45

Annual report ELEC 2012

The central location of the venue turned out to be both pleasant and efficient, with well-equipped and well-distributed rooms for parallel sessions and social events. The SYSID, which is held every three years, is organized by the International Federation of Automatic Control (IFAC). The symposium provides a forum for the diverse communities involved in the theory and applications of system identification. Previous SYSID symposia where held in Prague (1967, 1970), The Hague (1973), Tbilisi (1976), Darmstadt (1979), Arlington, Virginia, USA (1982), York, UK (1985), Beijing (1988), Budapest (1991), Copenhagen (1994), Kitakyushu (1997), Santa Barbara, California, US (2000), Rotterdam (2003), Newcastle, Australia (2006), Saint-Malo, France (2009). The 2012 symposium, which was the 16th, was the first SYSID to be held in Belgium. SYSID 2015 will be held in in Beijing, China, October 19-21, 2015, with Ji-Feng Zhang as the national organizing chair and Robert Bitmead as the international program committee chair. 329 Participants from 37 countries attended the conference, including 100 Ph.D. students. France, Belgium, Sweden, USA and the Netherlands had the highest level of participation for both delegates and paper authors.

The plenary program was designed to highlight both progress and open avenues for the field of system identification. The plenary lectures were the following: “Advanced Control of High Tech Systems” by Maarten Steinbuch, Eindhoven University of Technology, The Netherlands

46

Introduction to the department ELEC

“Certified System Identification – towards distribution-free results” by Marco Campi, University of Brescia, Italy “Learning and Inference for Graphical and Hierarchical Models: A Personal Journey” by Alan S. Willsky, Massachusetts Institute of Technology, Cambridge, MA, USA “Optimized behavioral interventions: what does system identification and control engineering have to offer?” by Daniel E. Rivera, Arizona State University, Tempe, AZ, USA “Machine Learning, Probalistic Inference, System Identification and Control” by Carl Edward Rasmussen, University of Cambridge, UK “Compressive Information Extraction: A Dynamical Systems Approach” by Mario Sznaier, Notheastern University Boston, MA, USA We would like to thank all members of the International Program Committee and the National Organizing Committee for their work on the symposium. We would also like to thank our family, friends and colleagues for their help and support, especially for their help in organizing the registration desk and the technical assistance in the session rooms. Special thanks are due to Pradeep Misra for his outstanding service.

47

Annual report ELEC 2012

2. Short Description of the Research Projects/ Team 2.1

TEAM A: AUTOMATIC MEASUREMENT SYSTEMS, TELECOMMUNICATIONS AND LABORATORY OF UNDERWATER ACOUSTICS

2.1.1

Introduction to Team A

The activities of team A directly related to fundamental electricity, deal with the set-up of computer controlled measurement systems, the design of new intelligent instruments, the processing of the measured data (DSP and algorithmic treatment) and the implementation of A.I.- techniques in instruments for the automatic interpretation of the acquired measurements. The technical applications areas are quite diverse and cover areas in the field of electrical and electronic systems, but a lot of special care is also given to earth science applications. This team is also active in telecommunication projects and studies, which deal with voice coded transmissions

(telephony)

and

with

robust

coded

digital

transmission

(ADSL,

VDSL).

Communication channels are modelled, with emphasis on the physical layer. Moreover, in the field of underwater acoustics important research projects are developed since 1985. Fundamental as well as applied research is carried out in the field of the modeling of marine systems, marine acoustics, sub bottom profiling and sediment classification. Prof. Dr. ir. L. Van Biesen has been the Belgian delegate in the board of IMEKO since 1993 and chairman of the Technical Committee TC-7 on Measurement Science since 1994, up to 2000, has been vice-chairman of the Technical Committee TC-19 on Environmental Measurements since 1999, President-Elect of IMEKO (2000-2003) and was the President of IMEKO (2003-2006), and has been Post-President of IMEKO (2006-2009)

48

Short Description of the Research Projects/ Team

2.1.2

Short Description of the Research Projects of Team A

2.1.2.1

Modeling of the channel transfer function and the crosstalk for specific historical connectivity practices in DSL copper networks and assessment of the effect on the achievable data rates (Carine Neus, Leo Van Biesen, Lieven Mertens*, and Kurt Coulier*) (*) Belgacom

Digital Subscriber Line technologies like ADSL, ADSL2+, VDSL2 and its evolutions offer high-speed data services (e.g. internet connection, digital television,...) to customers over the existing telephone lines. For practical reasons, the telephone lines are not simply a single straight line between the central office and the customer. Lines are bundled for cost effectiveness, repairs have been made, different line types are cascaded,.... As a consequence, very peculiar topologies can be found in the field. In this research, we focus on one specific type of connection, namely ‘retour pairs’, which have a different topology than the ‘direct’ pair (see Figure 1). The retour pair passes the customer’s house and returns after a certain distance “r”, typically at the splice at the end of the cable. Hence their name ‘retour pairs’. Due to historical reasons many customers receive two twisted pairs at home, and in the majority of these cases it is one direct pair and one retour pair. Due to the characteristics and spectrum usage of VDSL2, the use of the direct pair for service offering is obvious and recommended. But with the ever increasing request for bandwidth driven by strong competition and customer demand, and with the evolution of VDSL2 towards VDSL2 Vectoring, the operators show a need to better understand the behaviour of the retour pair. After all, the use of the retour pair is economically very interesting since civil works might be avoided in some cases, despite the expected degraded overall performance of the retour pair.

Figure 1. Two twisted pair lines arriving at one customer: one direct pair and one retour pair

This research aims at predicting the achievable data rate on retour pairs. We assume that they will behave differently from a direct pair, as their physical construction is different (i.e. they are folded back on themselves after a certain distance). Basically, two aspects must be studied:

49

Annual report ELEC 2012

1. The channel transfer function needs to be modelled. 2. The crosstalk needs to be evaluated. During 2011, a model has been developed for the channel transfer function of a retour pair. The model has been validated in the ELEC laboratory and in the field (in an operational distribution center). In 2012, the crosstalk in the presence of retour pairs will be evaluated. This project is performed in cooperation with Belgacom.

50

Short Description of the Research Projects/ Team

2.2

TEAM B: SYSTEM IDENTIFICATION AND PARAMETER ESTIMATION

2.2.1

Introduction to Team B

The main interest of the identification team is situated in the development of new identification methods and their application to real life problems. The team is involved with many aspects of the identification theory: experiment design development of estimators modelling problems The identification of linear and nonlinear systems is studied. Throughout our work we make the following choices: Use of periodic excitations whenever it is possible, use random excitations if it is imposed by the user. Use of non-parametric noise models to characterize the stochastic disturbances Use of an errors-in-variables framework: all measured signals are assumed to be disturbed by noise. Believe your data, not your prejudices. It will not always be possible to act along these clear principles, but whenever we face a choice, they are an important factor in our decision process. Another important aspect of our general approach is that we start the process by gathering system knowledge from the measurements. This gives in a very early phase of the identification process an idea of the global complexity of the modelling problem. For the linear systems this phase boils down to the non-parametric measurement of the frequency response function, which is usually of a high quality due to the periodic excitation approach. It contains a lot of information about the system. For nonlinear systems the situation is not that clear, and a number of ideas are under study at this moment. IDENTIFICATION OF LINEAR SYSTEMS In the eighties, ELiS, an estimator to identify single input single output (SISO) linear dynamic systems in the presence of uncorrelated input/output noise, was developed at the department ELEC. In a next step this estimator was generalized to multiple input/multiple output systems in the presence of correlated input and output noise.

51

Annual report ELEC 2012

Since linear dynamic systems are used as a basic modelling tool in a very wide range of applications, it is clear that this work has a lot of applications. We applied this modelling approach to chemical problems (traction batteries, diffusion processes), power engineering (electrical machines, power transformers; fault localisation on a cable), mechanical problems (flutter analysis, flexible robot arm) measurement area (compensation of the dynamics of an acquisition channel, modelling the dynamics of a sensor), and signal processing (design of digital filters). In all these applications the identified transfer function model was an intermediate step, for example leading to a better physical insight in the structure of the DUT, or to be used to improve the quality of the process control. Nowadays we are further developing the estimators, making them more robust and user friendly, resulting in a wider applicability of ELiS. Moreover a lot of effort is spent to generalize frequency domain identification methods so that they also can be applied under exact the same conditions as time domain methods, but still offering the advantages of using a non-parametric noise model.

robustification: what happens if the true model is not contained in the considered model class, for example when non-linear distortions or unmodelled dynamics are present what can be done if no noise information is available identification of high order systems, identification of an over parametrized model generation of improved starting values

user friendly: is it possible to extract the noise information from the same measurements that are used to identify the system without user interaction is it possible to guide the inexperienced user to a good solution of his identification problem development of fully automated processing methods: from raw time domain data to validated models (including automatic detection and processing of the periodicity of the signals; extraction of non-parametric noise models; model selection; model validation)

Generalization: identification of MIMO systems: design of excitation signals, development of new algorithms, constraint estimation (stability, positive definite systems) identification of continuous time models using arbitrary excitations developing a general theoretic framework linking time and frequency domain identification

52

Short Description of the Research Projects/ Team

generalization to other fields like electrochemical reactions (÷s) and microwave systems (Richardson)

IDENTIFICATION OF NON LINEAR SYSTEMS In the last 10 years we started to study more intensively identification of non-linear systems. It is not a good idea to define the class of models under study by a negation. Just as it is an impossible task to make a study of the zoology of the ‘non elephants’, it is impossible to grasp all non-linear systems in one framework. We should be more specific and give a positive definition of those systems that we will consider. In this project we focus on those systems where a periodic input results in a periodic output with the same period as the input (we call it PISPOT systems). This excludes a lot of phenomena like chaos, bifurcation, but it is still a wide class, including hard nonlinear systems like clippers, relays, deep saturation etc. Selecting a specific model structure for these system is still a very complex task. Not only the non-linear behaviour should be properly characterised, also the dynamics should be captured. This requires that many choices from the user, compared to the linear problem, that it is not a good idea to start immediately with a parametric model when identifying a non-linear system. Too many questions would remain unanswered. For that reason, we prefer to start with a non-parametric representation of the system, trying to get a first insight in its behaviour. Only in a second step, we will eventually move to a specific parametric model that might be well adapted to the given problem, using the previously gained insights to select a dedicated model structure. Our work on identification of nonlinear systems is split in two parts. In the first part we study the impact of nonlinear distortions on the linear identification framework. In the second part we aim at modelling the nonlinearities.

Linear modelling in the presence of nonlinear distortions Linear models are successfully applied to a wide range of modelling problems although they are based on very restrictive assumptions. The real world is not linear, and hence intrinsically the theory is not applicable. However, in practice, the linear approximations are useful and offer important advantages: They result in useful models that give the user a lot of intuitive insight in his problem. Many design techniques, that cannot be easily generalised to nonlinear models, are available. Nonlinear model building is mostly difficult and time consuming, while the additional performance that is obtained might be small. So it is not obvious that the gain in model performance warrants the required efforts to build a nonlinear model.

53

Annual report ELEC 2012

No general framework is available for nonlinear systems as it is for linear systems. Often dedicated models are needed, complicating the development/use of general software packages. For these reasons we work on a generalised linear framework that can be used in a nonlinear environment. Using this framework it is possible to get: A better understanding of the impact of nonlinear distortions on the model. Optimized measurement techniques that reduce the required measurement time significantly. Generalised uncertainty bounds that describe the model variations due to the nonlinear distortions. This allows for a better balanced design that accounts for the model limitations. A lower risk of being fooled by the classical linear identification methods that are widely used in commercial packages. Simple design rules that help to reduce the undesired impact of nonlinear distortions on practical designs. The whole approach is based on the observation that for random excitations, a nonlinear system can be represented under very mild conditions by a linear system plus an additive noise source. The linear system GR(j

k)

gives the best linear approximation of the output for the considered class

of excitation signals, while the noise source ys(j

k)

represents all nonlinear effects that are not

captured by the model.

Identification of nonlinear systems Nonlinear modelling is extremely difficult. The major reason for this problem is the enormous variability that exists. Even when we zoom in to for example Volterra systems, there are still many additional degrees of freedom compared to the modelling of linear system. The only model structure question for SISO-transfer functions is the order of the numerator and denominator. For SISO-Volterra systems, the situation is much more complicated, because the model uses multiple frequency variables that can appear in any possible combination. Hence no simple prior structure selection is possible. For that reason it is our strong belief that the parametric modelling step should be preceded by a non-parametric one. First the user should get an impression of the nonlinear behaviour, and only in the next step he can propose a parametric model structure. For that reason we studied and still look at a number of non-parametric methods like: Restoring force method, the power transfer method, the non-parametric Volterra models. For the parametric modelling we follow different approaches: nonlinear block-structured models and nonlinear state-space models

54

Short Description of the Research Projects/ Team

Non-linear block-structured models consists of linear dynamic blocks combined with static non-linear blocks. In its most general form feedback loops can be present. The major advantage of this structure is that it still provides some physical insight in the system, and it uses a ‘small’ number of parameters. The major disadvantage is that it is extremely hard to find good initial estimates for the individual blocks. Non-linear state-space models cover a large class of non-linear systems with a very rich behaviour. They can be considered as a potential candidate for ‘black-box’ modelling of non-linear systems. The major advantage of this structure is that it is much easier to find reasonable initial estimates for the structure for systems with a dominating linear behaviour. The major disadvantage is the large number of parameters that are used. These models provide also less physical insight. Finally we study also dedicated non-linear structures to model a number of typical highfrequency components like mixers. For these systems we use either Volterra-based descriptions, or dedicated block-structured models.

METHUSALEM: Centre for Data Based Modelling and Model Quality Assessment This is a project setup by the Flemish government to give long term stability to established research groups. The project runs for 7 years, with a budget of more than 500 000 Euro/ year. The ultimate goal of the project is to set up and maintain a centre for system identification, summarized in the motto: “From data to model” The aim of this centre is to develop, acquire and disseminate advanced system identification methods.

Structure of the project This project follows three lines towards this long term goal: Development of advanced identification methods for dynamic systems: Development of robust and user friendly methods for the identification of (non)linear dynamic systems. This is the home base of the identification team, and it consists of a ‘natural’ continuation and extension of the activities of the team over the last 20 years. High risk new challenges: Start-up of completely new high risk research lines that need a substantial development compared with the existing theory and methodology. It consists of two sub-projects: Development of a ‘non-asymptotic’ identification theory: How to deal with modelling problems where the number of data points is in the same order as the number of parameters?

55

Annual report ELEC 2012

Identification in the presence of noise and model errors: Development of a new paradigm that balances/tunes noise disturbances and model errors. Learning, dissemination, networking: The aim is to acquire and disseminate actively knowledge from/to other fields by offering one year research grants for visitors. Learning: On the one hand we want to acquire system identification methods that are developed in other fields like statistics, econometrics, and are unknown within the controland measurement society (learning), by hosting specialists of these fields for longer periods. Dissemination: On the other hand, many scientific disciplines face the problem of extracting mathematical models from experimental data, while this does not belong to their core business. Here, we want to give an active support to transfer sound identification methods to these groups, by hosting and paying PhD-students of external (VUB-) groups for a longer period. Networking: Extend the existing international network of post-doc and senior researchers by combining short term and/or long term visits.

2.2.2

Short Description of the Research Projects of Team B

2.2.2.1

An application example for the B-spline based nonparametric system identification* (Peter Zoltán Csursia, Johan Schoukens, István Kollár) *Cotutelle: Vrije Universiteit Brussel (J. Schoukens) and Budapest University of Technology and Economics (I. Kollár)

Introduction The goal of this paper is to present a smoothing techniquefor the identi cation of systems whose frequency response function (FRF) is smooth.

Using a generalized and modi ed B-spline based

methodology an FRF can be estimated. With respect to the system dynamics, it can be possible 1) to reduce the disturbing noise 2) to decrease the effect of transients 3) to reduce the number of model parameters.

Spline interpolation technique For LTI systems the modi ed B-spline method [1][3] can be used as a tool for estimating nonparametric smooth FRFs in the frequency domain.

The simple B-spline interpolational

estimation are based on the carrier estimates, for instance Cross Power Spectral Density (CPSD),

56

Short Description of the Research Projects/ Team

Hann windowed, Local Polynomial Method (LPM) [2]. In other words, after computing the FRF estimation as a second step, a B-spline method is used. It can be shown that B-splines improve the statistical properties of these estimators, for instance the expected value of the error and the variance level are smaller.The distribution of errors may change also: in many cases it convergences faster to the normal distribution.

Figure 1. On the left a random realization is shown. Pink line is the observed system with transient and noise. The red line is the CPSD estimate, the black line is the Hann windowed estimate, the green line is the LPM estimate and the blue line is the Bspline estimate based on LPM. The rhombuses are the variances of the estimators. On the right the empirical pdf of the rms error of the observed system is shown.

An example A second order Chebyshev lowpass lter is observed (10 dB of peak-to-peak ripple in the passband, the cutoff frequency is 0.2 f s /2, the sampling frequency (f s ) is 512 Hz,the length of the measurement is 512 samples) excited by a random noise input [2]. The observations also include transient. The empirical probability density function (pdf) of the rms error (after running 10.000 simulations) and a random realization of the FRF estimations together with the variance are shown. The expected value of the rms value of the noise is 1% of the expected value of the rms of the random input excitation. The LPM order is 2, bandwidth is 3 [2], B-spline order is 4,step size is 2 [3].

Summary In this work a powerful estimation method is developed for smooth LTI systems. This technique is illustrative, exible, user friendly. The only drawback is that it requires more computing time.

References [1] H. Prautzsch, W. Boehm, M. Paluszny, ”Bezier and B-spline techniques,” Springer-Verlag, Berlin, 2001 [2] R. Pintelon, J. Schoukens: ”System identi cation, a frequency domain approach, second edition”, 2012

57

Annual report ELEC 2012

[3] P.Z. Csurcsia, J. Schoukens, I. Kollár: ”Identi cation of time-varying systems using a two-dimensional Bspline algorithm,” I2MTC, pp. 1056-1061, Graz, Austria, 2012

2.2.2.2

Design of optimal inputs for nonlinear dynamic systems (Alexander De Cock)

The quality of an identified model strongly depends upon the quality of the experiment. By making a proper experiment design it is possible to maximize the information that can be retrieved from the experiment. For linear dynamic systems and static nonlinear systems the optimal input design problem is well understood. However for nonlinear dynamic systems optimal input design remains problematic. Most difficulties arise from the fact that not only the power spectrum but also the amplitude distribution of the excitation signal is involved and both of them should be simultaneously considered during optimization. Additionally nonlinear models are often only valid within a limited input range. As a result, approximation errors should also be taken into account. Depending on the application, different criteria can be used to express optimality of the input with respect to the properties of the estimated model. This research will focus on the determinant of the fisher information matrix as a criterion for optimality. By selecting the input that maximizes this determinant, the uncertainty volume in the parameter space will be minimized. In the case of linear time invariant (LTI) and static nonlinear systems (SNL) an efficient algorithm [1], based on the dispersion function, exists that finds an optimal input that maximizes the determinant of the fisher information matrix. Initially the research will focus on how the methods described in the LTI and SNL case can be extended to dynamical nonlinear systems. To this end the framework will be gradually build up by considering series of different subclasses such as Wiener systems, Hammerstein Systems, Wiener Hammerstein systems, Parallel structures and Feedback structures. Once the framework is in place additional constraints will be imposed that take into account the specific application of the model, limitation of the measurement setup and robustness toward parameter uncertainty.

Reference [1] Johan Schoukens, Rik Pintelon , ‘Identification of Linear Systems: A Practical Guidline to Accurate Modeling’, Pergamon Press, 1991

2.2.2.3

User-friendly identification of massive MIMO systems (Egon Geerardyn, Johan Schoukens, Yves Rolain)

The goal of this project is to develop semi-automatic identification methods for massive multipleinput multiple-output (MIMO) systems, i.e. systems with a very large amount of in- and outputs. Future state-of-the-art systems such as wafer steppers, will have many inputs ni (actuators) and outputs no (sensors), resulting in many (ni × no , e.g. 30 × 30 = 900) transfer functions to be

58

Short Description of the Research Projects/ Team

modeled. This prohibits manual tuning. However, modern control design methods require parametric plant models with reliable confidence bounds. Therefore identification methods with limited human interaction are needed. This affects all the individual parts of the identification cycle: Experiment design, where classical methods assume the system to be known beforehand, which leads to a chicken-and-the-egg problem. Therefore, robust excitation signals and measurement strategies are needed that guarantee a minimal model quality for a broad class of systems. In this project, quasilogarithmic multisines were already shown to be an optimal solution for single-input single-output systems [1]. For a given lower-bound on the damping of the system poles, this allows for a simple but efficient design of the excitation signal. This work will be extended to the MIMO case. Nonparametric estimation methods such as the Local Polynomial Method (LPM) [2] and Local Rational Method (LRM) [3] are to be studied further for the massive MIMO case. A hybrid LPM/LRM method will be developed to leverage the advantages of each. An automatic tuning of the local bandwidth and order of these methods will be developed taking the needs of large MIMO systems into account. In parametric estimation of massive MIMO systems, the nonparametric plant and noise models need to be considered to improve numerical convergence and minimize numerical artifacts. Validation of massive MIMO systems remains a challenge at the moment. E.g. for a 30 x 30 system, 900 individual transfer functions are to be considered. This is no longer practically feasible by hand. To overcome this, each separate transfer function is to be locally validated over the frequency in an automated fashion. By doing so, nonvalidated parts of the model can be reported and inspected manually. The human interaction can therefore be reduced and focused on the most critical parts of the model.

References [1] Geerardyn, E.; Rolain, Y. & Schoukens, J. “Quasi-logarithmic Multisine Excitations for Broad Frequency Band Measurements” IEEE International Instrumentation and Measurement Technology Conference, May 13-16, 2012, 737-741 [2] Pintelon, R.; Schoukens, J.; Vandersteen, G. & Barbé, K. “Estimation of nonparametric noise and FRF models for multivariable systems -- Part I: Theory”, Mechanical Systems and Signal Processing, 2010, 24, 573 - 595 [3] McKelvey, T. & Guérin, G. “Non-Parametric Frequency Response Estimation Using a Local Rational Model” Proceedings of the 16th IFAC Symposium on System Identification, July 11-13, 2012

59

Annual report ELEC 2012

2.2.2.4

Measurement and estimation of Linear Parameter-Varying systems (J. Goos, J. Lataire & R. Pintelon)

Nowadays, the Linear Time-Invariant (LTI) framework [1] has become well-known and it is widely used in the industry. An LTI model predicts the behavior of a plant to a given input, which can be used by a controller to better match the targeted output, like production specifications. Although the LTI system identi cation framework has proven its merits for many years, in quite some applications the linearity and time-invariance hypotheses are only approximately true or not valid at all. The need to operate processes with higher accuracy and ef ciency has therefore resulted in the realization that the non-linear (NL) and time-varying (TV) nature of many physical systems must be handled by the control design. In this Phd. thesis we will study Linear Parameter-Varying (LPV) systems [2] through state-space (SS) models, because they are very suited to deal with systems that have multiple inputs and outputs. The classical (discrete) LTI state space equations are

where k denotes the time index,

x are the states, u the inputs and y the outputs. In the LPV

framework, the matrices A, B, C and D can now vary with time.

Because of the time-varying dynamics, the classical identifications techniques have to be adapted. Our proposed approach consists of two steps. First, a rough initial model is estimated. Next, this model will be refined by a non-linear optimization algorithm. The resulting LPV model can then be used directly for more precise control [3].

References: [1] R. P. &. J. Schoukens, System Identi cation: A Frequency Domain Approach, IEEE Press, 2001. [2] R. Tóth, Modeling and Identi cation of Linear Parameter-Varying Systems, Springer Germany, 2010. [3] F. W. &. K. Dong, "Gain-scheduling control of LFT systems using parameter-dependent Lyapunov functions," Automatica, vol. 42, no. 1, pp. 39-50, 2006.

2.2.2.5

A frequency-domain formulation of Least-Squares support vector machines to deal with coloured noise. [John Lataire, in collaboration with Dario Piga (TU Eindhoven) and Roland Tóth (TU Eindhoven)]

This research topic investigates the benefits of formulating the Least-squares support vector machines (LS-SVMs) in the frequency domain. One clear advantage is the robustness of the frequency domain formulation to correlated noise. Significant improvements have been observed

60

Short Description of the Research Projects/ Team

when comparing the frequency domain LS-SVM with its time domain counterpart on a linear, static, time-varying system with coloured disturbing noise. Least-squares support vector machines have been introduced in the machine learning community [Suykens et al. 2002] to perform, among other goals, regression on measured (noisy) data. Their strength with respect to classical least squares regression is that the set of basis functions of LSSVMs is allowed to be potentially infinite dimensional. This allows LS-SVMs to be incredibly versatile. It can be thought of as a model that could describe virtually any type of input-output relation. However, an additional task linked with this versatility is that the model structure must be tuned thoroughly for a any new data generating process. Otherwise the model is exposed to heavy over- or undermodelling.

Figure 2. Calibration

objective

functions.

For

the

time

domain

formulation, the optimum is shifted w.r.t. the ideal objective function.

Figure 1. Comparison

between

the

frequency

domain and time domain estimates.

Tuning the model structure involves splitting the available data set into a training and a validation data set. The model is estimated on the training data set, and its ability to predict the behaviour of the validation data set is optimized by tuning the structure. This tuning, however, will be fooled if the disturbing noise on the training and the validation data sets are correlated. This is where the frequency domain formulation is beneficial: correlated noise in the time domain is not correlated in the frequency domain. This is (asymptotically) true if the noise is stationary. Tuning the model structure in the frequency domain therefore yields significantly better results. These results are illustrated on the simple simulated system: y(t) = a(t)u(t) + v(t) where u(t) and y(t) are the measured (noiseless) input and (noisy) output signals, respectively. v(t) is time-correlated noise, and a(t) is a modulating signal to be identified. The results of the

61

Annual report ELEC 2012

estimation are given in Figure 1. The true a(t) was a sinc function, in black. Its estimate is given in white (obtained after tuning the model), and its uncertainty region (standard deviation) is given by the grey shaded area. The frequency domain estimation clearly has a lower uncertainty than the time domain one. The reason for the better frequency domain estimation is given in Figure 2. The frequency domain calibration cost function (used for tuning the model) has an optimum which coincides with the optimum of the ideal calibration cost function, thus yielding an optimal model structure. The calibration cost function in the time domain is fooled by the correlation between the noise on the training and calibration data sets. It has, therefore, a shifted optimum, yielding a higher variance on the estimated model. Building on these results, the extension to the identification of dynamic systems from data corrupted by coloured noise, by using frequency domain LS-SVMs will be investigated in future work.

References: [1] A.

Suykens, T. Van Gestel, J. De Brabanter, B. De Moor, and J. Vandewalle. Least Squares Support

Vector Machines. World Scientific Publishing, Singapore, 2002. [2] R. Pintelon, J. Schoukens, and P. Guillaume. Continuous-time noise modelling from sampled data. IEEE Trans. on Instrumentation and Measurement, 55(6):2253–2258, Dec. 2006.

2.2.2.6

Nonlinear dynamic modeling with Neural Networks (Anna Marconato, Johan Schoukens)

Considered problem Nonlinear dynamic system modeling is an active research field where several challenges still need to be solved. In particular, capturing both the nonlinear behavior and the dynamics of an unknown system represents a difficult identification problem. On one hand, many different model structures have been proposed within the system identification area to characterize nonlinear dynamic systems (e.g. Volterra kernels, NARX models, block structures and nonlinear state-space models) [1]. On the other hand, there has been an increasing attention towards techniques developed in other fields, in particular in the machine learning community, where Neural Networks (NNs) and Support Vector Machines (SVMs) allow one to accurately approximate nonlinear functions [2,3]. There, a big challenge is how to incorporate dynamics in methods that are essentially designed to model static nonlinearities. To address these issues, we have recently proposed a NN-based technique for the identification of nonlinear state-space (NLSS) models to combine the best of these two worlds [4,5].

62

Short Description of the Research Projects/ Team

Proposed method The NN-NLSS algorithm aims at identifying nonlinear state-space models of the form:

x(t 1) y (t )

Ax(t )

Cx(t )

Bu (t )

Du (t )

f NL ( x(t ), u (t ))

g NL ( x(t ), u (t ))

In this way, it is possible to estimate the dynamics and the nonlinearities in the system separately, which simplifies the identification procedure and makes it more efficient. The identification algorithm consists of a three-step initialization scheme: estimation of the BLA; approximation of the unknown nonlinear state; estimation of the nonlinear terms as two independent one-hidden-layer feedforward NNs; and a final parameter optimization step.

Results and comments The proposed method has been successfully applied on real measurement problems such as the Wiener-Hammerstein, the Silverbox, and the Crystal Detector benchmark examples. In all the problems considered so far, the NN-NLSS method has yielded a good trade-off between model accuracy and model simplicity, i.e. the obtained models were characterized by a quite low validation RMSE and by a low number of parameters (especially when compared to other machine learning related methods). The results can be considered quite satisfactory, given that no a priori knowledge about the underlying system is used in the modeling procedure. Moreover, the obtained results show that the tuning of user-specified parameters is not a very critical step, especially since model selection can be performed in an efficient way already after the initialization stage, thus saving a considerable amount of time in the estimation procedure.

References [1] 1. L. Ljung, "Perspectives on system identification", Annual Reviews in Control, vol. 34, pp. 1-12, 2010. [2] 2. T. Hastie, R. Tibshirani and J. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Springer-Verlag, 2009. [3] 3. J.A.K. Suykens, J. Vandewalle, B. De Moor, Artificial Neural Networks for Modeling and Control of NonLinear Systems, Springer, 1996. [4] 4. A. Marconato, J. Sjöberg and J. Schoukens, "Initialization of nonlinear state-space models applied to the Wiener-Hammerstein benchmark", Control Engineering Practice, vol. 20, no. 11, pp. 1126-1132, 2012.

63

Annual report ELEC 2012

[5] 5. A. Marconato, J. Sjöberg, J.A.K. Suykens and J. Schoukens, "Identification of the Silverbox benchmark using nonlinear state-space models", 16th IFAC Symposium on System Identification (SYSID 2012), Brussels, Belgium, 2012.

2.2.2.7

Identification and comparison of several nonlinear models for the glucoregulatory system (Anna Marconato, Maarten Schoukens, Koen Tiels, Dhammika Widanage, Amjad Abu-Rmileh*, Johan Schoukens) *Research Group on Statistics, Econometrics and Health (GRECS), University of Girona, Spain

Insulin-glucose system modeling for control To develop an artificial pancreas for T1DM treatment, closed-loop controllers play a crucial role in regulating the glucose level in the blood. The mathematical models currently available to characterize a diabetic patient are too complex to be considered for control purposes. Therefore, the application of advanced identification techniques to model insulin-glucose systems, based on available input/output data, represents a crucial step towards the development of the artificial pancreas for T1DM patients. In particular, in [1] it is shown that identification of nonlinear models is needed, since linear modeling does not guarantee an accurate approximation of the (nonlinear) glucoregulatory system especially when the model is required for a wide operating range.

Nonlinear modeling approach Figure 1 depicts the considered glucoregulatory system and the controller. In this example, the input to the system is the insulin delivered to the patient, and the output is the glucose level.

The objective of this work is the identification of

Figure 1. Schematic representation of the considered

nonlinear models to describe the glucoregulatory

glucoregulatory system (S). A model-based controller (C)

system, based on input-output data. Block structures

is needed to regulate the level of glucose in the patient’s blood.

(Wiener and Wiener-Schetzen models, [2,3]) as well as nonlinear state-space representations (both polynomial and neural networks-based state-space models, [4,5]) are considered to model both the dynamics and the nonlinear behavior of the insulin-glucose system. The different methods have been successfully applied on two identification examples, yielding a significant improvement in terms of accuracy when compared with linear dynamic models. In

64

Short Description of the Research Projects/ Team

particular, the obtained Wiener models are very accurate (which is essential to guarantee that no damage is caused to the patient), and simple enough to make the controller implementation easy. As shown in Figure 2, the Wiener model (with neural networks to model the nonlinearity) approximates very well the input-output behavior of the system.

Figure 2.

Results obtained with the NN-Wiener model. True output (solid black line) and modeled output (dashed

grey line) are shown

References [1] 1. A. Abu-Rmileh, J. Schoukens. "Frequency domain analysis of nonlinear glucose simulation models", 8th IFAC Symposium on Biological and Medical Systems, Budapest, Hungary, 2012. [2] 2. P. Crama, J. Schoukens, "Initial estimates of Wiener and Hammerstein systems using multisine excitation", IEEE Transactions on Instrumentation and Measurement, vol. 50, pp. 1791–1795, 2001. [3] 3. K. Tiels, J. Schoukens. "Identifying a Wiener system using a variant of the Wiener G-Functionals", 50th IEEE Conference on Decision and Control and European Control Conference (CDC-ECC11), Orlando, FL, USA, 2011. [4] 4. J. Paduart, L. Lauwers, J. Swevers, K. Smolders, J. Schoukens, R. Pintelon, "Identification of nonlinear systems using polynomial nonlinear state space models", Automatica, vol. 46, pp. 647–656, 2010. [5] 5. A. Marconato, J. Sjöberg and J. Schoukens, "Initialization of nonlinear state-space models applied to the Wiener-Hammerstein benchmark", Control Engineering Practice, vol. 20, no. 11, pp. 1126-1132, 2012.

2.2.2.8

Detection and Quantification of the Influence of Time Variation in Frequency Response Function Measurements Using Arbitrary Excitations, (R. Pintelon, Ebrahim Louarroudi, and John Lataire)

The linear time-invariant (LTI) framework has proven to be very successful for describing real life systems in all kinds of applications such as prediction/control in the process industry, structural health monitoring in civil engineering structures, modal analysis in mechanical engineering,

65

Annual report ELEC 2012

monitoring of the corrosion of metals, … Therefore, nonparametric frequency response function (FRF) estimates are very useful to get quickly insight in the dynamic behavior of complex systems that are only approximately linear and time-invariant. While the impact of nonlinear distortions on classical FRF measurements is well understood, much less is known concerning the influence of time-variation on FRF estimates. The goal of this project is to fill this gap. Recently it has been shown that time-variant effects act as frequency-correlated noise in classical FRF measurements using random excitations [1]. Therefore, it is difficult to distinguish them from the leakage errors: in [1] the presence of time-variant effects and leakage errors in classical FRF measurements can be detected, but they cannot be quantified. In this paper we present a method for quantifying the noise level, the leakage errors, and the level of the time-variant effects in FRF estimates from one experiment with a random excitation. It is based on the equivalence between a single-input, single-output linear time-variant system and a multiple-input, single-output (MISO) linear time-invariant system. Therefore, classical nonparametric methods for estimating the FRF of MISO LTI systems can be used to quantify the time-variant effects. A similar idea has been used in [2] to model periodically time-varying systems excited by periodic excitations. In this project we estimate the best linear time-invariant approximation of arbitrary time-varying systems excited by arbitrary excitations. Compared with the classical FRF estimate via cross- and autopower spectra, SA (

)=

(

)/

(

), the proposed method (i) estimates the best linear approximation with a

(significant) smaller uncertainty; (ii) separates the leakage errors from the time-variant effects; and (iii) quantifies the approximation error of the time-invariant framework. First, the whole idea has been elaborated for known input, noisy output observations of timevariant systems operating in open loop [3]. Next, the theory has been generalized to noisy inputoutput measurements of time-variant systems operating in closed loop [4]. The performance of the method is illustrated on the time-varying electronic circuit shown in Fig. 1. Figure 2 shows the estimated best linear time-invariant approximation. From Fig. 3 it can be seen that its variance is much smaller than that of the spectral analysis estimate, especially in the frequency band where the influence of the time-variation is significant.

66

Short Description of the Research Projects/ Team

Figure 1.

Time-varying electronic circuit with input

), output

), and scheduling parameter

). It is a second

order bandpass filter with time-varying resonance frequency and damping ratio consisting of the following components: a high gain operational amplifier (CA741CE); a JFET transistor (BF245B); three resistors ( 10 k

and

= 470 ); and two capacitors (

=

= 10

).

=

Figure 2. Estimated best linear time-invariant approximation

Figure 3. Spectral analysis estimate

(black), its variance (light grey), and its bias (dark grey) –

time-invariant approximation (BLTI) – time-varying electronic

time-varying electronic circuit.

circuit. Black: estimate

var

SA (

SA

BLTI (

) of the best linear

); dark grey: variance estimate

)); light grey: difference

black dashed line: var(

SA

)).

SA (

)

BLTI (

)| ; and

References [1] J. Lataire, E. Louarroudi, and R. Pintelon, “Detecting a time-varying behavior in frequency response function measurements,” IEEE Trans. Instrum. and Meas., vol. 61, no. 8, pp. 2132-2143, 2012. [2] E. Louarroudi, R. Pintelon, and J. Lataire, “Nonparametric tracking of the time-varying dynamics of weakly nonlinear periodically time-varying systems using periodic inputs,” IEEE Trans. Instrum. and Meas., vol. 61, no. 5, pp. 1384-1394, 2012. [3] R. Pintelon, E. Louarroudi, and J. Lataire, “Detection and quantification of the influence of time-variation in frequency response function measurements using arbitrary excitations,” IEEE Trans. Instrum. and Meas., vol. 61, no. 12, pp. 3387-3395. [4] R. Pintelon, E. Louarroudi, and J. Lataire, “Detection and quantification of the influence of time-variation in closed loop frequency response function measurements,” IEEE Trans. Instrum. and Meas., accepted for publication.

67

Annual report ELEC 2012

2.2.2.9

Brain Tissue Differentiation and Characterization using Piezoelectric Actuators driven by multisine excitation (David Oliva Uribe, Ralf Stroop, Johan Schoukens and Joerg Wallaschek)

This project is related to the estimation of the mechanical parameters of biological tissues (in particular brain tissue) using a piezoelectric tactile sensor (shown in Figure 1). The main purpose is to provide the tactile sensor a reliable measurement procedure for the differentiation of two biological materials with slightly differences (e.g. tumour and healthy tissue in brain). Potential applications of this research can be addressed in the development of assisting tools for intraoperative tumour delineation and tumour resection in neurosurgery, where is intended to help neurosurgeons with the difficult task of determining tumour boundaries during resection of brain tumours. In order to have a reliable instrument that can be used in surgical procedures, it is necessary to enhance the capabilities of the tactile sensor system. This project involves the improvement in measurement time and accuracy of the sensor system using multisine excitation. In addition, we aim to provide the sensor system with the function to characterize the mechanical properties by the estimation of viscoelastic parameters using system identification techniques.

Effect of the load

Piezoelectric bimorph tactile sensor

Figure 1. Left: Piezoelectric tactile sensor. Right: Comparison of two frequency response functions to show the effect of the load in the sensor system. Contact with soft material will lead to changes in the maximum amplitude and the resonance frequency.

At current, the tactile sensor has been successfully tested with tissue phantoms and ex vivo animal samples where it has been shown the capability of the sensor to differentiate minimal differences in tissue consistency. It is expected to perform in the short term, measurements on human pathological samples and the implementation of the measurement setup directly in the operation theatre.

68

Short Description of the Research Projects/ Team

This project is done in cooperation between the VUB dept ELEC, the Institute of Dynamics and Vibration Research of the Leibniz University of Hannover and the Neurosurgery Clinic of the University Hospital of Bochum.

Figure 3. Tactile sensor in contact with

Figure 2. Electromechanical model of the piezoelectric tactile sensor.

animal brain tissue.

2.2.2.10 Fast algorithms and software for weighted mosaic Hankel low-rank approximation (K. Usevich and I. Markovsky)

Introduction Structured low-rank approximation is a prototypical problem in systems, control, and signal processing. Problems of model reduction, system identification, approximate deconvolution, distance to controllability, etc., can be posed and solved as a structured low-rank approximation for a suitably chosen structure, rank, and approximation criterion (see [1] for an overview). The structured low-rank approximation (SLRA) is formulated as follows: Given

p

R

np

, structure

S , norm

minimize p np

and natural number

r 24h). However, the short term.

step of the heat input (q)

response is less accurate and a constant heat input is required. The numerical techniques can handle both long and short time scales and variable heat input. However, both analytical and numerical techniques start from an approximate heat conduction equation, where in general the heat convection is neglected for simplification purposes. The new approach estimates the relation between the load and the temperature with a rational model (black-box) in the frequency domain.

Where and

) and

( )=

1+

( )

) the Discrete Fourier Transform of the measured heat and temperature signal,

represents the Laplace variable.

No assumptions on the parameters of this model are be made. This approach will lead to a model that reflects the relation between the temperature and the load present in the measurement data. The interpretation of the estimated model is less intuitive than the one found with the standard analytical and numerical approaches, but the model is expected to be more accurate on the short time scale (< 24h). In a following step, this more accurate model can be used to develop a control strategy that minimizes the energy consumption.

References [1] Mogensen, P. 1983. Fluid to Duct Wall Heat Transfer in Duct System Heat Storages. Proceedings of the International Conference on Subsurface Heat Storage in Theory and Practice. Swedish Council for Building Research. June 6-8. [2] Gehlin, S., 2002, Thermal response test: method development and evaluation, PhD thesis, Luleå tekniska universitet. [3] Yavuzturk C., Spitler J.D., Rees S.J., 1999, A Transient Two-Dimensional Finite Volume Model for the Simulation of Vertical U-Tube Ground Heat Exchangers, ASHRAE transactions, Vol. 105, 2, p 465-474.

82

Short Description of the Research Projects/ Team

2.3.2.5

An identification algorithm for parallel Wiener-Hammerstein systems (Maarten Schoukens and Yves Rolain)

Introduction An

identification

method

for

parallel

Wiener-

Hammerstein systems (see Figure 1) is presented. The proposed approaches

identification presented

method in

[1,

combines

2],

where

the

parallel

Hammerstein systems and parallel Wiener systems are considered, with the approach presented in [3] where the identification single branch Wiener-Hammerstein

Figure 1. A 3-branch parallel Wiener-Hammerstein system:

a

MIMO

static

nonlinear

block

f(x)

sandwiched in between the LTI blocks Hi(q) and Sj(q). The noise source v(k) is additive colored noise.

systems is considered.

Model and method The model class: A block-structured parallel Wiener-Hammerstein model as shown in Figure 1 is considered. The static nonlinear blocks are modeled using polynomial functions. The LTI blocks are represented with a parametric rational function in the Laplace variable s or the discrete time shift operator q 1. Noise framework: An additive noise perturbation at the output of the device under test only is considered. The identification framework: uses a black box approach

starting

from

input-output

measurements of the system only. First, the overall dynamics are estimated in least squares sense. Second, these dynamics are decomposed [1, 2], and partitioned [3] giving an estimate of the LTI blocks. Finally, the static nonlinearities are

estimated

using

a

linear

least

squares

estimator.

Figure 2. Output spectra of the simulated parallel Wiener-

Results: The identification method is tested on a simulation example, giving promising results as is shown in Figure 2. The proposed method provides good initial estimates, which are refined in a

Hammerstein system. Dark gray line: output of the parallel Wiener-Hammerstein system. Grey dots: error of the initial estimate of the parallel Wiener-Hammerstein model, light gray

circles:

error

of

the

refined

parallel

Wiener-

Hammerstein model.

nonlinear optimization step. The refined model has an error of -60 dB on the validation data, which corresponds to the noise level in the simulation example.

83

Annual report ELEC 2012

Conclusion An identification method for parallel Wiener-Hammerstein systems is proposed. To the author’s knowledge, This is the first method that can estimate a model with a parallel Wiener-Hammerstein structure. The proposed method gives promising results on different simulation examples.

References [1] M. Schoukens, R. Pintelon, and Y. Rolain. Parametric Identification of Parallel Hammerstein Systems. IEEE Trans. Instrum. Meas., 60(12):3931 – 3938, 2011. [2] M. Schoukens and Y. Rolain. Parametric Identification of Parallel Wiener Systems. IEEE Trans. Instrum. Meas., 61(10):2825 – 2832, 2012. [3] J. Sjöberg and J. Schoukens. Initializing Wiener-Hammerstein models based on partitioning of the best linear approximation. Automatica, 48(2):353 – 359, 2012.

2.3.2.6

Error bound on polynomial approximation of steep functions (Diana Ugryumova, Gerd Vandersteen, Rik Pintelon)

System identification is an important research field for a better understanding of dynamic systems. It is used to build an accurate model of a system that can be used for control. A distillation column is an example of such a system. One of the challenges in system identification is the reduction of errors due to transient effects and disturbing noise. A nonparametric transfer function estimation method was developed recently, called local polynomial method (LPM) [1]. This method approximates the transfer function and the transient contributions locally by a polynomial in frequency. For slowly varying transfer functions a low-order polynomial is sufficient. The results are at least as good as those of existing time-domain windowing methods. For quickly varying transfer functions, like integrators, the errors of the polynomial approximation increase at the low frequencies. This research is based on [2], where the bound for the error of polynomial approximation was derived for a symmetrical interval. Here, the result is extended for a nonsymmetrical polynomial approximation interval. This error bound is proportional to the relative bandwidth (between the one used for the local polynomial approximation and the bandwidth of the system) to the power related to the chosen order of the local polynomial: =

1

represents the bandwidth of the local polynomial approximation, 3dB bandwidth starting from the first excited frequency approximation, and

corresponds here to the

, R is the order of the local polynomial

is a constant related to the number of frequency lines used and the order of

the local polynomial approximation. The constant is calculated in the same way as in [2].

84

Short Description of the Research Projects/ Team

As an example, we explore the possibility of distinguishing between a pure integrator, a first-order transfer function with a large time constant (“slow” pole), error bound on simulation data.

=

=

, and

, using the above

is here the Laplace variable. We assume that there is no noise

and/or transient present. For a polynomial of order two we found the error bounds on the approximation of both frequency responses. The absolute value of the difference between the two frequency responses is compared to the sum of the error bounds and shown in Fig. 1. In this figure, the difference between the frequency responses is larger than the sum of error bounds in the frequency region from 0.02 to 0.07 Hz. The green line indicates that our measurements should contain at least frequency of 0.07 Hz or lower in order to be able to see the difference between the integrator function and a first-order function with a pole at 0.03 Hz. This is again shown in Fig. 2 together with the frequency responses in the frequency region from 0.02 to 0.2 Hz. This lowest frequency of 0.07 Hz is directly related to the total measurement time, which in this case should be at least 14.3 seconds. More research needs to be done to find out how conservative this error bound is.

Magnitude

cannot differentiate!

Phase (deg)

Figure 1. Comparison

of

the

absolute

value

of

the

difference between the frequency responses (black) and

Frequency (Hz)

the sum of the error bounds (dashed).

Figure 2. Frequency responses of a pure integrator (black) and a first-order rational function with a pole at 0. 03 Hz (red). In order to differentiate between the two frequency responses, the lowest frequency that we should measure is 0.07 Hz or lower (green line).

References [1] R. Pintelon, G. Vandersteen, J. Schoukens, and Y. Rolain, “Improved (non-)parametric identification of dynamic systems excited by periodic signals - the multivariate case,” Mechanical Systems and Signal Processing, vol. 25, no. 8, pp. 2892–2922, 2011. [2] J. Schoukens, G. Vandersteen, R. Pintelon, Z. Emedi, Y. Rolain, “Bounding the polynomial approximation errors

of

frequency

response

functions”,

I2MTC

2012,

IEEE

International

Instrumentation

Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pages 1395 – 1399.

85

and

Annual report ELEC 2012

2.4

TEAM D: MEDICAL MEASUREMENTS AND SIGNAL ANALYSIS (M2ESA)

2.4.1

Introduction to project M2ESA

Problem statement Physicians have an enormous amount of practical experience which is most of the time based on old medical principles. The public opinion requires that diagnoses are accurate and infallible. To obtain this for all medical cases, the actual way of working which relies on experience becomes insufficient. Hence, a more robust, rigorous and well-founded approach is desired. Step by step, the medical world realizes the need and advantages of mathematical models in the quest for accurate diagnoses. Although the modelling science is a rather young part of science, the techniques which the modelling grocery has to offer are so abundant that without proper guidance one cannot see the wood for trees. This is the main reason why the medical community is rather reluctant to use the advanced modelling methods.

Objectives In the

M 2 ESA

project, the measurement and modelling knowledge available at the department

ELEC will be combined. The successful marriage between practice and theory offers the medical world simple, practical, risk-free and powerful tools to satisfy the users’ needs. A trade-off between theoretical/mathematical optimality and user-friendliness will be studied. Indeed, it is important that these emerging modelling techniques can be safely applied and interpreted by the layman user. Some of the projects that

M 2 ESA

focuses on, are the following:

Accurate oscillometric blood pressure measurements: One of the most popular medical instruments used at home is the automatic non-invasive blood pressure meter (NIBP). Most medicine cupboards contain one and a lot of people use it on a daily basis. A cuff is wrapped around the arm of the patient and inflated until the circulation stops. Most classical automatic blood pressure meters are based on the oscillometric principle, which records the oscillations in the cuff pressure during deflation of the cuff. Out of this oscillometric waveform a mean arterial pressure (MAP) as well as a systolic and diastolic pressure is deducted by means of a mathematical algorithm. Each manufacturer of blood pressure meters, however, has developed his proper algorithm which is most of the time patented. On top of that, these algorithms are often not scientifically founded or transparent.

86

Short Description of the Research Projects/ Team

Non-invasive glucose measurements: In the EU and US, 7,8% of the population have diabetes (www.diabetes.org), of which up to one third undiagnosed. Up to 20% of the population has a variable degree of glucose intolerance, and might also benefit from early glucose monitoring. Management of diabetes, in particular insulin-dependent forms, requires intensive control of blood glucose. Currently this is done by pricking the fingertip to draw blood and measure capillary blood glucose on an external sensor strip. This is a time-consuming, relatively painful procedure and offers only discontinuous monitoring. Developing a non-invasive glucose measurement procedure would be a considerate relief in the social, the physical as well as the financial field for these patients. Earlier attempts to generate a similar application have failed due to the lack of a solid background on the measurement concept. Functional Magnetic Resonance Imaging (fMRI): MRI and fMRI is daily used to explore tumour tissue, infections, brain problems, muscles diseases, etc... Analysing functional MRI data is often a hard task due to the fact that these periodic signals are strongly disturbed with noise. In many cases the signals are even buried under the noise and not visible, so that detection is quite impossible. In the past, different modelling approaches have been used without proper validation and comparison. Vital signs of foetuses: The premature mortality rate is still very high although the precaution and advances in the medical world. The current society puts a large responsibility on the shoulders of gynaecologists. Therefore it is important to have proper access to the foetus’ vital signs. This is not only a major measurement challenge but also requires powerful signal analysis techniques. The main problem is that all non-invasive measurements are indirect, very noisy and on top of that nonlinear. Breast cancer: The actual detection of breast cancer, called mammography, is based on Xrays and is a very painful experience for the patient. New detection techniques based on low-power microwaves are emerging. The advantages are apparent: no breast compression, no ionizing radiation, no risk of heating the breast tissue. These new techniques can benefit from a strong signal analysis and modelling to reconstruct the images of interest.

2.4.2

Introduction to project MEMON

Problem statement Wireless communication systems are widely available in our daily life, from mobile communications to cognitive radios and wireless ad hoc/sensor networks. In order to achieve a standard performance, more requirements are put on the radio frequency (RF) systems. Signals used in today’s wireless systems are characterized by high and fast dynamics in their envelope due to the

87

Annual report ELEC 2012

use of sophisticated modulation techniques. Hence, they have large bandwidths and high peak-toaverage power ratios. As a consequence, high requirements are put on the high power RF transmitters which are considered to be nonlinear systems and on the RF receivers whose digital bandwidths are limited due to shortage in high speed analog-to-digital converters, based on acceptable measurement accuracy. One of the most important components in wireless communication systems is the radio frequency power amplifier (PA). Due to the high linearity and efficiency requirements, it is of the highest importance to accurately characterize the linear as well as the nonlinear behaviour. Characterizing the linear behaviour of RF components is a well-known technique. However, characterizing the nonlinear behaviour of these components is a completely different and more involved problem. The new high efficiency PA and transmitters, based on switched mode amplifiers, are emerging technologies that certainly require novel measurement methods.

Objectives In this research project, excitation signals will be designed in order to reduce the requirements on the linear behaviour of RF transmitters. New sampling and reconstruction strategies are developed which allow measuring accurately wide-band waveforms based on relatively slow analog-to-digital converters. Furthermore, the different measurement and modelling techniques that are nowadays available to analyse the nonlinear behaviour of RF devices will be analysed. A zillion methods have been published and used, but what is clearly missing is guidance for the user. This study will clarify why and how all these methods differ from each other based on either a system theoretical or stochastic point of view. Will all the different techniques be able to co-exist or is there a need for a new evolution? This research is in collaboration with Prof. Niclas Björsell of the University of Gävle (Sweden) and is funded by a research grant of the Research Foundation-Flanders (FWO).

2.4.3

Short Description of the Research Projects of TEAM D

2.4.3.1

A simple non-parametric pre-processing technique to correct for non-stationary effects in measured data (Kurt Barbé, Wendy Van Moer, Lieve Lauwers and Niclas Björsell)

In many measurement applications, one aims at characterizing the dynamic behavior of a deviceunder-test, observing periodic component buried in the observed signal, studying the coloring of the disturbing noise sources. The standard technique to reach this objective is the measurement of the Frequency Response Function (FRF). Hence, the measurement of the FRF can be considered as a fundamental measurement topic. Looking at the fundamentals of FRF measurements, it is clear

88

Short Description of the Research Projects/ Team

that the classical technique to obtain the FRF operates under strict conditions. Departures of these conditions may lead to systematic errors in the measured FRF. Classical methodologies available in literature to compute the FRF operate under the assumption that the measured signals are stationary, such that the underlying signal’s properties remain invariant over time. This assumption is clearly violated when dealing with time-varying systems. However, also for time-invariant systems the stationarity assumption can become invalid due to the measurement process/errors. This is for instance the case when the following three scenarios occur: 1. Transient effects due to initial and final experimental conditions; 2. Missing data due to sensor failure; 3. Amplitude variations of the excitation signals due to actuator problems. Transient effects typically arise when a

0

finite measurement time-window is used.

-20 Amplitude

Missing data

occur when the

sensor

-40

performing the measurements incorrectly

-60

registers or simply fails to register the

[dBm]

measurements. Actuator problems are

-80 -100 0

present when the actuator fails to apply

50

100

150

200

Frequency [MHz] Figure 1.

the intended excitation signal to the system. One way to tackle this problem is by measuring the actual applied input

Signal fall-out, no automatic sensing algorithm:

stationary output spectrum (black), corrected output spectrum (light gray), uncorrected output spectrum

signal. Unfortunately, this is for many applications difficult or even infeasible (e.g., for wireless systems).

(dark gray)

Hence, actuator problems lead to systematic differences between the applied and the intended excitation. In these three scenarios, the measurement engineer needs to take into account the influence of the non-stationary effects, since they introduce systematic errors in the measured Frequency Response Function (FRF) of the DUT. Different methods exist to deal with each of the abovementioned non-stationary effects specifically. However, most of these techniques require a parametric model, on top of the use of time windows to eliminate transients. For example, in the time series analysis literature, the problem of missing data is dealt with by representing the missing data as additional parameters or unknowns. As a result, knowledge of the time interval where the non-stationary effect appears is required. In this paper, we propose a general technique to correct for non-stationary effects in the measured data, based on processing overlapping sub-records. The technique is based on the Welch method for non-parametric power spectrum estimators to suppress transient influences due to nonstationary effects present in the data.

89

Annual report ELEC 2012

2.4.3.2

Fractional order time series models for extracting the haemodynamic response from functional Magnetic Resonance Imaging data (Kurt Barbé, Wendy Van Moer and Guy Nagels)

Functional Magnetic Resonance Imaging (fMRI) is without doubt one of the most popular techniques to map the brain functions. Its popularity follows from the fact that this technology is relatively cheap and completely non-invasive. In general, fMRI can be discriminated in two objectives: (i) signal detection and (ii) characterization of the haemodynamic response function. The first objective deals with detecting those areas of the brain that reveal an increased neural activity upon stimulus. The second objective deals with describing the changes in the blood flow upon stimulus. Indeed, upon stimulus the activated areas of the brain require more oxygen and glucose to operate which results in changes in the blood flow known as the haemodynamic response. Besides these two main objectives, looking

1.5 Signal

at recent fMRI research the contributions

1

can be partitioned into (i) hardware and

intensity

instrumentation, (ii) image enhancements

0.5

and pre-processing, (iii) denoising and noise 0 0

characterization, (iv) post-processing and 5

10

15

20

25

30

Times [s] Figure 1.

Modeled

haemodynamic

strongly activated

estimation of the haemodynamics). This paper’s contribution deals with the post-

response

voxel (simulated

SNR:

of

a

2)

–

noisefree (black), FIR (blue), IIR (Magenta), SPM

processing

and

estimation

of

the

haemodynamics.

software (green), Fractional model (red)

Characterization of the haemodynamics is of importance for neuropsychology, to monitor neurological diseases… The haemodynamics provide access to physiological processes in the brain and can be linked to for instance: muscle reaction times, observing the glucose and oxygen levels in the brain, … A second reason to model the haemodynamics is the possibility of simulating the neurological processes of the brain. As a result the modeling of the haemodynamic response requires a model structure that allows extracting physiological information. The main choice, faced when choosing a model for hemodynamic response, is the parameterization or basis set of the hemodynamic response function.

Current choices

Figure 2. Significantly detected voxels in an odd-ball

range between two extremes. At one extreme, finite

experiment by means of the fractional model. The left

impulse response models, Random-effect polynomial

top plot is the back view, the right top plot is the top

Analysis of Variance, Fourier and Wavelet analysis,

view and the bottom plot is the side view. The level of

90

activation varies from weakly (yellow) to strongly (red)

Short Description of the Research Projects/ Team

Volterra kernels have a large number of parameters and impose minimal constraints on the shape of the hemodynamic response function. These provide a very good fit to the data. However, in most instances, they are over parameterized and over fit responses. At the other end of the spectrum, there are physiologically informed (canonical) hemodynamic response functions, whose basic shape is based on the known form and time constants of underlying physiology: Poisson filter, Balloon model, Gamma and Gaussian filter and the double-gamma function. This means they have only two or three parameters and provide a less accurate fit to the data. However, their simplicity precludes over fitting. In brief, the choice of models depends upon the amount of biophysical knowledge one wants to use to constrain the models - in the trade-off between accuracy and complexity. In what follows, we introduce a biophysically informed model that generalizes the canonical forms currently used and yet provides a more flexible fit to the data. The only disadvantage of the model structure is the need of a non-convex optimization. It is well known that non-convex problems give rise to various local optima such that high quality starting values are required. To obtain a high quality starting value, we apply a method to transform an integer order model to a fractional order model. This fractional order model may serve as a good initial solution of the nonlinear least squares problem.

2.4.3.3

Fractional models for modeling complex linear systems under poor frequency resolution measurements (K. Barbé, J. Olarte Rodriguez, Wendy Van Moer and Lieve Lauwers)

Modeling the transfer function of a linear system based on noisy output observations with an errorfree input remains an important issue of statistical signal processing applications. One of the key model structures to obtain a parametric transfer function model is the rational form. This is due to the fact that the pole/zero configuration of the rational form provides the user some physical insight (e.g. damping of the system). Thus, for linear systems the model structure selection is often not a problem but rather a matter of preference or application. Indeed, modal forms or partial fraction decompositions are popular for mechanical applications, analogue filter design for instance often relies on the use of pole-zero-gain models, improving the numerical stability and accuracy advocates the use of orthogonal polynomials and state-space models are efficient for control engineering applications. Once the model structure is chosen a model order selection problem needs to be solved together with the model parameter estimation. Two popular model selection criteria are AIC and MDL. For a sufficiently good frequency resolution, the MDL tends to reveal the true model order with probability one given the correct choice of model structure. The AIC is known to select a model order that is larger than the true one. For problems linear in the parameters, the Mallows criterion is frequently used which is shown to be unbiased but not necessarily consistent for improving frequency resolutions.

91

Magnitude [dB]

Annual report ELEC 2012

50

The parameter estimation problem can be

0

solved with various fairly-related methods: a maximum likelihood

-50

approach

when the

probability density function of the disturbing noise sources is known; the (nonlinear) least

-100

squares approach when the spectral density -150 1.5

2

2.5 3 Frequency [log ]

3.5

4

10

Figure 1.

Illustration of the identification procedure for a

of the noise signals is assumed to be flat; or a

weighted

(nonlinear)

least

squares

SNR of 100 dB. The FRF is given by the black cross

approach for noise signals with a colored

markers, the green solid curve is the fractional order

spectral density.

model, the AIC selected integer order model is given by the blue curve. The error curves between the model and the FRF is given by the dashed curves

It is observed that when model order selection methods suggest a large model complexity, these systems can be modeled by a fractional order system which reduces the model order significantly. This observation implies that large integer order systems may hold a fractional behavior. In this paper, we show that a group of poles and zeros which cannot be completely resolved due to the frequency resolution, can be replaced by a fractional order pole or zero. This observation bears an important consequence: if such a system is modeled by an integer order system, the poles and zeros are so close to each other that the frequency resolution is too poor to resolve them. This implies that the condition number of the regression matrices is very high which may result in an increased lack of fit. By replacing such a group of poles and zeros by a fractional order pole or zero, the frequency resolution becomes sufficiently good such that the condition number of regression matrices is reasonable and the asymptotic theory of statistical system modeling holds. Fractional order modeling is a fast emerging engineering field due to its enhanced frequencydomain flexibility to model the data. Visionary publications postulate fractional order models as the systems/models of the 21st century. The main advantages of the proposed method in this paper are three-fold: (i) high order models can be replaced by a low order fractional model which enhances the frequency resolution, (ii) the method is fully automatic which detects possible fractional behavior, compresses a large integer order model to a low fractional order model and finally identifies the fractional order model parameters and (iii) the method can be used for any kind of system fractional or integer.

92

Short Description of the Research Projects/ Team

2.4.3.4

Discriminant Analysis for an automatic signal detection technique useful in cognitive radios. (Lee Gonzales Fuentes1,2, Kurt Barbe1, Wendy Van Moer1,2, and Niclas Bjorsell1,2) 1 2

Dept. ELEC/M2ESA, Vrije Universiteit Brussel-Brussels, Belgium University of Gavle- Gavle, Sweden

The efficient use of a natural resource like the electromagnetic spectrum can be achieved by means of cognitive radio. Cognitive radio is an intelligent wireless radio system which is aware of its environment and adapts its communication parameters to the variations of its surrounding [1]. Since it dynamically monitors the unused spectrum, it can relinquish it to a secondary user when free or transferring it when requested by the primary user such that in any case, interference between users is prevented. Signal detection of primary users, thus, enables spectrum use agility. In normal operation conditions, the sensed spectrum is nonflat i.e. the power spectrum is not constant. A novel method [2] based on statistical test is extended towards nonflat spectrum [3]. The method proposes the segmentation of the measured spectra into regions where the flatness condition is approximately valid. As a result, an automatic detection of the significant spectral components together with an estimate of the magnitude of the spectral component and a measure of the quality of classification becomes available as seen in Fig. 1. The aim of this project is to optimize the methodology for signal detection in cognitive radios such that the probability that a spectral component was incorrectly classified is iteratively reduced. Simulation and measurement results showed the advantages of the presented technique particularly in different types of spectra. Due to its superior advantages compared to other methods, this method can be useful for spectrum sensing in cognitive radios.

0

-10 -10

-20 -30

-30

Magnitude [dBm]

Magnitude [dBm]

-20

-40 -50 -60

-40 -50 -60

-70

-70

-80

-80 0

1

2

3 4 Frequency [GHz]

5

0

6

1

2 3 4 Frequency [radians/sample]

5

6

Figure 1. Automatic signal detection for a multisine with uniform amplitude and an SNR of -10 dB before (left) and after (right) updating the segment bounds: (gray) amplitude of frequency lines, (bold dark curve) discrimination curve, (black circles) center of discriminant heights, (solid gray line) discrimination heights, and (black arrows) eight signal tones.

93

Annual report ELEC 2012

References [1] S. Haykin, “Cognitive Radio: Brain-Empowered Wireless Communications,” IEEE Journal on Selected Areas in Communications, vol. 23, no.2, pp. 201-220, Feb. 2005. [2] K. Barbé, and W. Van Moer, “Automatic Detection, Estimation, and Validation of Harmonic Components in Measured Power Spectra: All-in-One-Approach,” IEEE Trans. Instrum. Meas., vol. 60, no. 3, pp. 10611069, Mar. 2011. [3] L. Gonzales Fuentes, K. Barbé, W. Van Moer, and N. Björsell, “Cognitive Radio: Discriminant Analysis finds the free space,” IEEE International Instrumentation and Measurement Tehcnology Conference I2MTC 2012, Graz, Austria, vol., no., pp. 2242-2247, May 13-16, 2012.

2.4.3.5

Characterization of the probability density functions of measured signals. (Lee Gonzales Fuentes1, Kurt Barbe1, Lee Barford2, and Wendy Van Moer) 1 2

Dept. ELEC /M2ESA, Vrije Universiteit Brussel-Brussels, Belgium Measurement Research Laboratory, Agilent Technologies- Nevada, USA

Graphical representation can provide crucial information regarding the data in a user-friendly format. However, simply plotting data points can lead to a wrong perception of the data information, especially when the number of variables increases. Plots of functions of these data points allow the data to exhibit all its properties. The probability density function describes how any continuous random variable is distributed. Its importance lies on the many statistical properties and applications that can be obtained from it. In practice, the distribution of any observable variable is unknown. 30

0.05 Histogram Gaussian kernel

25

0.045

Histogram Gaussian kernel

0.04 20

0.035

15

0.025

0.03

0.02 10

0.015 0.01

5

0.005 0 -0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0 100

0.25

150

200

250

density

300

350

function

of

400

the

450

Figure 1. Probability density function of the jitter in a digital

Figure 2. Probability

waveform: histogram (bars) and a Gaussian kernel density

transition intervals: histogram (bars) and a kernel density

jitter

inter

(red curve).

estimator (bold curve).

Most instruments automatically generate an estimate of the probability distribution of a measured continuous physical variable as a histogram. The fact that the shape of the histogram is dependent of the choice of the bin edges makes its plot to be discontinuous for continuous data showing unnecessary or nonexistent details. Statisticians have developed a set of techniques, called density estimation, to overcome these limitations. These estimators construct a smoother and often more accurate estimate of the probability distribution compared to the histogram [1], [2]. This project

94

Short Description of the Research Projects/ Team

brings out the benefits and limitations of these techniques, particularly their finite-sample robustness. Upon exploring these properties, these techniques may be optimized such that they can be used by the measurement engineer.

References [1]

B. W. Silverman, Density Estimation for Statistics and Data Analysis, Chapman and Hall, 1986.

[2] W. Navidi, Statistics for Engineers and Scientists, 3rd ed. Mc Graw Hill, 2011.

2.4.3.6

A robust signal detection method for fMRI data under correct Rice conditions (Lieve Lauwers, Kurt Barbé and Wendy Van Moer)

Introduction The goal of functional Magnetic Resonance Imaging (fMRI) is to detect regions in the human brain that show significant neural activity upon stimulus. To do so, a statistical analysis of the data needs to be performed in which the hypothesis that no signal is present (i.e., the null hypothesis) is tested. The main problem of the classical statistical tests is that they are based on the wrong assumption that fMRI data are normally distributed. It is however known from literature that fMRI data are Rice distributed [1], [2]. For high signal-to-noise ratios (SNRs), the Rice distribution converges to a Gaussian distribution [3]. Hence, for high SNRs the performance of the classical statistical tests will hardly suffer from the normality assumption. Unfortunately, in fMRI applications low SNRs are omnipresent such that the characteristics of the Rice distributed data should be taken into account in order to accurately analyze fMRI data. To conclude, it is important to develop statistical methods that yield reliable results in all circumstances.

Classical versus Rice-corrected t-test In [4], a test statistic (under the null hypothesis) is presented for an fMRI data set , in case of a known (square wave) fMRI paradigm

with

and an unknown standard deviation (std) : (eq.1)

=

an estimate of the std. Since the t-test given in [4] simply ignores the true Rician distribution of the data, it is in

general not suited for fMRI data. Using (eq.1), the user will not be able to detect small intensity variations due to neural activity and, hence, perform a wrong data analysis. We propose a correction for this t-test by taking into account the correct Rice conditions, resulting in the Rice-corrected t-test =

4

: (eq.2)

2

95

Annual report ELEC 2012

By comparing (eq.1) and, (eq.2) we observe the presence of the factor 2

factor which is independent of

or

. This correction

allows the user to perform a robust statistical analysis of fMRI

data, using a single statistical method.

Simulation experiment The performance of the test statistics (eq.1) and, (eq.2) were compared via a simulation experiment in which we determined the obtained significance level

of both t-tests. In a Monte

Carlo simulation of M = 10 000 runs, we calculated the test statistic in (eq.1) and, (eq.2) under the

null hypothesis. We specified the significance level

spec

to be equal to 5%. Translated in fMRI

terms, this means that we accept in 5% of the cases a false alarm: a brain element is considered as active, while it is not in reality (type I error). In Fig. 1, we show the resulting histogram from the Monte Carlo simulation of (eq.1) and, (eq.2) given by the black and grey curve respectively. The black vertical line corresponds to the specified significance level of 5%. Qualitatively, we see that the Rice-corrected t-test (eq.2) results in a higher rejection ratio. The cumulative distribution function (cdf) is a more appropriate tool to quantify the simulation results. Fig. 2 shows the empirical cdf, together with the theoretical significance level of 5% (black vertical line) and a horizontal dash-dotted line indicating a cumulative probability of 95%. We see that the significance level and the 95-percentile of (eq.2) perfectly match. This simulation experiment shows that the user-specified significance level

spec

is not obtained by

the classical t-test in (eq.1) while this is achieved by the Rice-corrected t-test in (eq.2). The main consequence of this result is that the rejection region of the null hypothesis will not correspond to the user’s demands, resulting in a test with a decreased detection performance. 0.7

1 classical t-test Rice-corrected t-test

0.6

0.9

classical t-test Rice-corrected t-test

Empirical cdf

Histogram

0.8 0.5

0.4

0.3

0.2

0.7 0.6 0.5 0.4 0.3 0.2

0.1 0.1 0 -4

-3

-2

-1

0

1

2

3

0 -5

4

-4

t Figure 1. Histogram

of

the

classical

-3

-2

-1

0

1

2

3

4

5

t Rice-

Figure 2. The empirical cdf for the classical (black) and Rice-

corrected t-test (grey). The black vertical line denotes the

(black)

and

corrected t-test (grey); the theoretical significance value of

user specified significance value of 5%.

5% (black vertical line); the cumulative probability of 95% (dash-dotted line).

96

Short Description of the Research Projects/ Team

References [1] S. O. Rice, “Mathematical analysis of random noise”, Bell System Technical Journal, Vol. 24, pp. 46 and further, 1945. [2] H. Gudbjartsson, S. Patz, “The Rician distribution of noisy MRI data”, Magnetic Resonance in Medicine, 34(6), pp. 910-914, 1995. [3] C. Carobbi and M. Cati, “The absolute maximum of the likelihood function of the rice distribution: existence and uniqueness”, IEEE Trans. on Instrumentation and Measurement, Vol. 57, No. 4, pp. 682-689, 2008. [4] B. A. Ardekani, I. Kanno, “Statistical methods for detecting activated regions in functional MRI of the brain”, Magnetic Resonance Imaging, Vol. 16, No. 10, pp. 1217-1225, 1998.

2.4.3.7

Odd Random Phase Electrochemical Impedance Spectroscopy (ORP-EIS) for Glucose Sensing (Oscar J. Olarte1, Wendy Van Moer1, Kurt Barbé1 and Yves Van Ingelgem2) 1 2

Dept. ELEC, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium Dept. SURF, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium

Introduction When developing a non-invasive glucose measurement system, one must deal with different noise sources that need to be identified and quantified in order to provide confidence bounds for the estimated glucose levels. One special source of noise is the effect of the Type B uncertainties or errors. These uncertainties depend widely of the environmental effects and/or respond to systematic errors that are “hidden” in the manufacturer’s specification. In order to obtain reliable measurement results, the repeatability and reproducibility of the experimental acquisition must be verified. The most important experimental circumstances that might vary for different measurements in electrochemistry are [1]: the electrolyte, the temperature of the electrolyte, the position of the electrodes and the electrodes.

Effect of the type B uncertainties in ORP-EIS Based on P time periods of the excitation signal, are calculated the following characteristics [2]: the BLA of the impedance standard deviation

total

Z BLA ( ) ,

the standard deviation of the noise

Ny

( )

and the total

( ) . In order to calculate the repeatability and reproducibility,

M repeated

experiments are realized and the following characteristics are calculated: the mean impedance Mean Z BLA ( ) M

and the standard deviation of the impedance

Ny MMean ( )

std Z BLA ( ) M

and the mean level of the total standard deviation

97

N total

. The mean level of noise

Mean M

( ).

Figure 1 presents an

Annual report ELEC 2012

example of these calculations. In this figure is shown that the standard deviation mean impedance does not coincide with the mean noise level std Z BLA ( ) M

So the level of the Notice that

std ( ) Z BLA M

std Z BLA ( ) M

of the

Ny MMean ( ) .

can be associated with the reproducibility of the measurements.

is larger than the mean level of the total standard deviation N total

Mean M

( ).

This indicates that the influence of the non controllable conditions is larger than the influence of the non-linearities in the system itself. However, According to our results, the reproducibility is in general over 96%. Impedance Spectra Of NaCl Solution at 700 mg/dL 80

Conclusions

ZMean BLAM

60

The capabilities of an ORP-EIS to perform

Impedance [dB]

40

glucose measurements are studied. Although

ZStd

20

BLAM

the

0

are preliminary,

they

indicate

reproducibility over 96%. Thus, the ORP-EIS

NtotalMean M

-20

results

clearly elucidate the capabilities of the EIS

NyMean M

-40

toward the development of a non-invasive

-60

glucose measurement system. 10

Figure 1.

1

10

2

10 Frequency [Hz]

3

10

4

Impedance spectra of NaCl solution at 700mg/dL for

M=10 experiment repetitions. The dark and light grey lines refer the noise level and the total standard deviation for each experiment.

The thick lines represent the mean values and

standard deviation along the repetitions.

References [1] E. V. Gheem, “A new methodology for electrochemical impedance spectroscopy in the presence of nonlinear distortions and non-stationary behaviour: Application to pitting corrosion af aluminium”, Vrije Universiteit Brussel VUB, 2005. [2] J. Schoukens, R. Pintelon, and T. Dobrowiecki, “Linear modeling in the presence of nonlinear distortions,” IEEE Trans. Instrum. Meas., vol. 16, pp. 786–792, August 2002.

98

Education

3. Education 3.1

THE INTRODUCTION OF THE BACHELOR-MASTER STRUCTURE

Since the academic year 2004-2005 the “bachelor - master” structure (replacing the candidate licentiate structure) has been introduced. The initiative for this thorough interference in the programmes of higher education was the Bologna Declaration. The Ministers of Education of 31 European Countries gave in 1999 the start to uniform the higher education in Europe. Although the Bologna process creates convergence, the fundamental principles of autonomy and diversity are still respected. The aim of the Bologna Declaration is to improve in Europe the exchangeability of degrees, the free mobility of students, quality assurance and a flexible study package by introducing credit systems.

3.2

BRUFACE (WWW.BRUFACE.EU/EN/)

Brussels Faculty of Engineering (in short "Bruface") is an initiative of the two universities in the centre of Brussels. The Université Libre de Bruxelles (ULB) and the Vrije Universiteit Brussel (VUB) jointly offer a broad spectrum of fully English taught master programmes in engineering. Starting from the academic year 2011-2012 Université Libre de Bruxelles and Vrije Universiteit Brussel jointly organise the following English taught Master of Science (MSc) programmes MSc in Architectural Engineering MSc in Chemical and Materials Engineering Options Materials | Process Technology MSc in Civil Engineering MSc in Electromechanical Engineering Options Aeronautics | Energy | Mechatronics-Construction | Vehicle Technology and Transport MSc in Electronics and Information Technology Engineering

99

Annual report ELEC 2012

The Université Libre de Bruxelles (ULB) and the Vrije Universiteit Brussel (VUB) are both active in several international higher education networks including T.I.M.E.1 and UNICA2.

3.3

COURSES LECTURED IN THE FACULTY OF ENGINEERING

3.3.1

Lectures and practical courses Lectures and practical courses 3

Credits

Elektromagnetisme - Prof. A. Barel Electromagnetism 3rd Bachelor of Sciences in Engineering (compulsory)

6

Gevorderde controletechnieken - Prof. R. Pintelon – Prof. Jan Swevers (KUL) Advanced Control 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd Electronic and IT Engineer (optional)

4

Identificeren van dynamische systemen - Prof. R. Pintelon Identification of dynamic systems 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd Electronic and IT Engineer (optional)

4

Netwerken en filters - Prof. R. Pintelon Network Analysis and Synthesis 3rd Bachelor of Sciences in Engineering (compulsory)

7

Systeem- en controletheorie - Prof. J. Schoukens – Prof. R. Pintelon System and Control Theory 3rd Bachelor of Sciences in Engineering (compulsory) 3rd Bachelor of Sciences in Engineering (optional) 3rd Bachelor of Sciences in Engineering (compulsory) 3rd Bachelor of Mathematics (optional) 2nd Master of Sciences in Engineering: Chemistry and Materials (optional) 3rd Chemical and Materials Engineer (optional) Hoogfrequent elektronica en antennes - Prof. Y. Rolain High-frequency Electronics and Antennes 1st Master of Sciences in Engineering : Electronics and IT-Engineering (compulsory) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) Meten en identificeren - Prof. Y. Rolain - Prof. J. Schoukens Measurement and Identification 1st Master of Sciences in Engineering : Electronics and IT-Engineering (compulsory) 2nd Master of Sciences in Engineering : Biomedical Engineering (optional) 3rd Electronic and IT Engineer (compulsory) 3rd Bachelor of Mathematics (optional) 2nd Master of Sciences in Engineering: Chemistry and Materials (optional) 2nd Master of Sciences in Engineering : Electro-Mechanical Engineering (optional) 3rd year Chemical and Materials

1

6 5 3

6

7

Top Industrial Managers Europe: is a network of 51 leading Engineering Schools and Faculties and Technical Universities

which offers, through a system of voluntary bilateral agreements between its members, promotion and recognition of academic excellence and relevance to the international labour market in the form of Double Degrees in engineering and in related fields. 2

Is a network of 42 Universities from the Capital cities of Europe, with a combined strength of over 120,000 staff and

1,500,000 students. Its role is to promote academic excellence, integration and co-operation between member universities throughout Europe. 3

The language of tuition of all these courses is Dutch

100

Education

Lectures and practical courses 3

Credits

Engineer (optional) 3rd Electro-Mechanical Engineer (optional)

3

Bio-informatica en datamining - Prof. Yves Moreau (KUL) - Prof. J. Schoukens Bioinformatics and Datamining 2nd Master of Sciences in Engineering : Applied Computer Sciences (optional) 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd Electronic and IT Engineer (optional)

4

Meten en modelleren van niet-lineaire systemen - Prof. J. Schoukens Measuring and Modelling of Nonlinear Systems 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (optional)

4

Statistiek voor ingenieurs - Prof. J. Schoukens Probability and Statistics 2nd Bachelor of Sciences in Engineering (compulsory)

3

Capita selecta Telecom - Prof. L. Van Biesen Capita Selecta Telecom 2nd Master of Sciences in Engineering : Applied Computer Sciences (optional) 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Computer Science (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (optional)

3

Industriële meetomgevingen - Prof. L. Van Biesen Industrial Measurement Environments 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (optional)

4

Navigatie en intelligente voertuigen - Prof. L. Van Biesen Navigation and Intelligent Vehicles 2nd Master of Sciences in Engineering : Electro-Mechanical Engineering (compulsory) 3rd Master Electro-Mechanical Engineer (compulsory)

3

Physical Communication - Prof. L. Van Biesen 1st Master of Sciences in Engineering: Computer Science (optional) 2nd Master of Sciences in Engineering: Applied Computer Science (optional) 2nd Master of Sciences in Engineering: Computer Science (optional)

6

Signaaltheorie - Prof. L. Van Biesen Signal Theory 1st Master of Sciences in Engineering : Electronics and IT-Engineering (compulsory) 2nd Master of Sciences in Engineering: Computer Science (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (compulsory)

5

Stage: elektrotechniek - Prof. L. Van Biesen Traineeship: Electrical Engineering 2nd Master of Sciences in Engineering : Applied Computer Sciences (optional) 2nd Master of Sciences in Engineering : Biomedical Engineering (optional) 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 3rd year Electronic and IT Engineer (optional)

6

Stem, beeld, navigatie en telemetrie - Prof. L. Van Biesen Voice, Image Coding, Media and Systems 1st Master of Sciences in Engineering: Computer Science (optional) (optional) 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Computer Science (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (optional)

6

Toegepaste elektriciteit - Prof. L. Van Biesen , Prof. Wendy Van Moer Basic Electricity 2nd Bachelor of Sciences in Engineering (compulsory) 3rd Bachelor of Sciences in Engineering (compulsory)

7

Transmissiemedia en -systemen - Prof. L. Van Biesen

101

Annual report ELEC 2012

Lectures and practical courses 3

Credits

Physical Communication 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (compulsory) 2nd Master of Sciences in Engineering : Applied Computer Sciences (optional)

4 6

CAE-tools voor het ontwerp van analoge elektronische schakelingen - Prof. G. Vandersteen CAE-tools for the Design of Analog Electronic Circuits 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (optional)

4

Design en karakteriseren van hoogfrequente (niet-lineaire) systemen - Prof. W. Van Moer Design and Characterisation of RF and Microwave Nonlinear Systems 2nd Master of Sciences in Engineering : Electronics and IT-Engineering (optional) 2nd Master of Sciences in Engineering: Photonics Engineering (optional) 3rd year Electronic and IT Engineer (optional)

4

3.4

MINOR “MEASURING, MODELLING, AND SIMULATION OF DYNAMIC SYSTEMS” OFFERED WITHIN THE 2ND MASTER ELECTRONICS AND INFORMATION TECHNOLOGY Course (Dutch or English)

Credits

An introduction to system identification - Prof. J. Schoukens

4

Measuring and modelling of nonlinear systems - Prof. J. Schoukens

4

Industrial measurement environments - Prof. L. Van Biesen

4

Identification of dynamic systems - Prof. R. Pintelon

4

Advanced control methods - Prof. R. Pintelon - Prof. J. Swevers, KULeuven

4

CAE-tools for the design of analog electronic circuits - Prof. G. Vandersteen

4

Design and characterization of high-frequency (nonlinear) systems - Prof. W. Van Moer

4

Bio-informatics and datamining - Prof. J. Schoukens - Prof. Y. Moreau (KULeuven)

4

Design of new systems and products is a complex process with a central role for the engineer. A good physical insight is combined with experimental results to come eventually to the best products (lowest price, good quality, no pollution) to concur the competition. Each step in this design process requires dedicated tools. We should collect good experimental data, often collected under poor conditions (course: Industrial measurement techniques). From these measurements a wide variety of models is extracted for the designer (course: Identification of dynamic systems) that should meet many requirements. Sometimes the models are used in simulation tools like ‘Spice’, or they are used to measure parameters that are difficult to access directly, like the damping of a wing, the time constants of an electrical machine. Models are also at the basis of control design (for example the active suspension of a car). A lot of commercial software packages are available on the market to support the design process, but using these tools without understanding can result in a bad or poor design or even create dangerous situations. In a simulation package, many user parameters are set to default values that are not always suitable for the specific situation, and this can jeopardize the results completely (course: CAE-tools for the

102

Education

design of analog electronic circuits). During the design, it is very tempting to rely on linear models because they are very intuitive, easy to use, many rules of thumb are available, and it is not too difficult to extract them from measurements. However, nature is not linear. What is the quality of the design under these conditions? Is the stability analysis of the controller still valid? How to design a controller in the presence of nonlinear distortions? How is the bit-error-rate of a high speed communication link affected by a nonlinear amplifier? The courses Measuring and modelling of nonlinear systems and Advanced control methods help to answer these questions. These ideas are practically applied in microwave designs where transistors are often pushed in their nonlinear operating region for power efficiency reasons (course: Design and characterization of highfrequency (nonlinear) systems). As an engineer, we are overwhelmed with information that is often out of the scope or our interest. The course on Bioinformatics and datamining opens a new scientific field that is directed to the problem how to extract the desired information out of enormous amounts of data. How can we turn data into information?

3.5

DESIGNING SYSTEMS FROM CONCEPTS: THE PING-PONG TOWER PROJECT

The design of complex systems demands that engineers possess significant set of abstract systemlevel thinking skills. Engineering students therefore need to be exposed to the art of solving problems systematically and have to learn the limitations and the backsides of ad-hoc methods, to ensure that they should only turn to these methods as last resort alternatives. To start the process of system-based thinking early, we use an experience based learning project during the students’ fourth semester to awaken them to a systematic engineering approach. This project is taken by all engineering students at our university. As a consequence, all the students are taught the crucial concepts that can lead to the inclusion of sustainable development in engineering practice irrespective of their final specialization in electronics, mechanics, chemistry…. A feasible toy engineering problem is proposed that includes a lot of practical engineering problems: The process to be controlled is the stabilization of the height of a Ping-Pong ball floating in a usercontrolled airflow inside a transparent Plexiglas tube. Although students get a strong guidance towards good engineering practice, they have to choose the method and decide on the practical implementation themselves. Pedagogically speaking, the major advantage of this project is that the students gain a lot of engineering attitudes. Firstly, they gain hands-on experience in a wide range of engineering applications: digital electronics, analog electronics, power electronics, control engineering, signal processing, optical system design, and computer engineering all have their role to play in the project. Secondly, they gain the insight that system-level thinking leads to complexity reduction and problem partitioning, and therefore allows to solve large-scale problems that would remain

103

Annual report ELEC 2012

untracktable otherwise. They learn that walking the lines of a systematic design framework leads to well-understood, high-quality, reproducible and reusable results. In the next section, we situate the project in the engineering study. Afterwards, we explain the different steps the student should take. Then we describe which engineering attitudes the students gain throughout this project.

3.5.1

Situating the project At the Vrije Universiteit Brussel, all engineering students follow the same courses during their first four semesters of their bachelor education. In order to help the engineering students to choose between different specializations, they are confronted with four different engineering problems in their fourth semester, one in civil engineering, one in chemical engineering, one in mechanical engineering and one in electronic engineering. The ping pong tower project has a number of inherent advantages due to its broad range of possible solutions. The solutions that can be built

are a combination of analog and digital electronics; use both hardware and software solutions; have the dynamics of the system are in a practical range, making it possible to demonstrate instabilities; involve no safety risks due to the use of low voltages; make it possible to introduce the students to control theory.

104

Education

3.5.2

The project

The students need to control the height of a Ping-Pong ball in a tube by means of a fan which blows air in the tube. They need to set and measure the height using a PC. Every group needs to present its work in a scientific way by the end of the project to train their scientific presentation skills. The goal is that the teams build up a complete solution starting from available basic building blocks: 1. A variable speed fan blows air into a Plexiglas tube whose diameter is 4 mm larger than the Ping-Pong bal. 2. An example interface between the PC and a PIC-based microcontroller board which has a USB connection to the PC. The firmware of the microcontroller board contains the implementation of an analog-to-digital convertor and a PID controller with user adjustable gains besides the present USB interface to the PC. The simple PC program allows the user to set the wanted height and read the actual height of the Ping-Pong ball from the microcontroller board. 3. Various pre-existing modules are available since – due to the limited timeframe – it is impossible that the students build everything from scratch. The pre-existing modules are an ultrasonic position sensor module, a fan power steering module, a microcontroller board and a simple PC program.

3.5.3

From concept to working system

The key idea is to illustrate the usefulness of a top-down design for the control of complex systems. The students are thought to first reason at the system level using simple system models. Next, good engineering practice rules are to be used to specify the modules separately. Then, the selected modules are designed separately from a set of discrete components. The final challenge is to combine everything in a performing system. During the first one and a half day, the students try to understand the problem by slicing it into smaller sub-problems. During the next day, they decide on their strategy and implementation. The next two days, they spend on implementing their different blocks. The final one and a half day before the presentation, they need to combine the different sub-blocks and tune their controller.

3.5.3.1

Step 1: Understanding the problem

What is a system? This question is the key problem tackled in this phase. Therefore, the students get a short introduction to the system-level concept. They discover the usefulness of a block diagram of their complete system to support high-level reasoning. They partition the problem into different logical blocks with a smaller complexity. They also have to think about the analog and/or digital interfaces between the different blocks. This brings them to a block diagram like Figure 1 the figure below. For the first time in their education, the students have to deal with

105

Annual report ELEC 2012

the sensors and sensor data, the controller, the power steering stage for the fan, interfacing the digital data between a PC and a microcontroller, and the design of a user-friendly interface on the PC.

+

Controller -

Fan power steering

Fan

Tube with ping-pong ball

Height sensor

Interface

USB

Figure 1. A possible block diagram for the problem.

In order to be able to interconnect all the blocks, they need to decide on the interfaces between the different blocks. Therefore, they write down specifications for the different blocks. The interface specifications of the pre-existing modules are given in a datasheet format. These specifications let the students reflect on the interconnection and make them acquainted with datasheets. The power steering module – for example – has two different inputs: an analog voltage and a digital PWM input that are both controlling the output voltage. They have to decide which type of signal they are going to use and consider the influence of their choice to the rest of the system. The students experience that most design problems do not have a unique solution, unlike what they found from their courses in mathematics. Although the control loop will always look the same at system level, the block diagrams of the different groups can differ. Some groups implement for example the controller in the PC program, which moves the interface between the PC and the system directly to the controller input. Such choices have a large impact. The students learn to weight off the different advantages and disadvantages, and learn to slice a complex problem into tractable and independent, co-operating blocks.

3.5.3.2

Step 2: Strategic thinking

Due to the limited timeframe and their lack of practical experience, it is not possible for every group to build up a complete solution from scratch. Therefore, depending on the number of students in the group, two to three blocks are selected for a design from scratch, depending on the

106

Education

student’s own interest. This allows training their negotiation skills and expressing their leadership. They use the pre-existing modules for the other blocks. The groups define work packages that fit the workload of two students for the next three days. Since they work in teams of two students, they all feel engaged in the project. They can make their own contribution to the project without being overshadowed by one brilliant student in the group who would otherwise be designing the complete project. Every team then focuses on the design strategy of their specific block. They have the freedom to choose their approach freely. The sensor – for example – can be realized under different forms: an ultrasonic height sensor, an optical scan line, an image-based webcam sensor, … They are also encouraged to propose their own solutions, although the feasibility needs to be checked beforehand.

3.5.3.3

Step 3: Creating the different blocks

When they develop a specific block, they learn that the system-based approach used before can be reused. The specific problem is further split into different sub-blocks until they reach the level where building and understanding each part becomes straightforward. Then, each element is realized and tested to meet the prescribed specifications. Afterwards, those elementary parts are combined into subsystems. For the ultrasonic sensor, this system-based approach results in the design of a transmitter, a receiver and a signal processing unit. The transmitter generates a pulsed 40 kHz signal together with its envelope. The receiver cleans up the measured reflected burst and generates a discrete envelope of the received signal. The signal processing unit measures the time T between the transmitted and reflected burst, which is proportional to the distance L between the sensor and the Ping-Pong ball. The system-based approach can be recycled ones more. For example, the pulse transmitter in the ultrasonic sensor can be split into different blocks: the 40 kHz oscillator, the envelope generator, the masking of the oscillator by the envelope and the driver for the loudspeaker. These different blocks are basic electronic building blocks that the students already studied or can be found easily in references on electronic circuits or on the internet. When the different basic building blocks are combined, the students experience the concepts of loading and interference. The loading of a circuit requires the design of an output driver such that it matches the input characteristics of the next block. The interference is minimized by a careful routing of the power line, proper decoupling and filtering of the appropriate signals. During this step, they learn to iteratively use the system-level approach to end up with a hierarchical design and how elementary building blocks in electronics or informatics can be combined into a sub-system that meets the specifications. They also experience that the basic building blocks seen in introductory courses are used in real systems.

107

Annual report ELEC 2012

Pulsetransmitter

40 kHz

Ultrasonic sensor

Processing Receiver Reflection time T

Distance L

Out Figure 2. In the case of the ultrasonic height sensor, the block can be further split into three sub-blocks: the transmitter, the receiver and the signal processing.

3.5.3.4

Step 4: Going back to the system level

Now that the separate blocks are operational, the challenge lies in their combination, i.e. by suppressing the interference and overcoming the loading of the different blocks. Finally, system-level testing is performed. Once the system is reaching specs, it becomes time to close the control loop. Here, students encounter feedback loops for the first time. This shows them the power of feedback control and nicely illustrates the properties of the different control actions: The proportional gain which lacks accuracy, the derivative action for speed improvement and the integral action for error removal and the consequent instability. First – by playing manually with the gains of the different actions – the students discover the advantages and disadvantages of the different actions. Second – by the end of the day – the students are explained how the relay test works. This allows them to obtain a working controller without any heuristic search. Hence, the way is now opened for them to understand the usefulness and the power of the otherwise so abstract control theory. They learn to evaluate the control behaviour by looking at the tracking behaviour and the disturbance rejection. During this step, they learn how to combine different blocks to a larger system and also to evaluate and tune the system’s performance by measurements. To finalize their effort and tune their controller, they need one and a half day.

108

Education

3.5.3.5

Step 5: Presentation

During the last half day, the groups have to demonstrate their system to the groups and give a scientific presentation of their solution. This presentation trains their communication skill. It also enables them to learn about other implementations from other groups.

3.5.3.6

Conclusion

The project is mainly intended to help the students to make an informative choice about their engineering specialization, and also teaches them a fundamental engineering concept: design a complex control system. The system-level systematic framework enables the engineer to develop well-understood, highquality, reproducible and reusable results.

3.6

COURSES LECTURED IN THE FACULTY OF SCIENCE AND BIOENGINEERING Lectures and practical courses

Credits

Berekenbaarheid en informatietheorie 4 - Prof. L. Van Biesen Computability and Information Theory 1st Year Master of Sciences in Engineering: Computer Science (compulsory)

6

Geographical Information Systems - Prof. L. Van Biesen 2nd year Master of Ecological Marine Management (compulsory) 2nd Year Master of Sciences in Engineering: Applied Computer Science (optional) 2nd Year Master of Science Ecological Marine Management (compulsory)

3

Theory of Computation and Information Theory - Prof. L. Van Biesen 1st Year Master of Sciences in Engineering: Computer Science (compulsory)

6

Analyse, WPO 5: ir. Griet Monteyne,Dr. Anna Marconato Analysis, Excercises 1st Year Bachelor of Science in Math

14

6

Complexe Analyse, WPO : Dr. Lieve Lauwers, Jan Goos Complex Analysis, Excercises, 2nd Year Bachelor of Sciences in Engineering and physics

5

Wiskundige statistiek, WPO7, Prof. Kurt Barbé Mathematical statistics 3rd Year Bachelor of Science in Math

6

3.7

DOCTORAL TRAINING PROGRAMME Lectures and practical courses

Statistics for PhD students: Prof. K. Barbé Central PhD training programme (R&D)

4 5 6 7

The The The The

language language language language

of of of of

tuition tuition tuition tuition

of of of of

this this this this

Credits

6

course course course course

is is is is

Dutch Dutch Dutch Dutch

109

Annual report ELEC 2012

3.8

NATIONAL AND INTERNATIONAL COURSES

3.8.1

National courses (since 2003):

3.8.1.1

Identificatie van systemen (Identification of Systems)

Organised by: University of Gent, “Instituut voor permanente vorming” Location:

IVPV - UGent, Technologiepark, 9052 Gent-Zwijnaarde

Lectured by:

Johan Schoukens and Yves Rolain

Dates:

7, 14 and 21 December 2004

Meten en modelleren is een basisactiviteit van vele ingenieurs: modellen worden gebruikt tijdens het ontwerp, in simulatoren en in eindproducten. Het modelleringsproces is een complexe activiteit die in 4 grote delen kan worden opgesplitst: verzamelen van de experimentele data; opstellen van een model; in overeenstemming brengen van een model en data; validatie van de resultaten. Systeemidentificatie biedt een systematische, optimale oplossing en wordt in deze module bestudeerd, met als toepassing the identificeren van dynamische systemen. Hierbij wordt de klemtoon gelegd op het aanbrengen van de ideeën, ondersteund door uitgewerkte Matlab illustraties. De behandelde topics zijn Systeemidentificatie: wat? waarom? Een verhelderend voorbeeld Goede schatters/slechte schatters, wat mag je ervan verwachten? Niet-parametrische identificatie van frequentieresponse functies Basisidee: van tijdsignaal tot frequentierespons (FRF) Experiment design: keuze van de excitatiesignalen, ruisgevoeligheid, uitmiddelen Nietlineaire distorties: detectie, kwalificatie en quantificatie Parametrische identificatie van de transferfunctie Basisidee: van data tot model Tijdsdomein- en frequentiedomein-identicatie Identificatie van tijdsvariërende systemen Basisidee Balans volgsnelheid/ruisgevoeligheid

3.8.1.2

Courses lectured at the Katholieke Universiteit Leuven (KUL) Rik Pintelon: “Identificeren van lineaire dynamische systemen” 18 HOC, 36 WPO (4 credits): keuze o.o. in de Master in de wiskundige ingenieurstechnieken

110

Education

Johan Schoukens: “Systeemidentificatie en modellering” (6 credits): Master in de ingenieurswetenschappen: wiskundige ingenieurstechnieken Master in de ingenieurswetenschappen: bouwkunde Master in de wiskunde Master in de ingenieurswetenschappen: wiskundige ingenieurstechnieken, programma voor industrieel ingenieurs of master industriële wetenschappen (aanverwante richting) (na toelating) Yves Rolain: “Meten en modelleren” keuze o.o. in de Master in de wiskundige ingenieurstechnieken

3.8.1.3

Open course program-IMEC academy: DSP concept explained with well-chosen exercises.

Organised by: imec-leuven Location:

imec, Leuven, Belgium

Lectured by:

Johan Schoukens and Yves Rolain

Dates:

November 14th, 28th, December 5th, 19th, 22nd (2011)

The course is a basic theoretical introduction to the concepts of digital signal processing. Introduction to system theory and signal processing impulse response and transfer functions of linear systems; stability-causality; poles and zeros; sampling DFF=FFT – all with Matlab exercises. Introduction to measurement and modelling of linear systems (measurement of the frequency response function; choice of excitation; the effect of noise and leakage; estimation of the parametric model) – including Matlab exercises. Handling non-linear distortion: detection; classification and qualification of linear distortion – including Matlab exercises. Design of digital filters and systems (basic choices and non idealities: filter examples; compression and expansion …) – including Matlab exercises. Wrap-up: further Matlab exercises applying the techniques on integrated problems. Analysis of implemented filters; evaluating non-linear distortions (impact of quantizing noise).

111

Annual report ELEC 2012

3.8.2

International courses (since 2003):

3.8.2.1

Characterisation of Multiport Systems through 3-port LSNA Measurements

Location:

Seminar at NIST, Boulder, CO, December 2003

Lectured by:

Wendy Van Moer

# attendees:

20

3.8.2.2

The use of multisines

Location:

Seminar at NIST, Boulder, CO, December 2003

Lectured by:

Daan Rabijns

# attendees:

20

3.8.2.3

GIS training in SEAFDEC, Thailand

Location:

Samut Prakan SEAFDEC/Training Department, Thailand, Bangkok, 2005

Lectured by:

Tesfazghi Ghebre Egziabeher

The theme of the course was interrelated to the use of Geographic Information System for Fishery Management. Participants of the course were members of the Southeast Asian Fisheries Development and Training Centre (SEAFDC/TD), who were professionally engaged in the fishing industry.

3.8.2.4

Measuring, Modeling, and Designing in a Nonlinear Environment

Location:

tutorial workshop organized at I2MTC08, Vancouver, Canada, May 2008

Lectured by:

Yves Rolain, Ludwig De Locht, Rik Pintelon, Johan Schoukens, Wendy Van Moer

Topics: Best linear approximation and design: A perfect marriage (L. De Locht) Measurement of the Best Linear Approximation of Nonlinear Systems (W. Van Moer) Impact of nonlinear distortions on the linear framework (J. Schoukens) Frequency Response Function Measurement in the Presence of Nonlinear Distortions (R. Pintelon)

3.8.2.5

VUB - doctoral school on Identification of Nonlinear Dynamic Systems

Location:

Dept. ELEC, Vrije Universiteit Brussel, Building K, 6th floor

Lectured by:

Rik Pintelon, Johan Schoukens, Gerd Vandersteen, Yves Rolain

# attendees:

18 (8 different nationalities) in 2008, 14 (9 different nationalities) in 2010 and 18

(9 different nationalities) in 2011

112

Education

The department ELEC of the VUB, organized a 4 weeks doctoral school (from the 26th of May till the 20th of June in 2008, from 16th May till the 21st of June in 2009; from the 6th of June till the 2nd of July in 2010 and from 7th of May until 5th of June in 2011) to give an intensive training on advanced modelling and simulation techniques of (non)linear dynamic systems, starting from experimental data. Half of the time has been spent on courses/exercises, the other half on a project to get hands-on experience. The material teached during the courses and exercises has been put into practice during a clearly defined project, in order to get hands-on experience. The course covered the following topics: A basic introduction to system identification, Identification of dynamic systems, Measuring and modelling of nonlinear systems, Simulation tools for nonlinear systems, Design and characterization of high-frequency (nonlinear) systems (optional: intended for those with an interest in microwave systems). The next doctoral school on Identification of Nonlinear Dynamic Systems 8 will be organised in May/June 2012 (from the 21st of May 2012 till the 15th of June 2012). Participation to this workshop offers a number of advantages. Besides the training, it can also be the start of a collaboration. To some of the participants we can offer a one year grants to start a research collaboration, or even a full four years grant for a (joint) PhD. Interested candidates are invited to send their curriculum vitae, together with a short motivation why they would like to follow this course, and this before the 20th of February 2012. They can also express their interest in the possibility for a longer cooperation. Please do not hesitate to contact us

if

you

would

like

to

have

more

information:

[email protected].

8

http://wwwtw.vub.ac.be/elec/doctoralschool.htm

113

email:

[email protected]

or

Annual report ELEC 2012

4. Bibliography9 4.1 b1.

BOOKS Identification of Linear Systems: A Practical Guideline to Accurate Modeling J. Schoukens, R. Pintelon Pergamon Press - London, 1991 (ISBN 0-08-040734-X) The book is concentrated on the problem of accurate modelling of linear time invariant systems. These models can be continuous time (Laplace-domain) or discrete time (Z-domain). The complete experimental procedure is discussed: how to create optimal experiments (optimization of excitation signals), how to estimate the model parameters from the measurements, how to select between different models, etc. These problems are thoroughly discussed in the first section of the book. A profound theoretical development of the proposed identification algorithm is also made in this section. The second part consists of detailed illustrations of the proposed algorithms on practical problems: modelling an electronic, electrical, acoustic, mechanical system. Finally the book is completed with a practical guideline to help the user making the correct choices. The book is intended for all those dealing with “practical” modelling problems and have to combine measurements and theory. A second group of interested people are those involved with identification theory.

b2.

System Identification: A Frequency Domain Approach Rik Pintelon, Johan Schoukens IEEE Press, Piscataway, NJ, 2001 (ISBN 0-7803-6000-1) How does one model a linear dynamic system from noisy data This book presents a general approach to this problem, with both practical examples and theoretical discussions that give the reader a sound understanding of the subject and of the pitfalls that might occur on the road from raw data to validated model. The emphasis is on robust methods that can be used with a minimum of user interaction. System Identification: A Frequency Domain Approach is written for practising engineers and scientists who do not want to delve into mathematical details of proofs. Also, it is written for researchers who wish to learn more about the theoretical aspects of the proofs. Several of the introductory chapters are suitable for undergraduates. Each chapter begins with an abstract and ends with exercises, and examples are given throughout

b3.

System Identification A frequency Domain Approach - second edition Rik Pintelon, Johan Schoukens IEEE Press, Wiley 2012 (ISBN 978-0-470-64037-1) Systems identification is a general term used to describe mathematical tools and algorithms that build dynamical models from measured data. Used for prediction, control, physical interpretation, and the designing of any electrical systems, they are vital in the fields of electrical, mechanical, civil, and chemical engineering. Focusing mainly on frequency domain techniques, System Identification: A Frequency Domain Approach, Second Edition also studies in detail the similarities and differences with the classical time domain approach. It high lights many of the important steps in the identification process, points out the possible pitfalls to the reader, and illustrates the powerful tools that are available. Readers of this Second Editon will benefit from: - Matlab ® software support for identifying multivariable systems that is freely available at the website http://booksupport.wiley.com - State-of-the-art system identification methods for both time and frequency domain data - New chapters on non-parametric and parametric transfer function modeling using (non-)period excitations - Numerous examples and figures that facilitate the learning process - A simple writing style that allows the reader to learn more about the theoretical aspects of the proofs and algorithms Unlike other books in this field, System Identification: A Frequency Domain Approach, Second Edition is ideal for practicing engineers, scientists, researchers, and both master's and PhD students in electrical, mechanical, civil, and chemical engineering.

9

More publications of the department ELEC can be http://www.vub.ac.be/infovoor/onderzoekers/research/team_pub.php?team_code=elec or http://wwwtw.vub.ac.be/elec/Papers%20on%20web/index.html

114

found

on:

Bibliography

b4.

Low Rank Approximation: Algorithms, Implementation, Applications. Ivan Markovsky Series “Communications and Control Engineering”, Springer (ISBN 978-1-4471-2227-2) Matrix low-rank approximation is intimately related to data modelling; a problem that arises frequently in many different fields. Low Rank Approximation: Algorithms, Implementation, Applications is a comprehensive exposition of the theory, algorithms, and applications of structured low-rank approximation. Local optimization methods and effective suboptimal convex relaxations for Toeplitz, Hankel, and Sylvester structured problems are presented. A major part of the text is devoted to application of the theory. Applications described include: - system and control theory: approximate realization, model reduction, output error, and errors-in-variables identification; - signal processing: harmonic retrieval, sum-of-damped exponentials, finite impulse response modeling, and array processing; - machine learning: multidimensional scaling and recommender system; - computer vision: algebraic curve fitting and fundamental matrix estimation; - bioinformatics for microarray data analysis; - chemometrics for multivariate calibration; - psychometrics for factor analysis; and - computer algebra for approximate common divisor computation.

b5.

Mastering System Identification in 100 Exercises Johan Schoukens, Rik Pintelon, Yves Rolain IEEE Press, Wiley 2012 (ISBN 978-0-470-9368-98-6) Systems identification is a general term used to describe mathematical tools and algorithms that build dynamical models from measured data. Mastering System Identification in 100 Exercises takes readers step by step through a series of MATLAB ® exercises that teach how to measure and model linear dynamic systems in the presence of nonlinear distortions from a practical point of view. Each exercise is followed by a short discussion illustrating what lessons can be learned by the reader. The book, with its learn-by-doing approach, also includes: - State-of-the-art system identification methods, with both time and frequency domain system identification methods - including the pros and cons of each - Simple writing style with numerous examples and figures - Downloadable author-programmed MATLAB ® files for each exercise--with detailed solutions - Larger projects that serve as potential assignments Covering both classic and recent measurement and identifying methods, this book will appeal to practicing engineers, scientists, and researchers, as well as master's and PhD students in electrical, mechanical, civil, and chemical engineering.

4.2 p411.

JOURNAL PAPERS (2012) Initializing Wiener-Hammerstein models based on partitioning of the best linear approximation J. Sjöberg, J. Schoukens Automatica Vol. 48, No. 2 (2012) 353-359 This paper describes a new algorithm for initializing and estimating Wiener–Hammerstein models which consist of two linear parts with a static nonlinearity in between. The algorithm makes use of the best linear model of the system, which is a consistent estimate of the systems dynamics for Gaussian excitations. The linear model is split in all possible ways into two sub-models. For all possible splits, a Wiener–Hammerstein model is initialized which means that a nonlinearity is introduced in between the two sub-models. The linear parameters of this nonlinearity can be estimated using least-squares. All initialized models can then be ranked depending on the fit. Typically, one is only interested in the best one, for which all parameters are fitted using prediction error minimization. The paper explains the algorithm in detail and consistency of the initialization is proven. Computational aspects are investigated, showing that in most realistic cases, the number of splits of the initial linear model remains low enough to make the algorithm useful. The algorithm is illustrated on an example where it is shown that the initialization is a tool to avoid many local minima.

p412.

Analyzing the Windkessel Model as a Potential Oscillometric Blood-Pressure Measurements

Candidate for

Correcting

Kurt Barbé, Wendy Van Moer and Danny Schoors IEEE Transactions On Instrumentation And Measurement, No. 2, Vol. 61, pp. 411 – 418 Developing a good model for oscillometric blood-pressure measurements is a hard task. This is mainly due to the fact that the systolic and diastolic pressures cannot be directly measured by noninvasive automatic oscillometric bloodpressure meters (NIBP) but need to be computed based on some kind of algorithm. This is in strong contrast with the classical Korotkoff method, where the diastolic and systolic blood pressures can be directly measured by a sphygmomanometer. Although an NIBP returns results similar to the Korotkoff method for patients with normal blood pressures, a big discrepancy exist between both methods for severe hyper-and hypotension. For these severe cases, a statistical model is needed to compensate or calibrate the oscillometric blood-pressure meters. Although different statistical models have been already studied, no immediate calibration method has been proposed. The reason is that

115

Annual report ELEC 2012

the step from a model, describing the measurements, to a calibration, correcting the blood-pressure meters, is a rather large leap. In this paper, we study a " databased" Fourier series approach to model the oscillometric waveform and use the Windkessel model for the blood flow to correct the oscillometric blood-pressure meters. The method is validated on a measurement campaign consisting of healthy patients and patients suffering from either hyper-or hypotension. p413.

Frequency domain based nonlinear feed forward control design for friction compensation David Rijlaarsdam, Pieter Nuij, Johan Schoukens, Maarten Steinbuch Mechanical Systems And Signal Processing, Vol. 27, February 2012, pp. 551-562 A frequency domain based method for controller design for nonlinear systems is presented. In a case study, this method is applied to optimally design a feed forward friction compensator for an industrial motion stage in a transmission electron microscope. It is shown that the frequency domain approach yields a tool to fast and easily design friction control in practice with high detection sensitivity and orthogonal tuning of the controller parameters, while providing a well defined notion of optimal performance.

p414.

Analysis of Best Linear Approximation of a Wiener–Hammerstein System for Arbitrary Amplitude Distributions Hin Kwan Wong; Schoukens, J.; Godfrey, K.R. IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 3, March 2012, pp. 645 - 654 For a nonlinear system with an input signal having a Gaussian probability distribution (this includes random-phase multisines), the best linear approximation (BLA) is proportional to the frequency response of the overall system. However, this is not the case for non-Gaussian input signals, for which the frequency response is biased with respect to the Gaussian BLA. In this paper, theoretical expressions for determining this bias for Wiener-Hammerstein systems are developed both for binary input signals and for white noise inputs with arbitrary probability distribution. Cubic and quintic nonlinearities are considered, but the methods can be extended to other forms of polynomial nonlinearity. Simple measures for quantifying the bias are also developed. It is shown that the bias decays rapidly to zero for a growing length of the impulse response.

p415.

Demonstration of sawtooth period locking with power modulation in TCV plasmas M. Lauret, F. Felici, G. Witvoet, T.P. Goodman, G. Vandersteen, O. Sauter, M.R. de Baar and the TCV team IOP Publishing and International Atomic Energy agency - Nucl. Fusion 52 (2012) 062002 (4pp), 2012 Corroborating evidence is presented that the sawtooth period can follow the modulation frequency of an externally applied high power electron cyclotron wave source. Precise, fast and robust open loop control of the sawtooth period with a continuously changing reference period has been achieved. This period locking is not associated with the crash, but with the phase evolution of the inter-crash dynamics. This opens new possibilities of open loop control for physics studies and maybe for reactor performance control.

p416.

Cross-term Elimination in Parallel Wiener Systems Transformation

Using

a Linear

Input

Maarten Schoukens, Yves Rolain IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 3, March 2012, pp. 845 - 847 Multivariate polynomials are often used to model nonlinear behavior, e.g., in parallel Wiener models. These multivariate polynomials are mostly hard to interpret due to the presence of cross terms. These polynomials also have a high amount of coefficients, and the calculation of an inverse of a multivariate polynomial with cross terms is cumbersome. This paper proposes a method to eliminate the cross terms of a multivariate polynomial using a linear input transformation. It is shown how every homogeneous polynomial described using tensors can be transformed to a canonical form using multilinear algebraic decomposition methods. Such tensor decomposition methods have already been used in nonlinear system modeling to reduce the complexity of Volterra models. Since every polynomial can be written as a sum of homogeneous polynomials, this method results in a decoupled description of any multivariate polynomial, allowing a model description that is easier to interpret, easier to use in a design, and easier to invert. This paper first describes a method to represent and decouple multivariate polynomials using tensor representation and tensor decomposition techniques. This method is applied to a parallel Wiener model structure, where a multiple-input-single-output polynomial is used to describe the static nonlinearity of the system. A numerical example shows the performance of the proposed method. p417.

Time Series Reconstruction from Unequally Spaced natural Archive Data Veerle Beelaerts, Maite Bauwens, Rik Pintelon Mathematical Geosciences, Vol. 44. Some natural substrates record environmental information and, as such, provide a means to reconstruct the environmental conditions from the period these substrates were formed. Samples from environmental archives are not always equally spaced in distance. When a periodic time series model is estimated from these unequally spaced proxy records, the search for reasonable starting values is the maindifficulty. In this work, a non-parametric method

116

Bibliography

based on the regressive Fourier series is first presented, which reduces averaging errors starting from unequally spaced records. The method is applied to synthetic data and generally performs well in all circumstances. Secondly, a parametric method for the construction of a time base and the elimination of averaging errors from unequally spaced records is presented. This parametric method uses the non-parametric method to produce starting values for the parameters. The method is compared with the time series construction method with the averaging effect taken into account and it is observed that only the current method produces acceptable results. The statistical performance of the method is verified with a Monte Carlo simulation and the estimator is proven to be an efficient estimator. The applicability of the method is demonstrated on the vessel density measurement in a mangrove tree, Rizophora mucronata, which is a proxy for the rainfall in tropical coastal regions. p418.

Recursive identification of Hammerstein systems with application to electrically stimulated muscle. F. Le, I. Markovsky, C. Freeman, and E. Rogers Control Engineering Practice, 20(4): pp. 386–396, 2012 Abstract: Modeling of electrically stimulated muscle is considered in this paper where a Hammerstein structure is selected to represent the isometric response. Motivated by the slowly time-varying properties of the muscle system, recursive identification of Hammerstein structures is investigated. A recursive algorithm is then developed to address limitations in the approaches currently available. The linear and nonlinear parameters are separated and estimated recursively in a parallel manner, with each updating algorithm using the most up-to-date estimation produced by the other algorithm at each time instant. Hence the procedure is termed the alternately recursive least square (ARLS) algorithm. When compared with the leading approach in this application area, ARLS exhibits superior performance in both numerical simulations and experimental tests with electrically stimulated muscle.

p419.

Non-parametric estimate of the system function of a time-varying system John Lataire, Rik Pintelon, Ebrahim Louarroudi Automatica, Vol. 48, No. 4, April 2012, Pages 666–672 The task of identifying an unknown dynamic system is made easier with prior knowledge on its behaviour. Using a frequency domain approach, the non-parametric maximum likelihood estimator of the system function, associated with the time-dependent impulse response of a time-varying system, is constructed. This is accomplished by the use of a simple linear least squares fitting algorithm, applied to the spectral response of the system to a multisine excitation. The noise variance on the system function is estimated simultaneously, and modelling errors can be detected, as illustrated on a simulation example.

p420.

Exploiting the Phantom-Mode Signal in DSL Applications Wim Foubert, Carine Neus, Leo Van Biesen and Yves Rolain IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 4, April 2012, pp. 896-902 In order to meet the ever-increasing bandwidth demand of the users, new “digital subscriber line” (DSL) technologies are being developed. In addition to the traditional differential mode of the telephone line, telecom operators are now looking in to the exploitation of the phantom-mode signal. New transmission line models, using the multiconductor transmission lines theory, were developed to support this extended use. In this paper, it is shown that the phantom mode is an eigenmode of the quad cable system, and hence, according to the theory, there is no crosstalk between this mode and the differential mode.

p421.

Nonparametric Tracking of the Time-Varying Dynamics of Weakly Nonlinear Periodically Time-Varying Systems Using Periodic Inputs Ebrahim Louarroudi, Rik Pintelon, John Lataire IEEE Transactions On Instrumentation And Measurement, Vol. 61, No. 5, May 2012, pp. 1384-1394 In this paper, a nonparametric estimation procedure is presented in order to track the evolution of the dynamics of continuous (discrete)-time (non)-linear periodically time-varying (PTV) systems. Multisine excitations are applied to a PTV system since this kind of excitation signals allows us to discriminate between the noise and the nonlinear distortion from a single experiment. The key idea is that a linear PTV system can be decomposed into an (in) finite series of transfer functions, the so-called harmonic transfer functions (HTFs). Moreover, a systematic methodology to determine the number of significant branches is provided in this paper as well. Making use of the local polynomial approximation, a method that was recently developed for multivariable (non)-linear time invariant systems, the HTFs, together with their uncertainties embedded in an output-error framework, are then obtained from only one single experiment. From these nonparametric estimates, the evolution of the dynamics, described by the instantaneous transfer function (ITF), can then be achieved in a simple way. The effectiveness of the identification scheme will be first illustrated through simulations before a real system will be identified. Eventually, the methodology is applied to a weakly nonlinear PTV electronic circuit.

p422.

Single and Piecewise Polynomials for Modeling of Pitched Sounds Miro Zivanovic and Johan Schoukens IEEE Transactions on Audio, Speech, and Language Processing, Vol. 20, No. 4, May 2012, pp. 1270-1281 We present a compact approach to simultaneous modeling of non-stationary harmonic and transient components in pitched sound sources. The harmonic and transient components are described by separate models which are built

117

Annual report ELEC 2012

from a common sinusoidal basis modified by a joint action of single and linear piecewise time polynomials respectively. A single polynomial accounts for slow and continuous signal time variations, while various piecewise polynomials can capture fast signal changes on smaller subintervals within the analysis window. The resulting model is linear-inparameters and the solution to the corresponding linear system of equations provides correct model parameter estimates according to the signal content in the analysis window. The model is extended to deal with mixtures of sounds, where harmonics clustered in a small bandwidth are jointly modeled as a single harmonic. The comparative results suggest that the proposed model outperforms two reference modeling methods in terms of modeling errors and number of parameters. p423.

Approximation of Confidence Sets for Output Error Systems Using Interval Analysis Sandor Kolumban, Istvan Vajk, Johan Schoukens Control Engineering And Applied Informatics, Vol. 14, No. 2, June 2012, pp. 73-79 Standard identification techniques usually result in a single point estimate of the system parameters. This is justified in cases when the number of observations is large compared to the number of system parameters. However in case of small sample count it is more reasonable to identify a set of possible parameters which contain the nominal parameters with a given probability. These confidence sets cannot be calculated directly. The paper proposes interval analytic techniques to approximate confidence sets of model parameters up to arbitrary precision. The origins of interval analysis lie in the field of reliable computing, giving certified results for every computation. It has been used to solve global optimization problems numerically providing theoretical certificates on the optimality of the results. This method of global optimization is modified in a suitable way to generate the needed confidence sets. Introduction to interval analytic techniques is given and the methodology of global optimization via these is also presented. The modifications of this algorithm needed to construct the confidence sets are discussed and the method is illustrated on a simple example. The presented algorithm is focused on the output error model structure but the methodology can be extended to more general cases as well.

p424.

On the calculation of the D-optimal multisine excitation power spectrum for broadband impedance spectroscopy measurements B Sanchez, C R Rojas, G Vandersteen, R Bragos and J Schoukens Measurement Science and Technology, Vol. 23, No. 8, (2012) 085702 (15pp) The successful application of impedance spectroscopy in daily practice requires accurate measurements for modeling complex physiological or electrochemical phenomena in a single frequency or several frequencies at different (or simultaneous) time instants. Nowadays, two approaches are possible for frequency domain impedance spectroscopy measurements: (1) using the classical technique of frequency sweep and (2) using (non-)periodic broadband signals, i.e. multisine excitations. Both techniques share the common problem of how to design the experimental conditions, e.g. the excitation power spectrum, in order to achieve accuracy of maximum impedance model parameters from the impedance data modeling process. The original contribution of this paper is the calculation and design of the Doptimal multisine excitation power spectrum for measuring impedance systems modeled as 2R-1C equivalent electrical circuits. The extension of the results presented for more complex impedance models is also discussed. The influence of the multisine power spectrum on the accuracy of the impedance model parameters is analyzed based on the Fisher information matrix. Furthermore, the optimal measuring frequency range is given based on the properties of the covariance matrix. Finally, simulations and experimental results are provided to validate the theoretical aspects presented

p425.

Fractional-Order Time Series Models for Extracting the Haemodynamic Response From Functional Magnetic Resonance Imaging Data Kurt Barbé, Wendy Van Moer and Guy Nagels IEEE Transactions on Biomedical Engineering, Vol. 39, No. 8, Augustus 2012, pp. 22642272 The postprocessing of functional magnetic resonance imaging (fMRI) data to study the brain functions deals mainly with two objectives: signal detection and extraction of the haemodynamic response. Signal detection consists of exploring and detecting those areas of the brain that are triggered due to an external stimulus. Extraction of the haemodynamic response deals with describing and measuring the physiological process of activated regions in the brain due to stimulus. The haemodynamic response represents the change in oxygen levels since the brain functions require more glucose and oxygen upon stimulus that implies a change in blood flow. In the literature, different approaches to estimate and model the haemodynamic response have been proposed. These approaches can be discriminated in model structures that either provide a proper representation of the obtained measurements but provide no or a limited amount of physiological information, or provide physiological insight but lacks a proper fit to the data. In this paper, a novel model structure is studied for describing the haemodynamics in fMRI measurements: fractional models. We show that these models are flexible enough to describe the gathered data with the additional merit of providing physiological information.

p426.

A simple non-parametric pre-processing technique to correct for non-stationary effects in measured data Kurt Barbé, Wendy Van Moer, Lieve Lauwers and Niclas Björsell IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 8, pp. 2085-2095, 2012 The general approach for modeling systems assumes that the measured signals are (weakly) stationary, i.e., the power spectrum is time invariant. However, the stationarity assumption is violated when: 1) transient effects due to

118

Bibliography

experimental conditions are dominant; 2) data are missing due to, for instance, sensor failure; or 3) the amplitude of the excitation signals smoothly varies over time due to, for instance, actuator problems. Although different methods exist to deal with each of these nonstationary effects specifically, no unified approach is available. In this paper, a new and general technique is presented to handle nonstationary effects, based on processing overlapping subrecords of the measured data. The proposed method is a simple preprocessing step where the user does not need to specify which nonstationary effect is present, nor the time interval where the nonstationary effect appears. The merits of the proposed approach are demonstrated on an operational wireless system suffering from interrupted link effects. p427.

Uniquely connecting frequency domain representations of given order polynomial Wiener-Hammerstein systems David Rijlaarsdam, Tom Oomen, Pieter Nuij, Johan Schoukens, Maarten Steinbuch Automatica, Vol. 48, No. 9, September 2012, pp. 2381-2384 The notion of frequency response functions has been generalized to nonlinear systems in several ways. However, a relation between different approaches has not yet been established. In this paper, frequency domain representations for nonlinear systems are uniquely connected for a class of nonlinear systems. Specifically, by means of novel analytical results, the generalized frequency response function (GFRF) and the higher order sinusoidal input describing function (HOSIDF) for polynomial Wiener–Hammerstein systems are explicitly related, assuming the linear dynamics are known. Necessary and sufficient conditions for this relation to exist and results on the uniqueness and equivalence of the HOSIDF and GFRF are provided. Finally, this yields an efficient computational procedure for computing the GFRF from the HOSIDF and vice versa.

p428.

Basics of broadband impedance spectroscopy measurements using periodic excitations B Sanchez, R.C. Rojas, G Vandersteen, R Bragos and J Schoukens Measurement Science and Technology, Vol. 23, No. 10, 105501 (14pp), 2012 Measuring the impedance frequency response of systems by means of frequency sweep electrical impedance spectroscopy (EIS) takes time. An alternative based on broadband signals enables the user to acquire simultaneous impedance response data collection. This is directly reflected in a short measuring time compared to the frequency sweep approach. As a result of this increase in the measuring speed, the accuracy of the impedance spectrum is compromised. The aim of this paper is to study how the choice of the broadband signal can contribute to mitigate this accuracy loss. A review of the major advantages and pitfalls of four different periodic broadband excitations suitable to be used in EIS applications is presented. Their influence on the instrumentation and impedance spectrum accuracy is analyzed. Additionally, the signal processing tools to objectively evaluate the quality of the impedance spectrum are described. In view of the experimental results reported, the impedance spectrum signal-to-noise ratio (SNRZ) obtained with multisine or discrete interval binary sequence signals is about 20–30 dB more accurate than maximum length binary sequence or chirp signals.

p429.

Parametric Identification of Parallel Wiener Systems Maarten Schoukens and Yves Rolain IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 10, October 2012, pp. 2825 - 2832 This paper proposes a parametric identification method for parallel Wiener systems. The linear dynamic parts of the Wiener system are modeled by a parametric rational function in the Laplace or z-domain. The static nonlinearity is represented by a linear combination of multiple-input single-output nonlinear basis functions. The identification method uses a three-step procedure to obtain initial estimates. In the first step, the frequency response function of the best linear approximation is estimated for different input excitation levels. In the second step, the powerdependent dynamics are decomposed over a number of parallel orthogonal branches. In the last step, the static nonlinearities are estimated using a linear least squares estimation. Furthermore, a nonlinear optimization method is implemented to refine the estimates. The method is illustrated on a simulation and a validation measurement example.

p430.

Estimation of the FRF Through the Improved Local Bandwidth Selection in the Local Polynomial Method. P. Thummala and J. Schoukens. IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 10, October 2012, pp. 2833–2843. This paper presents a nonparametric method to measure an improved frequency response function (FRF) of a linear dynamic system excited by a random input. Recently, the local polynomial method (LPM) has been proposed as a technique to reduce the leakage errors on FRF measurements. The noise sensitivity of the LPM was similar to that of the classical windowing methods, while the leakage rejection is improved from an O(N-1) to an O(N-3). This paper shows a methodology, to automatically tune the parameter of the LPM, viz., the local bandwidth that sets how many neighboring frequency lines are combined in a single-point estimate, to get lower noise sensitivity by increasing the smoothening of the original LPM, without any user interaction. The balance between noise reduction and bias error will be automatically retrieved.

p431.

Frequency Response Function Measurements Using Concatenated Subrecords With Arbitrary Length Schoukens, J.; Vandersteen, G.; Rolain, Y.; Pintelon, R.

119

Annual report ELEC 2012

IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 10, pp. 2682 - 2688 This paper presents a nonparametric method to measure the frequency response function of a linear dynamic system that allows a series of subrecords with arbitrary length to be concatenated in one long data record without suffering from the leakage and transient errors. The method combines data blocks (of different lengths) into a single record, resulting in an increased frequency resolution. The analysis is based on the recent insight that leakage errors in the frequency domain have a smooth nature that is completely similar to that of initial transients in the time domain. The method is applicable to both single-input-single-output and multiple-input-multiple-output systems. The results are illustrated on simulations and experimental data. p432.

Initialization of nonlinear state-space models applied to the Wiener–Hammerstein benchmark Anna Marconato, Jonas Sjöberg, Johan Schoukens Control Engineering Practice, Volume 20, Issue 11, November 2012, Pages 1126 - 1132 In this work a new initialization scheme for nonlinear state-space models is applied to the problem of identifying a Wiener–Hammerstein system on the basis of a set of real data. The proposed approach combines ideas from the statistical learning community with classic system identification methods. The results on the benchmark data are discussed and compared to the ones obtained by other related methods.

p433.

Initial estimates of the linear subsystems of Wiener–Hammerstein models David Westwick, Johan Schoukens Automatica, Vol. 48, No. 11, November 2012, Pages 2931-2936 The iterative optimizations often used to identify Wiener–Hammerstein models, pairs of linear filters separated by memoryless nonlinearities, require good initial estimates of the linear elements in order to avoid them getting caught in local minima. Previous work has shown that initial estimates of the two linear elements can be formed by splitting the poles and zeros of the best linear approximation of the Wiener–Hammerstein system between the two linear elements, an approach which can generate a large number of initializations. This paper develops a scanning technique that can efficiently evaluate each of the proposed initializations using estimates of some carefully constructed nonlinear characteristics of the system, estimates which can be formed using linear system identification techniques after some data pre-processing. This approach results in a much smaller number, often only one, of potential starting points for the optimization. The proposed algorithm is demonstrated using a Monte Carlo simulation using data from the SYSID 2009 Wiener–Hammerstein Benchmark system.

p434.

Identification of a Wiener–Hammerstein system using the polynomial nonlinear state space approach J. Paduart, L. Lauwers, R. Pintelon, J. Schoukens Control Engineering Practice, Volume 20, Issue 11, November 2012, Pages 1133-1139 In this paper, the Polynomial NonLinear State Space (PNLSS) approach is applied to model a nonlinear system with a Wiener–Hammerstein structure. To obtain good initial estimates, the best linear approximation of the system under test is first identified. Next, this linear model is extended to a polynomial nonlinear state space model to capture also the system's nonlinear behavior. The identification procedure is applied to measurement data

p435.

Identification of Wiener–Hammerstein models: Two algorithms based on the best split of a linear model applied to the SYSID'09 benchmark problem J. Sjöberg, L. Lauwers, J. Schoukens Control Engineering Practice, Volume 20, Issue 11, November 2012, Pages 1119 - 1125 This paper describes the identification of Wiener–Hammerstein models and two recently suggested algorithms are applied to the SYSID'09 benchmark data. The most difficult step in the identification process of such block-oriented models is to generate good initial values for the linear dynamic blocks so that local minima are avoided. Both of the considered algorithms obtain good initial estimates by using the best linear approximation (BLA) which can easily be estimated from data. Given the BLA, the two algorithms differ in the way the dynamics are separated into two linear parts. The first algorithm simply considers all possible splits of the dynamics. Each of the splits is used to initialize one Wiener–Hammerstein model using linear least-squares and the best performing model is selected. In the second algorithm, both linear blocks are initialized with the entire BLA model using basis function expansions of the poles and zeros of the BLA. This gives over-parameterized linear blocks and their order is decreased in a model reduction step. Both algorithms are explained and their properties are discussed. They both give good, comparable models on the benchmark data.

p436.

Reducing the Analog and Digital Bandwidth Requirements of RF Receivers for Measuring Periodic Sparse Waveforms Charles Nader, Wendy Van Moer, Niclas Björsell, Kurt Barbé and Peter Händel IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 11, November 2012, pp. 2960-2971 In this paper, a prototype setup for measuring wideband periodic waveforms whose bandwidth surpasses the analog bandwidth of a radio-frequency receiver is presented. Three major challenges arise in the analog-to-digital stage when measuring such wideband waveforms: the availability of a high sampling rate based on a good amplitude resolution; the availability of the required analog bandwidth to capture the full waveform; and achieving the previous requirements in a cheap way. Those challenges are more pronounced when using wideband modulated signals to test nonlinear devices and when measuring/sensing wideband spectra for cognitive radio applications. For periodic signals,

120

Bibliography

undersampling techniques based on the evolved harmonic sampling can be used to reduce the sampling rate requirements while satisfying a good amplitude resolution. For sparse signals, a technique based on channelization and signal separation is proposed. This technique splits the spectrum of the waveform into parallel channels, downconverts them to the analog frequency band of the analog-to-digital converter (ADC), spreads the channel information, sums them, and then digitizes with a single ADC. Using reconstruction algorithms based on l1-norm minimization, the information of the parallel channels can be separated. The original wideband spectrum can be then reconstructed after de-embedding of the channelization process. p437.

Approximate ML Estimation of the Period and Spectral Content of Multiharmonic Signals Without User Interaction Mikaya L. D. Lumori, Johan Schoukens, John Lataire IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 11, November 2012, pp. 2953-2959 The goal of this paper is to construct an approximate maximum-likelihood estimator to accurately estimate the period and spectral contents of a noisy periodic signal that has many frequency components. This is accomplished without user interaction. The signal data record has a total number of periods that is not necessarily an integer but is greater than four. Furthermore, the number of samples per period may not necessarily be an integer number. It is shown that the accuracy of the estimated results is superior to estimates that are devoid of variance weighting, such as those engendered by the least squares estimator.

p438.

Detection and Quantification of the influence of Time Variation in Frequency Response Function Measurements Using Arbitrary Excitations Rik Pintelon, Ebrahim Louarroudi, John Lataire IEEE Transactions on Instrumentation and Measurement, Vol. 61, No. 12, December 2012, pp. 3387-3395 This paper presents a nonparametric method for detecting and quantifying the influence of time variation in frequency response function measurements. The method is based on the estimation of the best linear time-invariant (BLTI) approximation of a linear time-variant (LTV) system from known input, noisy output data. The key idea consists in feformulating the single-input, single-output time-variant problem as a multiple-input, single-output timeinvariant problem. In addition to the BLTI approximation of the LTV system, the contribution of the disturbing noise, the leakage error, and the time-varying effects at the output is also quantified. As such, the approximation error of the time-invariant framework is known.

p439.

Peak-Power Controlled Digital Predistorters for RF Power Amplifiers P. Landin, W. Van Moer, M. Isaksson, P. Handel IEEE Transactions on Microwave Theory and Techniques, Vol. 61, No. 11, November 2012, pp. 3582-3596 This paper investigates the issue of “predistorter blow-up,” i.e., uncontrolled peak expansion caused by the predistorter. To control the peak expansion, an extension of the multistep indirect learning architecture (MS-ILA) is proposed by adding a constraint that describes the allowed peak power of the predistortion signal. The resulting optimization problem is shown to be convex and an optimization method is formulated to solve it. Measurements on a class-AB power amplifier (PA) using orthogonal frequency-division multiplex signals show that the peak control works as intended and prevents the MS-ILA from generating high peaks when the PA is operated in compression. The restriction on the peak power also prevents the performance degradation occurring due to the “blow-up” problem. This makes the proposed controlled MS-ILA a safer option than the standard MS-ILA. In addition to controlling the peak input power to the PA, using the proposed method it was possible to increase the output power by 1.3 dB while fulfilling requirements of less than 40-dB adjacent channel leakage power ratio, compared to the standard five-step MS-ILA. Reduced peak power also reduces the requirements on linearity in signal generation, resolution in computations, and analog-to-digital and digital-to-analog conversion.

p440.

Peak-Power Controlling Technique for Stabilizing Digital Pre-distortion of RF Power Ampli ers C. Nader, P. Landin, W. Van Moer, N. Björsell, P. Handel, D. Rönnow IEEE Transactions on Microwave Theory and Techniques, Vol. 61, No. 11, November 2012, pp. 3571-3581 In this paper, we present a method to limit the generation of signal peak power at the output of a digital pre-distorter that is applied to a RF power amplifier (PA) operating in strong compression. The method can be considered as a joint crest-factor reduction and digital pre-distortion (DPD). A challenging characteristic of DPD when applied to a PA in strong compression is the generation of relatively high peaks due to the DPD expansion behavior. Such high peaks generation, which may be physically unrealistic, can easily damage the amplification system. Such a phenomenon, referred in this study as DPD-avalanche, is more noticed when the signal exciting the PA is compressed due to crestfactor reduction. The suggested method for controlling such DPD-avalanche is based on shaping the input signal to the DPD in such a way to keep the pre-distorted signal peak power below or near the maximum allowed peak power of the PA. The suggested method is tested experimentally on a Class-AB and a Doherty PA when excited with a wideband orthogonal frequency-division multiplexing (OFDM) signal. Scenarios for an OFDM signal with and without crest-factor reduction are evaluated. Measurement results when using the proposed DPD-avalanche controller show smooth deterioration of the in-band and out-of-band linearity compared to steep deterioration when no controller is used. In addition, the suggested controller offers a higher operating power range of the DPD while fulfilling out-ofband distortion requirements and preserving low in-band error.

121

Annual report ELEC 2012

p441.

Modeling and digital predistortion of class-D outphasing RF power amplifiers P. Landin, J. Fritzin, W. Van Moer, M. Isaksson, A. Alvandpour IEEE Transactions on Microwave Theory and Techniques, Vol. 60, No. 6, June 2012, pp. 1907-1915. This paper presents a direct model structure for describing class-D outphasing power amplifiers (PAs) and a method for digitally predistorting these amplifiers. The direct model structure is based on modeling differences in gain and delay, nonlinear interactions between the two paths, and differences in the amplifier behavior. The digital predistortion method is designed to operate only on the input signals' phases, to correct for both amplitude and phase mismatches. This eliminates the need for additional voltage supplies to compensate for gain mismatch. Model and predistortion performance are evaluated on a 32-dBm peak-output-power class-D outphasing PA in CMOS with on-chip transformers. The excitation signal is a 5-MHz downlink WCDMA signal with peak-to-average power ratio of 9.5 dB. Using the proposed digital predistorter, the 5-MHz adjacent channel leakage power ratio (ACLR) was improved by 13.5 dB, from -32.1 to -45.6 dBc. The 10-MHz ACLR was improved by 6.4 dB, from -44.3 to -50.7 dBc, making the amplifier pass the 3GPP ACLR requirements.

4.3 c831.

CONFERENCE PAPERS (2012) Combining Broad-Band Multisines Excitations and Dielectric Spectroscopy for Non-Invasive Glucose Measurements Oscar Olarte, Wendy Van Moer, Kurt Barbé, Yves Van Ingelgem and Annick Hubin 5th International Conference on Advanced Technologies & Treatments for Diabetes (ATTD), Barcelona, February 8th and 11th, 2012 A common non-invasive glucose measurement technique is based on dielectric spectroscopy (DS). This technique allows measuring the properties of the system as a function of the frequency as well as distinguishing between the different processes that could be involved. When developing a non-invasive glucose measurement system, one must deal with different noise sources that need to be identified and quantified in order to provide confidence bounds for the estimated glucose level. Hence advanced signal processing techniques are needed to quantify, detect and discriminate the presence of noise sources as well as the non-linear distortions inherent to the system. Combining non-linear identification techniques and odd random phase multisines (ORPM), together with the (DS) will allow for an accurate non-invasive glucose measurement system. Method and results: Results based on a matrix of glucose and blood-protein at different concentrations show the ability of the approach to differentiate the glucose levels in amplitude and phase with great precision, as well as the estimation of the noise level. Experiments using more complex matrices in an incremental experimentation (saline solution, serum, plasma and cells culture, complete blood and finally over the skin) can show the influence of the different blood-components and environmental factors over the glucose impedance measurement. Conclusions: The capabilities of a ORPM to perform glucose measurements are studied. The measurement technique is based on DS and uses a ORPM excitations signal analysis. Although the results are preliminary, they clearly elucidate the capabilities of random phase multisine analysis for glucose measurements.

c832.

Separate Initialization of Dynamics and Nonlinearities in Nonlinear State-Space Models Anna Marconato, Jonas Sjöberg, Johan Suykens, Johan Schoukens I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2104-2108 This work focuses on the identification of nonlinear dynamic systems. In particular the problem of obtaining good starting values for the identification of nonlinear state-space models is addressed. A fast and efficient initialization algorithm is proposed, combining the use of methods from the statistical learning community to model the nonlinearities and classic system identification tools to capture system dynamics. The performance of the method is evaluated on simulation examples.

c833.

Identification and Modeling of Distillation Columns From Transient Response Data Diana Ugryumova, Gerd Vandersteen, Bart Huyck, Filip Logist, Jan Van Impe, Bart De Moor I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2098-2103 Mathematical models that predict the behavior of the column are highly useful to improve the working conditions of distillation columns, i.e. increase throughput or lower energy consumption. They are a prerequisite for advanced control strategies. White-box models of distillation columns already exist, but they are often too time consuming to calibrate in real industrial cases. This paper models a binary distillation column using a black-box model in the frequency domain. One of the identification challenges of distillation columns is the long transient response that causes leakage errors in the frequency domain. This problem is solved using a recently developed local polynomial method. Furthermore, we propose a simple expedient to model the variations in the column response that are caused by the changes in ambient temperature. The results are illustrated using both simulation data of a white-box model and using real measurements.

122

Bibliography

c834.

Nonparametric Estimation of the Instantaneous Transfer Function of Linear Periodically Time-Varying Systems Excited by Arbitrary Signals Ebrahim Louarroudi, John Lataire, Rik Pintelon I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1435-1440 Recently, a nonparametric identification scheme was developed to achieve high quality estimates of the evolution of the time-varying dynamics of continuous-time periodically timevarying systems (PTV). The method was founded upon a local polynomial approximation of the harmonic transfer functions using a single experiment, and this within an output-error framework. The proposed method imposes restrictions on the type of input (i.e. a broad band periodic signal), which could be a limitation in certain applications, especially where the user is not able to impose the type of excitation signal. In this work, this assumption is relaxed, such that the methodology presented here allows for arbitrary inputs as long as its power spectrum is band-limited. On top of that, we assume that an integer number of periods of the time-variation are observed. Due to the non-periodicity of the input-output a transient term will pop up in the expression of the frequency domain model which can also locally be very well described by a polynomial. The price being paid to relax the assumption to arbitrary inputs is that we can no longer distinguish the nonlinear distortions from the noise disturbances if the system behaves to some extent nonlinearly. Therefore, in this paper we restrict ourselves to linear PTV systems. The identification scheme is supported by simulations and real measurements to show the robustness of the proposed method.

c835.

Quasi-logarithmic Multisine Excitations for Broad Frequency Band Measurements Egon Geerardyn, Yves Rolain, Johan Schoukens I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 737-741 Logarithmically distributed excitation signals are widely used to measure the transfer function of dynamic systems in a wide frequency band, covering several decades. Periodic signals are very popular in advanced dynamic signal analyzers. Generating periodic signals requires an equidistant frequency grid which conflicts with the logarithmic distribution, especially at the low frequencies. In this paper we offer an elegant solution to get around this problem that also makes an improved choice for the amplitude spectrum.

c836.

Transient Suppression in Non-Parametric Frequency Response Function Estimates of Heat Diffusion Phenomena Griet Monteyne, Gerd Vandersteen, Rik Pintelon I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1983-1987 The modeling of heat diffusion phenomena is important to develop optimal control strategies for ground coupled heat pump systems. In order to find the best heat diffusion model, a good estimate of the non-parametric Frequency Response Function (FRF) is desired. Due to the large transients in the diffusion system, it is mandatory to take transient effects in the data into account. This paper compares two methods to suppress the transient effects. Both methods assume that the excitation signal is periodic. The first method uses time domain windowing to reduce the transient, while the second method uses the property that the influence of the transient can locally be approximated in the frequency domain by a low degree polynomial. It will be shown that the latter method reduces the variance of the transfer function estimate because it suppresses the transient more efficiently. The experimental verification is done using measurements on the heat diffusion problem in a thermally isolated bar.

c837.

Study of the maximal interpolation errors of the local polynomial method for frequency response function measurements Johan Schoukens, Gerd Vandersteen, Rik Pintelon, Z. Emedji, Yves Rolain I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1395-1399 Frequency response function measurements take a central place in the instrumentation and measurement field because many measurement problems boil down to the characterisation of a linear dynamic behaviour. The major problems to be faced are leakage- and noise errors. The local polynomial method (LPM) was recently presented as a superior method to reduce the leakage errors with several orders of magnitude while the noise sensitivity remained the same as that of the often most information about the system is to be retrieved, the dominating error is the interpolation error. In this paper it is shown that the interpolation error for sufficiently low damping is bounded by (BLPM/B3dB)R+2 with BLPM the local bandwidth of LPM, R the degree of the local polynomial that is selected to be even (user choices), and B3dB the 3dB bandwidth of the resonance, which is a system property.

c838.

Non-Parametric Best Linear Time Invariant Approximation of a Linear TimeVarying System John Lataire, Ebrahim Louarroudi, Rik Pintelon I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1050-1055 The consequences of estimating the frequency response function of a continuous-time, linear time-varying (LTV) system with tools for linear time invariant (LTI) systems are studied. To this end, the best linear time invariant approximation of an LTV system is defined, and is related to a general model for LTV systems. A recently introduced

123

Annual report ELEC 2012

frequency response function estimation method for LTI systems is used to compute the best linear time invariant approximation, the properties of which are discussed. An analysis of the residual error specifies whether the system under consideration is time-varying or not. Also, the frequency band where the contributions from the time variation are higher than the noise floor is determined. All concepts are illustrated on simulation examples. c839.

Parameter Reduction of MISO Wiener-Schetzen Models Using the Best Linear Approximation Koen Tiels, Peter S. C. Heuberger and Johan Schouken I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2114-2118 This paper concerns the identification of MISO (multiple inputs single output) Wiener systems. For each input-output path, the linear dynamics are modeled by a set of orthonormal basis functions (OBFs). The static nonlinearity is modeled through a multivariate polynomial. The parameters of the model are the coefficients of this polynomial. In this paper, an identification procedure for SISO (single input single output) Wiener systems is extended towards MISO Wiener systems. The poles of the OBFs are estimated using an extension of the best linear approximation (BLA) towards MIMO (multiple input multiple output) systems. As the number of parameters can increase significantly compared to the SISO case, a parameter reduction step, first developed for the SISO case, is extended in this paper to the MISO case. In each set of OBFs, one OBF is replaced by the BLA of the input-output path corresponding to that set. It is shown that in this way the number of relevantly contributing terms in the multivariate polynomial is significantly reduced. Simulation results show a major reduction of the number of parameters, with only a minor increase in the rms error on the simulated output.

c840.

Efficient use of short data records for FRF modeling by using fractional poles Kurt Barbé, Wendy Van Moer, Lieve Lauwers, Clare Ionescu I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1337-1342 Modeling systems based on measurements is a well established field of research and engineering practice. The techniques available to build and identify these models operate under the assumption that sufficiently many” measurements are available. In most cases, the model quality improves when the number of measurements increases. Unfortunately, measurement time is expensive and in some applications it is even infeasible to increase the number of measurements. For these kinds of applications, classical modeling tools become untrustworthy and no alternatives are available. In this paper, we introduce fractional order differential equations instead of ordinary differential equations to model linear systems. The major advantage of the presented technique is that only a small number of parameters is needed to obtain a very flexible model. We propose an identification technique which replaces the ordinary differential equations by fractional order differential equations with a smaller number of parameters.

c841.

Cognitive Radios: Discriminant Analysis Finds the Free Space Lee Gonzales Fuentes, Kurt Barbé, Wendy Van Moer, Niclas Björsell I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2242-2247 Spectrum sensing is an essential pre-processing step for cognitive radio technology. This paper presents a novel method to detect the significant spectral components in measured nonflat spectra, and to estimate the magnitude of the spectral components. Furthermore, the probability that the spectral component was incorrectly classified is available. The algorithm is able to detect the presence or absence of signals in any kind of spectrum since no prior knowledge about the measured signal is needed. Hence, this method becomes a strong basis for a high-quality operation mode of cognitive radios. Simulation results prove the advantages of the presented technique.

c842.

A robust signal detection method for fMRI data under correct Rice conditions Lieve Lauwers, Kurt Barbé, Wendy Van Moer I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1972-1976 In this paper, we tackle the problem of signal detection in functional Magnetic Resonance Imaging (fMRI) data by means of a statistical analysis. The main problem of the commonly used statistical tests is that they are based on the assumption that the data are Gaussian distributed, which is only valid for high signal-to-noise ratios (SNRs). Hence, for low SNRs the classical statistical tests are inadequate due to the wrong normality assumption, since it is known from literature that fMRI data follow a Rice distribution. In order to handle both high and low SNRs, we present in this paper a correction for the simplest and most widely used t-test by incorporating the correct Rice conditions. The performance of the Rice-corrected statistical test is shown through simulations and compared with its uncorrected counterpart.

c843.

Parallel Wiener identification starting from linearized models Maarten Schoukens, Yves Rolain I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1411-1416 This paper proposes a parametric identification method for parallel Wiener systems, starting from linearized models. Nonlinear models are often hard to obtain, while nonlinear measurements benches are hard to design. The method proposed in this paper starts from linear measurements and ends with obtaining a good nonlinear model. First, a

124

Bibliography

two-dimensional linear model depending on the input frequency and input signal power is measured. Next, the initial estimates of the linear time invariant blocks of the nonlinear model are obtained by extracting information of the dependency about the frequency dynamics on the input power. The static nonlinearities of the model are estimated using a linear least squares approach. Finally, these initial estimates are further refined using nonlinear optimization. The method is illustrated on a simulation and a validation measurement example to illustrate the theoretical efficiency and the practical usefulness of the estimator. c844.

Estimation of the Period and Spectral Content of Multi-Frequency Signals Using Minimal Data Without User Interaction Mikaya L.D. Lumori, Johan Schoukens, John Lataire, Rik Pintelon I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2502-2506 Least Squares (LS) and Weighted Least Squares WLS) cost functions are applied to estimate the period and spectral content of a periodic signal comprising many frequency components. The signal data record is more than two periods, and there is no user interaction. There is no need for synchronization between the generator and the data acquisition, where the sampling rates may be different and the number of samples per period may not be an integer number. An initial LS estimate of the signal spectrum is used to obtain an initial sample variance of the noise for weighting an initial WLS estimator. Using a sliding frequency domain window, a smooth sample variance of the noise is estimated from the signal spectrum. The smooth sample variance is then used as a weight in a second WLS estimator, different from the initial WLS estimator. It is shown that the estimate of the second WLS cost function is more accurate and superior to the estimates of both the initial LS cost function and the initial WLS cost function.

c845.

Spectrum Sensing through Spectrum Discriminator and Maximum Minimum Eigenvalue Detector: A Comparative Study Mohamed Hamid, Kurt Barbé, Niclas Björsell, Wendy Van Moer I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2252-2256 In this paper we present a new spectrum sensing technique for cognitive radios based on discriminant analysis called spectrum discriminator and compare it with the maximum minimum eigenvalue detector. The common feature between those two techniques is that neither prior knowledge about the system noise level nor the primary user signal, that might occupy the band under sensing, is required. Instead the system noise level will be derived from the received signal. The main difference between both techniques is that the spectrum discriminator is a non-parametric technique while the maximum minimum eigenvalue detector is a parametric technique. The comparative study between both has been done based on two performance metrics: the probability of false alarm and the probability of detection. For the spectrum discriminator an accuracy factor called noise uncertainty is defined as the level over which the noise energy may vary. Simulations are performed for different values of noise uncertainty for the spectrum discriminator and different values for the number of received samples and smoothing factor for the maximum minimum eigenvalue detector.

c846.

Using the Best Linear Approximation as a First Step to a New Non-Invasive Glucose Measurement Oscar Olarte, Wendy Van Moer, Kurt Barbé, S. Verguts, Yves Van Ingelgem and Annick Hubin I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2747-2751 In this paper an approach to develop a non-invasive glucose measured system based on the best linear approximation BLA) and electrical impedance spectroscopy (EIS) is presented. In order to test the advantages of the technique, three blood components at reference blood concentration are tested. In these solutions the level of glucose was varied from normal to pathological concentrations. The results show the ability of BLA-EIS to detect the glucose changes by combining the amplitude and phase behavior of the impedance.

c847.

Identification Algorithm

of

Time-varying

Systems

using

a

Two-dimensional

B-spline

Péter Zoltán Csurcsia, Johan Schoukens, István Kollár I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1056-1061 This paper presents a new method which non-parametrically estimates a two dimensional impulse response function of slowly time-varying systems. A generalized B-spline technique is used for double smoothing: once over the different excitation times (which refers to the system memory) and once over the actual excitation time (referring to the system behavior). If the change of the parameters of the observed system is sufficiently slow, with respect to the system dynamics, we will be able to 1) reduce the disturbing noise by additional smoothing 2) reduce the number of model parameters that need to be stored. c848.

The Best Linear Approximation of Nonlinear Systems Operating in Feedback Rik Pintelon, Johan Schoukens I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2092-2097

125

Annual report ELEC 2012

In the literature the best linear approximation BLA) has been defined and studied for nonlinear systems operating in open loop. The difficulty of the closed loop configuration is that the nonlinear distortions also perturb the input via the feedback loop. These input distortions bias the estimate of the BLA. To handle this problem we introduce a generalized definition of the BLA that is valid for nonlinear systems operating in feedback. The classical definition for open loop systems follows as a special case. A measurement procedure is proposed and illustrated on open loop gain measurements of an operational amplifier. c849.

Robust optimization method for the identification of nonlinear state-space models. Van Mulders, A., Vanbeylen, L., Schoukens, J. I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 1423-1428 A partially constrained optimization method is presented to estimate the parameters of a discrete-time nonlinear state-space model. Its advantage is its robustness towards instabilities: it can even be used to model unstable systems. A nonlinear least-squares optimization strategy is used, allowing to estimate the model parameters together with a user-selected set of states. The fraction (over time) of selected states determines the number of constraints in the optimization. Depending on this fraction, the algorithm is more robust towards instabilities but rather slow (many constraints), or faster but less robust (few constraints). A strategy (with effective state selection) is proposed that benefits from the advantages of both situations. An experimental data example illustrates how large data sets can be handled via this strategy, and that unstable regions can be crossed.

c850.

From two frequency response measurements to the powerful nonlinear LFR model Vanbeylen, L. I2MTC 2012, IEEE International Instrumentation and Measurement Technology Conference, Graz, Austria, May 13-16, 2012, pp. 2109-2113 Until now, in constrast to other block-oriented model structures, the nonlinear LFR model has received relatively little attention by the system identification and instrumentation and measurement communities. However, since it comprises a general multiple-input-multiple-output (MIMO) linear time-invariant part and a static nonlinearity (SNL), it allows one to represent any (complex) block-structure consisting of linear dynamic blocks and one SNL. This flexibility makes the LFR model an attractive candidate in real measurement applications. In this paper, a method is proposed for generating initial estimates of the nonlinear LFR model, starting from frequency response measurements carried out at 2 input amplitudes. In a first step, the MIMO linear dynamics are extracted from subspace representations of the linear models at both amplitudes, and in a second step, the SNL is identified from the input-output data, through the MIMO linear part. To support the theory, simulation examples are included, showing superior results compared to the linear models.

c851.

Using Kalman filtering to analyze oscillometric blood presssure waveforms José Antonio de la O Serna, Wendy Van Moer, Kurt Barbé MeMeA 2012 IEEE Symposium on Medical Measurements and Applications, Budapest, Hungary, May 18-19, 2012, pp. 29-34 Measuring the blood pressure as possible can save a lot of human lives. Hence, it is very important to find an optimal method to determine the systolic and diastolic pressures out of the measured oscillometric blood pressure wave-form. Recently, studies have been showing that, by working in the frequency domain, outperforming results could be obtained. Using the digital Taylor-Fourier transform (DTFT) even allows separating the breathing and cardiac activity that is present in the oscillometric waveform. Furthermore, an estimate of the frequency fluctuation can easily be obtained. In this paper, we will investigate whether or not a Kalman filtering implementation can provide better results that the DTFT analysis. In theory both approaches should be equally performing. Both techniques will be compared on measured oscillometric waveforms. Even if the alternating Kalman filter does not excel the DTFT algorithm in interharmonic rejection, it offers interesting signal decomposition alternatives.

c852.

An innovative oscillometric blood pressure measurement: getting rid of the traditional envelope Kurt Barbé, Wendy Van Moer MeMeA 2012 IEEE Symposium on Medical Measurements and Applications, Budapest, Hungary, May 18-19, 2012, pp. 261-266 Osillometric blood pressure devices are popular and are considered part of the family’s medical chest at home. the popularity of these devices for private use is not shared by the physicians mainly due to the fact that the blood pressures are computed instead of measured. The classical way to compute the systolic and diastolic blood pressure is based on the envelope of the oscillometric waveform. The algorithm to compute the blood pressures from the waveform is firm dependent, ofthen patented and lacks scientific foundation. In this paper, we propose a totally new approach. Instead of the envelope of the oscillometric waveform, we use a statistical test to pin-point the time instances where the systolic and diastolic blood pressures are measured in the cuff. This technique has the advantage of being mathematically well-posed instead of the ill-posed problem of envelope fitting. Hence, in order to calibrate the oscillometric blood pressure monitor is sufficient the make the statistical test unbiased.

c853.

Saving lives by integrating cognitive radios into ambulances Wendy Van Moer, Niclas Björsell, Mohammed Hamid, Kurt Barbé, Charles Nader MeMeA 2012 IEEE Symposium on Medical Measurements and Applications, Budapest, Hungary, May 18-19, 2012, pp. 141-144

126

Bibliography

A brain stroke is defined as a disturbance in the blood supply of the brain. This can be due to either an obstruction in the blood vessels of the brain or a rupture in the blood vessels which causes a leakage of blood in the brain. In many cases, a stroke results in the death of the patient within 24 hours. Hence, it is crucial that the neurologist has immediately contact with the patient in the first 30 minutes after the stroke. This means that a direct broadband communication link between the ambulance and the hospital is needed in order to transmit all necessary physiological parameters, such as blood pressure and glucose level as well as video images. In this paper, we present a new architecture of a wireless communication link between the ambulance and the hospital based on the concept of cognitive radios. The sender/receiver module in the ambulance will allow measuring the wideband spectrum and search for a suitable empty frequency band to send the data. c854.

A 52-66GHz Subharmonically Injection-Locked Quadrature Oscillator with 10GHz Locking Range in 40nm LP CMOS G. Mangraviti, B. Parvais, V. Vidojkovic, K. Vaesen, V. Szortyka, K. Khalaf, C. Soens, G Vandersteen, P. Wambacq 2012 IEEE Radio Frequency Integrated Circuits (RFIC) Symposium (RFIC), Montréal, Canada, June 17-19, 2012 A mm-wave subharmonically injection-locked quadrature oscillator is demonstrated in a 40nm low-power LP) digital CMOS technology. A large locking range 10GHz), tunable over the 52-66GHz band, is achieved using transformercoupled resonators. A simple calibration scheme is proposed that only relies on a relative power measurement of the oscillator output signal. The wide locking range, the wide tunability and the simple calibration scheme make this injection-locked quadrature oscillator design suitable for frequency synthesis i mm-wave CMOS communication systems.

c855.

Synchronizing modulated NVNA measurements on a dense spectral grid Y. Rolain, M. Schoukens, R. Pintelon, G. Vandersteen 79th ARFTG Microwave Measurement Conference, Non-Linear Measurement Systems, Convention Center, Montréal, Canada, June 22, 2012 An alternative synchronizer is proposed for mixer-based NVNA measurements of nonlinear devices excited by a modulated signal generated by an AWG. It consists of a binary signal generated by the AWG itself. Experimental results confirm a phase accuracy of ± 2o around 1 GHz

c856.

Dynamical systems and control mindstorms Ivan Markovsky proceedings of the 20th Mediterranean Conference on Control and Automation, Barcelona, Spain, July 3-6, 2012, pp.54-59 An unorthodox programme for teaching systems and control is developed and tested at the School of Electronics and Computer Science of the University of Southampton. Motivation for the employed teaching methods is Moore's method and S.~Papert's book ``Mindstorms: children, computers, and powerful ideas''. The teaching is shifted from lecture instruction to independent work on computer based projects and physical models. Our experience shows that involvement with projects is more effective in stimulating curiosity in systems and control related concepts and in achieving understanding of these concepts. The programme consists of two parts: 1) analytical and computational exercises, using \matlab/Octave, and 2) laboratory exercises, using programmable Lego mindstorms models. Both activities cut across several disciplines---physics, mathematics, computer programming, as well as the subject of the programme---systems and control theory.

c857.

How effective is the nuclear norm heuristic in solving data approximation problems? Ivan Markovsky Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 316-321 The question in the title is answered empirically by solving instances of three classical problems: fitting a straight line to data, fitting a real exponent to data, and system identification in the errors-in-variables setting. The results show that the nuclear norm heuristic performs worse than alternative problem dependent methods—ordinary and total least squares, Kung’s method, and subspace identification. In the line fitting and exponential fitting problems, the globally optimal solution is known analytically, so that the suboptimality of the heuristic methods is quantified.

c858.

Identification of the Silverbox Benchmark Using Nonlinear State-Space Models Anna Marconato, Jonas Sjöberg, Johan Suykens, Johan Schoukens Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 632-637 This work presents the application of an initialization scheme for nonlinear state-space models on a real data benchmark example: the Silverbox problem. The goal of the proposed approach is to transform the identification of a nonlinear dynamic system into an approximate static problem, so that system dynamics and nonlinear terms are identified separately. Classic identification techniques are used to handle dynamics, while regression methods from the statistical learning community are introduced to estimate the nonlinearities in the model. Results obtained on the Silverbox problem are discussed and compared with the performance of other related methods.

127

Annual report ELEC 2012

c859.

Structured low-rank approximation as a rational function minimization K. Usevich and I. Markovsky Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 722-727 Many problems of system identification, model reduction and signal processing can be posed and solved as a structured low-rank approximation problem. In this paper a reformulation of the structured low-rank approximation problem as minimization of a multivariate rational cost function is considered. We show that in two different parametrizations the problem is reduced to optimization on a compact manifold or to a set of optimization problems on bounded domains of Euclidean space. We make a review of polynomial algebra methods for global optimization of the rational cost function.

c860.

Robust Input Design for Resonant Systems under Limited a Priori Information Christian Larsson, Egon Geerardyn, Johan Schoukens Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 1611-1616 Optimal input design typically depends on the unknown system parameters that need to be we consider robust input design for resonant systems that may span over a large frequency use classical D-optimal design combined with a robust excitation signal which guarantees the regardless of resonance frequency. Simulations show that the proposed signal has thedesired

c861.

identified. In this paper band. The concept is to same estimate variance properties.

New Connections between Frequency Response Functions for a Class of Nonlinear Systems David Rijlaarsdam, Tom Oomen, Pieter Nuij, Johan Schoukens, Maarten Steinbuch Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, 280-285 The notion of frequency response functions has been generalized to nonlinear systems in several ways. However, a relation between different approaches has not yet been established. In this paper, frequency domain representations for nonlinear systems are uniquely connected. Specifically, by means of novel analytical results, the generalized frequency response function (GFRF) and the higher order sinusoidal input describing function (HOSIDF) for polynomial Wiener-Hammerstein systems are explicitly related. Necessary and sufficient conditions for this relation to exist and results on uniqueness and equivalence of the HOSIDF and GFRF are provided. Finally, a numerically efficient computational procedure is presented that allows to compute the GFRF from the HOSIDF and vice versa.

c862.

Classification of the Poles and Zeros of the Best Linear Approximations of WienerHammerstein Systems David T. Westwick, Johan Schoukens Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 470-475 The parameters of a Wiener-Hammerstein model, a nonlinear block structure comprising two linear filters separated by a memoryless nonlinearity, may be identified using an iterative nonlinear least squares optimization, however avoiding suboptimal local minima in the error surface requires a good initial estimate of the parameter vector. The Best Linear Approximation (BLA) of a Wiener-Hammerstein model will contain all the poles and zeros of both linear elements, but does not provide any information regarding which poles and/or zeros should be assigned to either of the linear elements. This information is contained in the nonlinear terms in the system response. One such nonlinear term is the BLA fitted between a suitably chosen nonlinear transformation of the input, and the output residuals remaining after all linear terms have been removed. The poles and zeros present in this nonlinear transfer function are used to classify the poles and zeros in the initial linear fit as belonging to either the first or second linear element in the Wiener-Hammerstein model. The procedure is illustrated by applying it to experimental data from a WienerHammerstein benchmark system.

c863.

Mean-Squared Error Experiment Design for Linear Regression Models Diego Eckhard, Häkan Hjalmarsson, Cristian R. Rojas, Michel Gevers Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, 1629-1634 This work solves an experiment design problem for a linear regression problem using a reduced order model. The quality of the model is assessed using a mean square error measure that depends linearly on the parameters. The designed input signal ensures a predefined quality of the model while minimizing the input energy.

c864.

Parametric Identification of Elastic Modulus of Polymeric Material in Laminated Glasses Erliang Zhang, Jean-Daniel Chazot, Jérôme Antoni Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 422-427

128

Bibliography

This paper addresses an inverse approach to characterize the frequency-dependent elastic modulus of the polymer layer in laminated structures. Represented by fractional derivative models, the modulus is identified based on a finite element model of the laminated structure from the experimental frequency response functions. An efficient Markov Chain Monte Carlo method is implemented to learn the identification parameters from a Bayesian perspective. A surrogate model is applied to alleviate Bayesian computation through the use of artificial neutral network. The proposed approach is experimentally validated on a laminated glass. c865.

Extension of Local Polynomial Method for Periodic Excitations Griet Monteyne, Diana Ugryumova and Gerd Vandersteen Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, 61-65 This paper extends the Local Polynomial Method (LPM) for linear and time invariant systems excited by periodic signals. LPM is a robust and fast method for finding a non- parametric Frequency Response Function (FRF) estimate. A good FRF estimate is important in designing a good controller. Since both the system FRF and the transient behave smooth as a function of the frequency, LPM assumes that these functions can be approximated locally by a low degree polynomial. However, if the FRF varies strongly as a function of the frequency this assumption results in bias errors due to under-modeling. That is why this paper presents a transient LPM. This transient LPM suppresses the transients as well as the original LPM but does not introduce bias errors due to under-modeling. The variance of the FRF estimate via the transient LPM will be slightly larger than the variance of the FRF estimate via LPM. However, when these non-parametric FRF estimates are used to find a parametric estimate, this variance difference will not affect the result. Thus, the reduced bias of the FRF estimate via the transient LPM will lead to a better parametric FRF estimate. A disadvantage is that the transient LPM cannot estimate the level of the nonlinear distortions.

c866.

The Use of Binary Sequences in Determining the Best Linear Approximation of Nonlinear Systems Hin Kwan Wong, Johan Schoukens, Keith Godfrey Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 1323-1328 This paper compares the performance of three different types of periodic binary sequences in the identification of the Best Linear Approximation of a nonlinear system. The signal types considered are discrete interval random binary sequences (DIRBS), maximum length binary sequences (MLBS) and inverse-repeat binary sequences (IRBS). It is found that MLBS’s offer advantages when experiment time limitation prohibits a large amount of averaging. IRBS’s have the advantage that even order nonlinear contributions do not affect the quality of the estimate, but the disadvantage of either a longer experiment time or a lower frequency resolution.

c867.

Interpolated Modeling of LPV Systems Based on Observability and Controllability J. De Caigny, R. Pintelon, J. F. Camino, J. Swevers Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, 1773-1778 This paper presents a State-space Model Interpolation of Local Estimates (SMILE) technique to compute linear parameter-varying (LPV) models for parameter-dependent systems through the interpolation of a set of linear timeinvariant (LTI) state-space models obtained for fixed operating conditions. Since the state-space representation of LTI models is not unique, a suitable coherent representation needs to be computed for the local LTI models such that they can be interpolated. In this work, this coherent representation is computed based on observability and controllability properties. It is shown that compared with the state of the art in the literature, this new method has three strong appeals: it is general, fully automatic and results in numerically well-conditioned LPV models. An example demonstrates the potential of the new SMILE technique.

c868.

User Choices for Nonparametric Preprocessing in System Identification J. Schoukens, G. Vandersteen, M. Gevers, R. Pintelon Y. Rolain Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 37-42 Most research on system identification is focused on the identification of parametric models, for example a transfer function or a state space model where the information is condensed in a few parameters. In the daily practice, nonparametric methods, like frequency response function measurements, are intensively used. Recently, it was indicated that nonparametric identification methods could be used to robustify the parametric identification framework. A nonparametric preprocessing step can also be used to reduce or even eliminate the required user interaction, making system identification accessible for a much wider user group. For that reason, there is an increasing interest in nonparametric identification. In order to choose, compare, and to benchmark these nonparametric methods, it is very important to select the proper criteria. In this paper we identify and discuss the important choices that should be considered. It will be shown that these strongly depend on the intended use of the nonparametric model.

c869.

Reducing the Number of Parameters in a Wiener-Schetzen Model Koen Tiels, Peter S.C. Heuberger, Johan Schoukens Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 1120-1124

129

Annual report ELEC 2012

The class of Wiener-Schetzen models can describe a large variety of nonlinear systems. In this paper the dynamical part of these models is formulated in terms of orthonormal basis functions, while the nonlinearity is modeled through a multivariate polynomial. The parameters of the model are the coefficients of this polynomial. Generally this polynomial contains a relatively large number of significant terms, resulting in a large number of parameters. This paper considers a reduction of the significant parameters, by replacing one of the basis functions by the so-called best linear approximation of the system. It is shown that in this way the number of relevantly contributing terms in the multivariate polynomial is significantly reduced. Simulation results show a major reduction in the number of parameters, with only a minor increase in the rms error on the simulated output. c870.

Initial Estimates for the LFR Nonlinear Model Structure Via the Best Linear Approximation Laurent Vanbeylen Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 13-18 In this paper, a novel method is proposed for the identification of a fairly general nonlinear structure called Linear Fractional Representation (LFR), also related to Lur'e type systems. It consists of one static nonlinearity (SNL) connected to the input and the output via a multiple-input-multiple-output (MIMO) linear time-invariant (LTI) block. The nonlinear LFR structure encompasses, e.g., Wiener-Hammerstein and nonlinear feedback models. The procedure starts from 2 state-space models corresponding to the best linear approximation at 2 input variance levels. The MIMO LTI block is estimated by exploiting the approximate structural relationships, taking the state transformations carefully into account. Using the measured input and output, the input and the output of the SNL block are then reconstructed, yielding a nonparametric estimate of the SNL, which is finally converted into a parametric estimate. In the whole procedure, the internal signals, the linear and the nonlinear part need not be known. No stability requirement was imposed on the linear models used. The functional form of the SNL is not needed to find the MIMO LTI block and the nonparametric SNL estimate. The simulation results, supporting the theory, show the superior quality of the obtained initial estimate of the LFR nonlinear model compared to both linear models. This shows that the method is a very promising approach in the field of block-oriented nonlinear modelling.

c871.

Model Reference Control Design by Prediction Error Identification Luciola Campestrini, Diego Eckhard, Alexandre S. Bazanella, Michel Gevers Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 1478-1483 This paper studies a one-shot (non-iterative) data-based method for Model Reference (MR) control design. It shows that the optimal controller can be obtained as the solution of a Prediction Error (PE) identification problem that directly estimates the controller parameters through a reparametrization of the input-output model. The standard tools of PE Identification can thus be used to analyze the statistical properties (bias and variance) of the estimated controller. It also shows that, for MR control design, direct and indirect data-based methods are essentially equivalent.

c872.

Identification of Hammerstein-Wiener Systems Maarten Schoukens, Er-Wei Bai, Yves Rolain Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 274-279 This work introduces a new formulation of the Hammerstein-Wiener system for identification purpose. Two methods to calculate the estimates are implemented: the overparametrization approach, and the iterative approach. The good performance obtained by both methods is shown with a simulation example.

c873.

The Transient Impulse Response Modeling Method and the Local Polynomial Method for Nonparametric System Identification Michel Gevers, Per Hägg, Håkan Hjalmarsson, Rik Pintelon, Johan Schoukens Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 55-60 This paper analyzes two recent methods for the nonparametric estimation of the Frequency Response Function (FRF) from input-output data using Prediction Error identification. Such FRF estimate can be the main goal of the identification exercise, or it can be a tool for the computation of a nonparametric estimate of the noise spectrum. We show that the choice of the method depends on the signal to noise ratio and on the objective. The method that delivers the best FRF estimate may not deliver the best estimate of the noise spectrum. Our theoretical analysis is illustrated by simulations.

c874.

Errors-In-Variables Identification of Linear Dynamic Systems Using Periodic Excitations R. Pintelon, J. Schoukens, and G. Vandersteen Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, 1365-1370 Using nonparametric noise models the complexity of the errors-in-variables problem is reduced to that of a generalised output error problem. Via experiments with periodic excitation signals one can easily obtain nonparametric estimates of the input-output noise models in a preprocessing step. The following assumptions are

130

Bibliography

hereby made: (i) the system operates in steady state, (ii) at least P=7 signal periods are available, and (iii) consecutive signal periods are independently distributed. Due to the noise colouring, assumption (iii) is an approximation. In addition assumptions (i) and (ii) reduce the frequency resolution of the experiment. In this paper we present a method that handles these three restrictions: 2 periods of the transient response to a periodic excitation are sufficient, and the correlation among consecutive signal periods is suppressed. c875.

Identification of a block-structured model with localised nonlinearity. Van Mulders, A., Vanbeylen, L. and Schoukens, J. Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 1125-1130 This paper considers the identification of a rather general nonlinear time-invariant system, consisting of a MultipleInput Multiple-Output (MIMO) linear dynamic part and one static nonlinear part. It is sometimes referred to as Linear Fractional Transformation (LFT) or Linear Fractional Representation (LFR). The structure will be called nonlinear LFR and includes many standard block-structured models, such as Wiener, Hammerstein, Wiener-Hammerstein and nonlinear feedback. The identification does assume neither the states, nor the internal signals over the static nonlinearity to be measured. The static nonlinearity (SNL) is assumed to be polynomial. After estimation of a nonlinear state-space model with certain structural properties, the SNL can be separated from the MIMO linear part. Next, the linear system is represented by a combination of four linear dynamic blocks, yielding extra insight. The method is illustrated via an experimental-data example.

c876.

Design and Application of Signals for Nonlinear System Identification W.D. Widanage, J. Stoev, J. Schoukens Proceedings of the 16th IFAC Symposium on System Identification, Brussels, Belgium, July 11-13, 2012, pp. 1605-1610 This paper discusses the design, implementation and the advantages of three types of signals for nonlinear system analysis and identification. They belong to the class of multisine signals and are the random phase, positively skewed and crest factor optimised multisine signals. A straightforward routine to combine such a signal with the system's typical input signal is discussed. The advantages of using such signals is illustrated through the results obtained from identifying the dynamics of a mechanical wet-clutch system.

c877.

Minimally invasive electrical bioimpedance characterization of in-vivo human lung tissue during the bronchoscopy procedure: a feasibility study Benjamin Sanchez, Gerd Vandersteen, Irene Martin, Diego Castillo, Alfons Torrego, Pere J. Riu, Johan Schoukens and Ramon Bragos 34th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC'12), San Diego, California, USA, August 28 - September 1, 2012. Lung biopsies are the basis for the diagnosis of lung cancer. Bronchoscopic lung biopsies fail to provide useful information in a significant number of cases, especially in diffuse lung disease, so more aggressive procedures are required. Success could be improved by using an electronic biopsy based on electrical bioimpedance (EBI) measurements. EBI can be obtained from several locations on the lung parenchyma through a bronchoscope, without removing any sample, resulting in a much less invasive procedure. EBI provides information about the dielectric characteristics of the tissue as a function of the frequency, which is related to biologically significant parameters such as cell size and shape, cell membrane state, cell concentration, intra- and extracellular compartments among others. Thus, EBI can reveal different physiological conditions by itself and also guide the biopsy during the procedure. The motivation of the manuscript is to evaluate the clinical feasibility for the materials and the measurement methods of a new instrument for minimally invasive in-vivo human lung tissue characterization.

c878.

Frequency Domain Analysis of Nonlinear Glucose Simulation Models Amjad Abu-Rmileh, Johan Schoukens 8th IFAC Symposium on Biological and Medical Systems, Budapest, Hungary, 29–31 August, 2012 Simulation models are frequently used in the development of the artificial pancreas for patients with diabetes. In this paper, frequency domain measurement techniques are used to perform a comparative analysis of widely used nonlinear simulation models of the glucose regulation system in type 1 diabetes. The analysis highlights the main differences between the models under study, based on a nonparametric estimate of their frequency response functions. The underlying linear dynamics, the nature and level of model nonlinearity, and the effect of nonlinear behavior on linear modeling are used as comparison criteria. The analysis shows that, a better understanding of the behavior of such nonlinear systems and the limitations of their linear approximates provides the means to a more careful use in simulations and control design.

c879.

Simple Polynomial Method For Optimizing Upstream Performance In Multiuser Vdsl2 Lines Hernan X. Cordova, Leo Van Biesen and Rob F.M. van den Brink XX IMEKO World Congress Metrology for Green Growth September 9-14, 2012, Busan, Republic of Korea In telecommunications, much efforts are spent nowadays to make ICT more Green. Rising energy costs, an economic slowdown and environmental wareness have introduced serious strategic challenges to enterprises worldwide. There

131

Annual report ELEC 2012

are a ariety of new and existing technologies available that aid in green IT, such as more efficient hardware that demand less power. In access networks, often the copper telephony network ensuring POTS is used to offer Digital Subscriber Line services to industry and private customers; e.g. to offer high bandwidth Internet access. The reduction of the electrical power in the modems (both in the Central Office and at the premises of the customers) for the transmission of the digital data is a key issue in making xDSL a more green ICT service. In this paper attention is paid to optimize Power Back off methods in upstream VDSL2. We study the excess of power that is required and the performance degradation that occurs when no power control measures are taken at the upstream of a Very highspeed Digital Subscriber Line (VDSL2). We propose a two-step approach: first, to perform an off-line exhaustive search algorithm that can find the optimal upstream power back-off (UPBO) parameters for almost every constraint within a cable bundle, following the current spectrum management specifications and regulations. As any exhaustive search approach, finding the parameters becomes typically very time-consuming, thus, secondly, we propose a simplified polynomial method based on the initial results such that the time to find the optimal parameters is significantly reduced. The algorithm turns out to be a pragmatic approach for current operators and performs well in practice though it does not exploit dynamic spectrum management (DSM) capabilities (being considered a Level-1 DSM-sort of algorithm) and relies on accurate insertion loss calculations for the different users distributed along the cable. However, in practice, given specific conditions per country, higher levels of DSM coordination might not be feasible; thus the motivation of our proposal. Simulation results show very similar results between these steps but at a reduced complexity for the polynomial approach, making the second method attractive for practical VDSL2 lines. We do not maximize the minimum capacity of the system; as it is common practice for operators to try to maximize the performance up to certain distance, we pursue this approach instead. Finally, we show how our proposed method could be adopted in a specific region/country via a spectrum management policy such that it can be enforced by the regulatory local entities ensuring that all operators properly follow it. We also demonstrate the severe impact when the policy is not properly applied by all DSL operators. c880.

Enhancing Gps Positioning Accuracy From The Generation Of Ground-Truth Reference Points For On-Road Urban Navigation Mussa Bshara, Umut Orgune, Fredrik Gustafsson and Leo Van Biesen XX IMEKO World Congress Metrology for Green Growth September 9-14, 2012, Busan, Republic of Korea The global positioning system (GPS) is a Global Navigation Satellite System (GNSS) uses a constellation of between 24 and 32 Medium Earth Orbit satellites that transmit precise microwave signals, which enable GPS receivers to determine their current location, the time, and their velocity [1]. Initially, the GPS was developed for military applications, but very quickly became the most used technology in positioning even for end-user applications run by individuals with no technical skills. GPS reading are used also as reference points for many positioning techniques such as the techniques that depend on the transmitted electromagnetic signal to determine the position of the transmitter or the receiver, due to their superior accuracy comparing to such techniques. But how accurate are those readings, and how to obtain accurate reference points starting from raw GPS observations even when they are corrupted with errors. In this paper, a practical study about GPS positioning is provided. Generating the ground-truth reference points depending on GPS observations is also provided and discussed in details.

c881.

Frequency domain, parametric estimation of the evolution of the time-varying dynamics of periodically time-varying systems from noisy input-output observation E. Louarroudi, J. Lataire, R. Pintelon, P. Janssens, J. Swevers ISMA - USD2012 International Conference on Uncertainty in Structural Dynamics, 17-19 September, 2012, pp. 2785-2800 This paper presents a parametric, frequency domain identification method for modeling continuous-(discrete-) time, periodically time-varying systems from input-output measurements. In this framework both the output as well as the input are allowed to be corrupted by stationary noise (= errors-in-variables approach). Furthermore, it is assumed that the system under consideration can be excited by a broad-band periodic signal with a user-defined amplitude spectrum (i.e. multisine), and that the periodicity of the excitation signal, Texc, can be synchronized with the periodicity of the time-variation, Tsys. Under these conditions the system can reach a steady state. Besides, two different concepts of a transfer function for time-varying systems (called the frozen transfer function and the instantaneous transfer function) are also introduced. A clear distinction between both is made, and either can be estimated with the proposed identification scheme. It is up to the users to decide which definition suits best their purpose. Uncertainty bounds on all/most frozen model-related quantity (such as frozen transfer function, frozen poles, frozen resonance frequency, ...) are provided in this paper as well. Finally, the identification algorithm is demonstrated on an extendible robot arm.

c882.

Flutter speed prediction based on frequency-domain identification of a timevarying system J. Ertveldt, J. Lataire, R. Pintelon, S. Vanlanduit ISMA - USD2012 International Conference on Uncertainty in Structural Dynamics, 17-19 September, 2012, pp. 3013-3024 This article introduces as new approach for flight envelope clearance of aircraft. This can now be performed as one continuous test, resulting in a major time saving. Both analysis of the current behaviour of the structure, and prediction towards higher velocities are important for flight flutter testing, and are dealt with in this article. The timevarying weighted non-linear least-squares estimator is used to obtain the frozen system poles of a cantilever wing under test in a low speed wind tunnel. The obtained smooth variation of the transfer function coefficients is used as basis for predicting the damping ratio towards higher velocities. Comparison with time-invariant measurements have

132

Bibliography

shown that only one tenth of the measurement time is needed to obtain equal uncertainties on the estimates when performing a flight flutter test as one continuous test. c883.

Determining the Dominant Nonlinear Contributions in a multistage Op-amp in a Feedback Configuration Adam Cooman and Egon Geerardyn and Gerd Vandersteen and Yves Rolain SMACD 2012 International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, Seville, Spain, 19-21 September 2012, pp. 205 -208 In this paper a simulation based method is proposed to determine the position of the dominant nonlinear contribution in the schematic of multistage op-amp operated in a feedback configuration. The key idea is to combine the Best Linear Approximation (BLA) and a classical noise analysis to determine the dominant source of nonlinear contributions. This results in a powerful yet simple design tool which does not require special analyses or custom models. As an example, the method is applied to a folded-cascode op-amp.

4.4 a253.

ABSTRACTS (2012) Nonlinear block-oriented identification for insulin-glucose models Anna Marconato, Maarten Schoukens, Koen Tiels, Abu-Rmileh Amjad, Johan Schoukens 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands The application of advanced identification techniques to model insulin-glucose systems represents a crucial step towards the development of the artificial pancreas for diabetes patients. Type 1 diabetes mellitus (T1DM) is a disease characterized by the fact that the pancreas is not able to produce a sufficient amount of insulin. Therefore, when treating patients with exogenous insulin delivery, the level of glucose in the blood needs to be carefully regulated to avoid severe problems such as hypoglycemia, retinopathy or cardiovascular diseases. Several mathematical descriptions (mainly first principle models) have been considered to represent the diabetic patient, and automated closed-loop control systems based on these models are currently under study. The main difficulties associated to the existing models are related to the fact that the tuning of parameters differs for each patient, and that the model parameters cannot be identified in practice. The objective of this work is the identification of models to describe the glucoregulatory system, based on input-output data. In particular, nonlinear system identification methods for block structures are combined with the use of nonlinear functions from statistical learning, e.g. Neural Networks (NNs).

a254.

Piezoelectric Tactile Tissue Measurement Challenges

Differentiation

Sensor

System:

Concepts

and

David Oliva Uribe, Jörg Wallaschek, Johan Schoukens 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands In this contribution the development of a piezoelectric tactile sensor system for the differentiation of soft tissues and phantoms is presented. The aim of this system is to provide a medical tool that can help neurosurgeons with the critical task of brain tumour resection, where in particular, the most important characteristic is to offer the surgeon the capability to find accurately tumour boundaries during surgery. The differentiation among distinct tissues and phantoms that have similar mechanical characteristics is a technique based on the detection and evaluation of different electrical parameters, where frequency response function measurements utilizing multisine excitation are performed to obtain the transfer function of the bimorph's voltages USensor/UActuator. The system was tested on a series of gelatine gel phantoms at different concentrations. The bimorph sensor system is able to detect even minimal differences. a255.

Identification and modeling of distillation columns from transient response data Diana Ugryumova, Gerd Vandersteen, Bart Huyck, Filip Logist 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands In this research, we model a binary distillation column using a frequency domain identification approach. The goal is to find an accurate but simple black-box model of a distillation column to be used in e.g. model predictive control. A distillation column is in essence a multiple-input-multipleoutput MIMO) non-linear system whose system dynamics vary due to changes in the ambient temperature. Here, we consider the modeling of a system which is not in steadystate, and hence introduces leakage errors in the frequency domain, and whose dynamics depend on external factor, i.e. the ambient temperature. In the next sections, firstly, the modeling approach is introduced. Secondly, the results of a binary distillation column modeling using simulation data are presented.

a256.

Nonparametric Identification of Linear Periodically Time-Varying Systems Using Arbitrary Inputs Ebrahim Louarroudi, John Lataire, Rik Pintelon

133

Annual report ELEC 2012

31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands Periodically time-varying (PTV) systems appear in a lot of engineering applications [1]. Examples can be found in control, sampled data systems, multi-rate filter banks, mechanical processes, and so on. For instance, mechanical systems that sustain a periodic motion (at constant angular speed) exhibit a PTV behavior due to the rotating parts in the system. Recently, a nonparametric scheme was developed to obtain an estimate of the evolution of the timevarying dynamics of continuous-time PTV systems, [2]. The proposed method imposes restrictions on the type of input (i.e. a broad band periodic signal). In this work, this assumption is relaxed to arbitrary inputs. a257.

How to obtain a broad band FRF with constant uncertainty? Egon Geerardyn, Yves Rolain, Johan Schoukens 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands Initial FRF measurements of an unknown system are often a cumbersome endeavor. How does one design a suitable excitation signal when very little is known about the system? Excitation signals with a logarithmic power spectral density (PSD) are often used, since these spread the available signal power evenly over the features of the system. For multisine excitations, such a PSD is to be approximated in the sense that the PSD is discrete and confined to a linear frequency grid. This results in a quasi-logarithmic (quasi-log) multisine. In this paper we present an elegant and simple way to design such a quasi-log multisine. This consists of a well-chosen amplitude spectrum such that it approximates the spectrum of the logarithmic signal closely and suitable choice of the spacing between frequency lines. Such a signal allows for a robust identification of unknown systems.

a258.

Robust or fast local polynomial method: How to choose? Griet Monteyne, Diana Ugryumova, Gerd Vandersteen, Rik Pintelon 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands The Robust (RLPM) and Fast (FLPM) Local Polynomial Methods were developed to find a non-parametric Frequency Response Function (FRF) estimate [1]. In this contribution the methods are compared and applied on experimental data coming from a heat diffusion experiment. Both methods assume that the excitation signal is a periodic signal and that the transient can be approximated locally in the frequency domain by a low-degree polynomial. FLPM also approximates the FRF by a local polynomial with a low degree. This approximation results in a bias error in case a low-degree polynomial cannot approximate the FRF well. RLPM does not approximate the FRF by a local polynomial. Thus, using RLPM avoids this bias error due to undermodeling. Unfortunately RLPM cannot estimate the level of the nonlinear distortions unless data coming from at least two experiments with uncorrelated inputs is available.

a259.

State Space Identification for Linear Parameter-Varying Systems Jan Goos and Rik Pintelon 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands Although the linear time-invariant (LTI) system identification framework has proven its merits for many years, in quite some applications the linearity and time-invariance hypotheses are only approximately true or not valid at all. The need to operate processes with higher accuracy and efficiency has therefore resulted in the realization that the non-linear (NL) and time-varying (TV) nature of many physical systems must be handled by the control design. In the linear parameter varying (LPV) framework, the dynamic relation between the input and output signals is still assumed to be linear, but it is continuously adapted based on the actual value of the scheduling parameters. Besides the excitation one should also choose a periodic or non-periodic scheduling. A common assumption is the bounded rate of variation of the coefficients and parameters.

a260.

Detecting the time variation in an assumed linear, time invariant measurement John Lataire, Ebrahim Louarroudi, Rik Pintelon 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands

a261.

Identification of nonlinear systems via a powerful block-oriented nonlinear model Laurent Vanbeylen 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands Nonlinear operation of devices is ubiquitous in practical application examples. When the nonlinearities are small, a linear system may be sufficient. In such a situation, a well-established framework of system identification theory and methods can be used. However, in nonlinear identification, there is still a lot of work to be done. This topic is concentrating on the identification of the most general block-oriented nonlinear model structure with the restriction that there is no more than one static nonlinearity (SNL), a.k.a. Linear Fractional Representation (LFR) model in the literature. Its flexibility (power) comes from the multiple-input-multiple-output, linear time-invariant (MIMO-LTI) part of the model, which realizes an arbitrary interconnection between the model’s in- and output and the SNL’s in- and output. The model encompasses, e.g., the Wiener-Hammerstein and nonlinear feedback block-oriented models.

134

Bibliography

a262.

Parameter reduction of SISO Wiener-Schetzen models Koen Tiels, Peter S.C. Heuberger, Johan Schoukens 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands The class of Wiener-Schetzen models can describe a large variety of nonlinear systems. The dynamical part of these models is formulated in terms of orthonormal basis functions (OBFs), while the nonlinearity is modeled through a multivariate polynomial. The parameters of the model are the coefficients of this polynomial. Generally, this polynomial contains a relatively large number of significant terms, resulting in a large number of parameters. This abstract is based on “K. Tiels, P.S.C. Heuberger, and J. Schoukens, Reducing the number of parameters in a WienerSchetzen model,” submitted for presentation at the 16th IFAC Symp. Syst. Identification, Brussels, Belgium, 2012.

a263.

Dielectric Spectroscopy for Non-invasive Glucose Measurements Oscar Olarte, Wendy Van Moer, Kurt Barbé, Yves Van Ingelgem, Annick Hubin 31th Benelux Meeting on Systems and Control, March 27-29 2012, CenterParcs Heijderbos, Heijden, The Netherlands A common non-invasive measurement technique is based on dielectric spectroscopy (DS). This technique allows measuring the properties of the system as a function of the frequency as well as distinguishing between the different processes that could be involved [1]. When developing a non-invasive glucose measurement system, one must deal with different noise sources that need to be identified and quantified in order to provide confidence bounds for the estimated glucose level. Hence advanced signal processing techniques are needed to quantify, detect and discriminate the presence of noise sources as well as the non-linear distortions inherent to the system. Combining non-linear identification techniques and odd random phase multisines (ORPM) [2], together with the DS will allow for an accurate non-invasive glucose measurement system.

4.5 w162.

WORKSHOPS (2012) Imposing structure on identified nonlinear state-space models Anne Van Mulders, Laurent Vanbeylen and Johan Schoukens MHE Workshop 2012 , OPTEC Workshop on Moving Horizon Estimation and System Identification, Leuven, August 29-30, 2012 "The flexibility of nonlinear state-space models makes them powerful tools in system identification. However, they usually consist of a high number of model parameters and are difficult to interpret and to handle in a control context - this in contrast to most block-oriented models. The state-space model in this talk has polynomial nonlinearities. A recently developed stepwise method in which structure is gradually imposed on the identified polynomial state-space model will be presented. The final model is the nonlinear LFR (Linear Fractional Representation) or Lur'e model. This model structure includes many standard block-structures and is the most general representation of a system with one (here assumed Single-Input Single-Output) static nonlinearity. It offers a good flexibility-sparsity trade-off. Several other approaches to reduce the number of parameters can be considered. The remaining open issues related to these approaches will be discussed. Two examples are: - Regularization techniques such as l1 norm regularization. The standard forms of these techniques seem to be uneffective for use on the nonlinear state-space model. - Searching for the (nonlinear) state transformation that generates the most zeros (or in practice parameters that become so unimportant that they can be withdrawn). This appears to be a very difficult task."

w163.

TISSUE DIFFERENTIATION SENSOR SYSTEM FOR BRAIN TUMOUR RESECTION BASED ON MULTISINE EXCITATION D. Oliva Uribe, J. Schoukens, R. Stroop and J. Wallaschek ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands We developed a tactile sensor based on a piezoelectric bimorph for brain tissue differentiation. The sensor is driven using multisine excitation to obtain the frequency response function (FRF) during contact between the sensor and the tissue. Evaluation of FRF allows identifying changes in the mechanical conditions of the tissue. The sensor is intended to be used in brain tumour resection, where the ability to detect minimal differences in tissue consistency is a vital task for the surgeon in order to remove safely a brain tumour. The results of measurements performed on gelatine phantoms as well as animal brain tissue are shown in this contribution.

w164.

On the use of non-white input for identification of errors-in-variables dynamical system E. Zhang, R. Pintelon, J. Schoukens ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands This work deals with the identification of dynamical systems from noisy input-output observations contaminated by white and uncorrelated Gaussian noises. The system identifiability is guaranteed by considering the non-white

135

Annual report ELEC 2012

spectrum property of the input. Next a maximum likelihood based identification strategy is developed in the frequency domain. In addition, the uncertainty of the system estimates is returned. The proposed algorithm is illustrated on the identification of the transmissibility function of a distributed dynamical system. w165.

Frequency Domain Modeling of Heat Transport around Borehole Heat Exchangers Griet Monteyne, Gerd Vandersteen ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands Ground coupled heat pump (GCHP) systems provide an energy-efficient alternative to traditional air-conditioning systems for space heating and cooling. Heat is extracted from or injected into the ground through vertical borehole heat exchangers, which form a so-called borefield. The energy efficiency of the GCHP depends on the temperature of the fluid entering the GCHP at the borefield side. In other words, the temperature of the fluid coming out of the borehole influences the energy efficiency of the GCHP. Thus, the knowledge of this temperature is important to develop a control strategy that minimizes the energy consumption. That is why we aim to find an accurate model for the relation between the thermal (heating or cooling) load of the ground and this temperature. The thermal response test was introduced by Mogensen P. [1] to determine the main thermal characteristics of the ground (resistance and conductance) that relate the thermal load to the temperature. Analytical or numerical [2, 3] techniques were developed to find estimates for these characteristics. However, all the developed techniques make use of approximations. The analytical techniques simplify the heat transfer problem into a line- or cylindrical- source problem in order to be able to find an analytical solution. This makes these techniques well suited to model the relation on a long time scale (> 24h). However, the short term response is less accurate and a constant heat input is required. The numerical techniques can handle both long and short time scales and variable heat input. However, both analytical and numerical techniques start from an approximate heat conduction equation, where in general the heat convection is neglected for simplification purposes. The relation between the load and the temperature will be estimated with a rational model (black-box) in the frequency domain. No assumptions for the parameters of this model will be made. This approach will lead to a model that reflects the relation between the temperature and the load present in the measurement data. The interpretation of the estimated model is less intuitive than the one found with the standard analytical and numerical approaches, but the model is expected to be more accurate on the short time scale (< 24h). In a following step, this more accurate model can be used to develop a control strategy that minimizes the energy consumption.

w166.

Dealing with Estimators.

Correlated

Errors

in

Least-Squares

Support

Vector

Machine

John Lataire (Vrije Universiteit Brussel) Dario Piga (Technische Universiteit Delft) Roland Tóth (Technische Universiteit Eindhoven) ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands Least-squares support vector machines (LS-SVM's) are known to be versatile estimators of static functions. They are stochastically unbiased in an output error framework, even if the output noise is correlated. However, they require the calibration of a set of hyper parameters to avoid overmodelling (to minimize the variance), which can be hampered if the output noise is correlated. In this poster, a frequency-domain formulation of the LS-SVM estimator of a static function is proposed to deal with correlated output noise. It relies on the fact that correlated noise, if it is stationary, is not correlated in the frequency domain. This ensures efficient function estimation via sound calibration of the hyper parameters. The method is shown to be applicable to the identification of time-varying FIR filters, where the output signal is disturbed by correlated measurement noise. w167.

Improved, user-friendly initialization method nonlinear LFR block-oriented model

for the identification of the

Laurent Vanbeylen ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands Nowadays, there is high need for accurate, parsimonious nonlinear models. Block structures are known to be excellent candidates for this task. The nonlinear LFR model, composed of a static nonlinearity (SNL) and a MIMO-LTI part, is highly flexible since is creates an arbitrary MIMO-LTI interconnection between the model's input and output and the SNL's input and output. It can create nonlinear feedback (which is very important in oscillators and mechanical applications) and incorporates e.g. the Wiener-Hammerstein model as a special case. Starting from 2 best linear approximations (corresponding to 2 different excitation amplitudes or set points) of the system, representing the SNL provisionally by a variable gain, the method generates the best possible MIMO-LTI configuration and estimates the SNL in an automated, user-friendly, and efficient way. The resulting model parameters are finetuned via a subsequent optimization. The method will be illustrated via experimental data (and possibly some simulations). w168.

Identification of a nonlinear LFR block-structure with two static nonlinearities Anne Van Mulders, Laurent Vanbeylen ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands In this poster, a block-oriented nonlinear model with two separate static nonlinearities and arbitrary linear dynamics is considered. This model class includes for example Hammerstein-Wiener, Wiener-Hammerstein and nonlinear feedback systems. It is believed to be a rather parsimonious representation with good approximation capabilities. The

136

Bibliography

structure is a nonlinear LFR system that consists of the interconnection between a multiple-input multiple-output linear dynamic part (of which one in- and output are assumed to be measured) and two single-input single-output static nonlinearities. The method is based on (polynomial) state-space representations with rank-2 nonlinear part and uses an overparameterisation approach. The method will be demonstrated on some examples. w169.

Accuracy-complexity trade-off benchmark case studies

of

some

nonlinear

identification

methods:

Anne Van Mulders, Laurent Vanbeylen, Anna Marconato ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands This poster shows a comparison of the results of several nonlinear system identification techniques on a number of benchmark examples. An effective graphical representation is used to evaluate the trade-off between model quality and number of parameters. An interpretation of the results allows one to group the families of the considered nonlinear models: block-oriented, nonlinear state-space, neural networks models, ... Some theoretical open issues need still to be addressed and will be raised during the discussion. w170.

Determining the Dominant Nonlinear Contributions in a multistage Op-amp in a Feedback Configuration Adam Cooman, Gerd Vandersteen ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands A simulation based method is proposed to determine the position of the dominant nonlinear contribution in the schematic of a multistage op-amp operated in a feedback configuration. The key idea is to combine the Best Linear Approximation (BLA) and a classical noise analysis to determine the dominant source of nonlinear contributions. This results in a powerful yet simple design tool which does not require special analyses or custom models.

w171.

Polynomial approximation errors for integrator frequency response functions Diana Ugryumova, Gerd Vandersteen ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands System identification is an important research field for a better understanding of dynamic systems. It is used to build an accurate model of a system, which could be used for control. One of the challenges in system identification is the reduction of errors due to transient effects and noise. A nonparametric transfer function estimation method was developed recently, called local polynomial method. This method approximates the transfer function and the transient contributions locally by a polynomial in frequency. For slowly varying transfer functions a low-order polynomial is sufficient. The results are at least as good as those of existing time-domain windowing methods. For quickly varying transfer functions, like integrators, the errors of polynomial approximation increase in the low frequencies. Here we derive a bound for those errors. Also, we explore the possibility of distinguishing between a pure integrator and a first-order transfer function with a slow pole from measurements.

w172.

Quasi-logarithmic Multisines for Broad Frequency Band Measurements Egon Geerardyn, Yves Rolain, Johan Schoukens ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands If only little is known about a dynamic system under test, but one wants to obtain a reasonable model for this system, it is necessary to use robust excitation signals that cover a wide frequency band. A logarithmic distribution of spectral lines in excitation signals is commonly used to measure the transfer functions of dynamic systems over a wide frequency band covering several decades. However, generating multitone periodic signals requires an equidistant frequency grid which conflicts with the logarithmic spacing of the excitation lines, especially at the low frequencies. In this work we offer an elegant solution to get around this loss in power density using an improved choice for the amplitude spectrum of the multisine. We also offer a simple way to tune the frequency spacing of such a signal to allow for reliable identification of the system under test. This guarantees that different systems with identical damping will be identified with identical uncertainty if the 3 dB bandwidth of their resonance is excited by four lines of the quasi-logarithmic multisine.

w173.

Iterative update of the pole locations in a Wiener-Schetzen model Koen Tiels, and Johan Schoukens ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands A large variety of nonlinear systems can be described by a Wiener-Schetzen model. In this model, the linear dynamics are formulated in terms of orthonormal basis functions (OBFs), while the nonlinearity is modeled by a multivariate polynomial. The coefficients of the polynomial are the parameters of the model. The use of OBFs allows to incorporate prior knowledge about the system dynamics into the model. A mismatch between the poles used to construct the OBFs and the true poles of the underlying system can be handled by increasing the number of OBFs. However, this results in a large number of parameters to be estimated, and eventually in a larger uncertainty of the estimated model. We propose an iterative method to update the pole locations of the model without making repeated

137

Annual report ELEC 2012

experiments. An initial estimate of the pole locations is obtained from the best linear approximation (BLA) of the system. The causes of the mismatch between the estimated and the true pole locations are analyzed. Each of the error sources is tackled, resulting in the iterative method. The method is illustrated on a simulation example. w174.

Combining the best linear approximation and dimension reduction to identify the linear blocks of parallel Wiener systems Maarten Schoukens (Dept. ELEC, VUB), Christian Lyzell (Division of Automatic Control, LiU), Martin Enqvist (Division of Automatic Control, LiU) ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands A Wiener model is a fairly simple, well known, and often used nonlinear block-oriented black-box model. A possible generalization of the class of Wiener models lies in the parallel Wiener model class. This poster presents a method to estimate the linear time invariant blocks of such parallel Wiener models from input/output data only. The proposed estimation method combines the knowledge obtained by estimating the best linear approximation of a nonlinear system with a dimension reduction method to estimate the linear time invariant blocks present in the model. The estimation of the static nonlinearity is fairly easy once the linear blocks are known.

w175.

Order selection of LPV State Space models using a subspace method in the frequency domain Jan Goos, John Lataire, and Rik Pintelon ERNSI'2012, European Research Network on System Identification, 23-26 September 2012, Maastricht, The Netherlands Although the linear time-invariant (LTI) system identification framework has proven its merits for many years, in quite some applications the time-invariance hypothesis is only approximately true or not valid at all. The need to operate processes with higher accuracy and efficiency has led the linear parameter-varying (LPV) model to gain in popularity. Here, the dynamics depend linearly on one or more scheduling parameters.The subspace method uses a series of algebraic operations to derive the dynamics of the device under test, purely from input and output behaviour. In the LTI case, first the order of the underlying system is estimated. Next, the extended observability matrix or the states are calculated, after which the state space coefficients can be deduced.In this poster, we extend this LTI subspace method to a LPV model that is affine in its scheduling parameters (that is, the dynamics depend linearly on one or more scheduling variables). We will show that for a periodic input and scheduling sequence, we can obtain a very sparse representation in the frequency domain.

w176.

Determining the Dominant Nonlinear Contributions in a multistage Op-amp in a Feedback Configuration Adam Cooman, Gerd Vandersteen DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, belgium, October 8, 2012 A simulation based method is proposed to determine the position of the dominant nonlinear contribution in the schematic of a multistage op-amp operated in a feedback configuration. The key idea is to combine the Best Linear Approximation (BLA) and a classical noise analysis to determine the dominant source of nonlinear contributions. This results in a powerful yet simple design tool which does not require special analyses or custom models.

w177.

Accuracy-complexity trade-off benchmark case studies

of

some

nonlinear

identification

methods:

Anne Van Mulders, Laurent Vanbeylen, Anna Marconato DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, belgium, October 8, 2012 This poster shows a comparison of the results of several nonlinear system identification techniques on a number of benchmark examples. An effective graphical representation is used to evaluate the trade-off between model quality and number of parameters. An interpretation of the results allows one to group the families of the considered nonlinear models: block-oriented, nonlinear state-space, neural networks models, ... Some theoretical open issues need still to be addressed and will be raised during the discussion. w178.

Accuracy-complexity trade-off benchmark case studies

of

some

nonlinear

identification

methods:

John Lataire , Dario Piga (TUDelft), Roland Tóth (TU Eindhoven) DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, belgium, October 8, 2012 Least-squares support vector machines (LS-SVM's) are known to be versatile estimators of static functions. They are stochastically unbiased in an output error framework, even if the output noise is correlated. However, they require the calibration of a set of hyper parameters to avoid overmodelling (to minimize the variance), which can be hampered if the output noise is correlated. In this poster, a frequency-domain formulation of the LS-SVM estimator of a static function is proposed to deal with correlated output noise. It relies on the fact that correlated noise, if it is stationary, is not correlated in the frequency domain. This ensures efficient function estimation via sound calibration

138

Bibliography

of the hyper parameters. The method is shown to be applicable to the identification of time-varying FIR filters, where the output signal is disturbed by correlated measurement noise. w179.

Combining the best linear approximation and dimension reduction to identify the linear blocks of parallel Wiener systems Maarten Schoukens, Christian Lyzell, Martin Enqvist DYSCO IAP study day, Château-Ferme de Profondval, Court-St-Etienne, belgium, October 8, 2012 A Wiener model is a fairly simple, well known, and often used nonlinear block-oriented black-box model. A possible generalization of the class of Wiener models lies in the parallel Wiener model class. This poster presents a method to estimate the linear time invariant blocks of such parallel Wiener models from input/output data only. The proposed estimation method combines the knowledge obtained by estimating the best linear approximation of a nonlinear system with a dimension reduction method to estimate the linear time invariant blocks present in the model. The estimation of the static nonlinearity is fairly easy once the linear blocks are known

w180.

Quasi-logarithmic Multisines for Broad Frequency Band Measurements Egon Geerardyn, Yves Rolain, Johan Schoukens DCT 24 hour meeting 2012, Dynamics and Control Technology, 1-2 November 2012, Deurne, The Netherlands. If only little is known about a dynamic system under test, but one wants to obtain a reasonable model for this system, it is necessary to use robust excitation signals that cover a wide frequency band. A logarithmic distribution of spectral lines in excitation signals is commonly used to measure the transfer functions of dynamic systems over a wide frequency band covering several decades. However, generating multitone periodic signals requires an equidistant frequency grid which conflicts with the logarithmic spacing of the excitation lines, especially at the low frequencies. In this work we offer an elegant solution to get around this loss in power density using an improved choice for the amplitude spectrum of the multisine. We also offer a simple way to tune the frequency spacing of such a signal to allow for reliable identification of the system under test. This guarantees that different systems with identical damping will be identified with identical uncertainty if the 3 dB bandwidth of their resonance is excited by four lines of the quasi-logarithmic multisine.

4.6

SEMINAR PRESENTATIONS ORGANISED BY THE DEPT. ELEC (2012)

1. Identifying a Wiener sytem using a variant of the Wiener G-Functionals Koen Tiels (Department ELEC) This paper concerns the identification of nonlinear systems using a variant of the Wiener G-Functionals. The system is modeled by a cascade of a single input multiple output (SIMO) linear dynamic system, followed by a multiple input single output (MISO) static nonlinear system. The dynamic system is described using orthonormal basis functions. The original ideas date back to the Wiener G-functionals of Lee and Schetzen. Whereas the Wiener G-Functionals use Laguerre orthonormal basis functions, in this work Takenaka-Malmquist orthonormal basis functions are used. The poles that these basis functions contain, are estimated using the best linear approximation of the system. The approach is illustrated on the identification of a Wiener system.

2. Multi- and Many-Core Processing in Embedded and Signal Processing Systems Lee Barford, Agilent Measurement Research Laboratory, USA, March 1, 2012 Multicore processors and graphics processing units (GPUs) have become available at sufficiently low cost and power consumption that it is desirable to use them in embedded and signal processing systems. Such systems are characterized by the need to operate continuously on streams of data. How then can the available parallelism effectively be used? Many operations in such systems (e.g., filtering, collecting time-varying summary statistics) may straightforwardly be parallelized in the common case that increasing latency slightly is acceptable in order to considerably increase throughput. One strategy is to buffer data then operate on it in a data parallel manner, either: Initiating parallel tasks operating on independent data segments, or Operating on the whole buffer with a series of GPU thread launches. Other algorithms common in embedded systems do not have readily apparent parallel implementations. One such operation is pattern matching with a finite state machine (FSM). The state of a FSM can depend on data received arbitrarily far in the past. Hence the simple strategy of buffering to provide data parallelism encounters considerable difficulties. However, an equivalent parallel prefix algorithm can be derived mechanically from any FSM. The talk concludes with an exploration of the performance of this approach on a practical example.

3. Computational inference in dynamical system Dr. Thomas Schön, Linköping University, Sweden, 6-7 June, 2012 The overall aim in this part of the course is to provide an introduction to the theory and application of computational methods (some of them only a couple of years old) for inference in dynamical systems. More specifically, the computational methods we are referring to are sequential Monte Carlo (SMC) methods (particle filters and particle smoothers) for nonlinear state inference problems and expectation maximisation (EM) and Markov chain Monte Carlo (MCMC) methods for nonlinear system identification. Dealing with the nonlinear system identification problem will require nonstandard combinations of these methods.

139

Annual report ELEC 2012

We will work almost exclusively with state-space models, linear models to introduce the methods and nonlinear models to illustrate the capabilities of the methods. It is our firm belief that even if you aim for solving nonlinear problems, you should always make sure that the method under study is capable of solving basic linear problems first. If that cannot be done, the method does not stand a chance in solving the nonlinear problem either. Furthermore, a good understanding of linear models is important in order to be able to understand nonlinear models. Our aim throughout this course is introduce the methods by answering simple questions relating to linear systems and then (most importantly) show that the methods are capable of tackling nonlinear problems as well. After a brief introduction we will derive general expressions for computing filtering and various smoothing densities for the states in nonlinear dynamical models. The basic strategies employed in both maximum likelihood (ML) and Bayesian system identification are then reviewed. After this short starting phase, we derive the EM algorithm and show how to use it to compute ML estimates in linear systems. We then turn our attention to the MCMC methods and show how these methods can be used to solve the linear system identification problem. This involves some interesting developments requiring the use of the matrix-Normal and the inverse-Wishart distributions and the so called simulation smoothers. The linear models have so far served the purpose of testing grounds for introducing the EM and the MCMC methods. However, it is now time to leave the linear models behind and turn our attention to nonlinear models instead. The SMC methods (focusing on the particle filter) will be introduced and the basic theory is provided. We will also show how the particle filter has been used to solve some nontrivial nonlinear filtering problems we have been working on together with various companies. The particle smoother is briefly introduced. Finally, we will show how the methods introduced above can be used to solve various problems in nonlinear system identification. We start by showing how to compute ML estimates using EM (involving particle smoothers and nonlinear optimisation) and show how this can be used to solve various problems, including some Wiener identification problems. Depending on how much time we have towards the end, we might also introduce a very recent (and exciting) development called the particle MCMC (PMCMC) method. Using PMCMC we are capable of solving nonlinear Bayesian system identification problems by a nontrivial combination of the MCMC methods and the SMC methods. One particular PMCMC algorithm will be illustrated by solving a rather challenging Wiener problem, where a non-monotonic nonlinearity is identified using a nonparametric description. This final part will also be presented (necessarily more compressed though) at SYSID in July (also conveniently located in Brussels as you all know).

4. Recent matrix and tensor decompositions in data mining and machine learning Mariya Ishteva, Georgia Institute of Technology, USA, 17 October 2012 Matrix low rank approximations have been successfully used to solve many data mining tasks. In this talk, we present a new bounded matrix low rank approximation that best approximates a given matrix with missing elements. The bounds are imposed on the approximation itself rather than individually on the low rank factors. This new approximation models many real world problems, such as recommender systems. We present an efficient algorithm based on coordinate descent method which outperforms existing algorithms in the literature. In the second part of the talk we will discuss existing and new connections between latent variable models from machine learning and tensors (multi-way arrays) from multilinear algebra. A few ideas have been developed independently in the two communities; however, there are still many useful but unexplored links. In particular, we will connect latent tree graphical models to state of the art tensor decompositions to solve structure learning problems and find tractable representations of probability tables of many variables. This is a joint work with Ramakrishnan Kannan, Le Song, and Haesun Park

5. Data-driven modeling: A low-rank approximation problem Ivan Markovsky, Department ELEC, 8 November 2012 The thesis of the presented research program is that various data-driven modeling problems can be posed and solved as a single core linear algebra problem: approximation of a structured matrix by a matrix with the same structure and lower rank. We call this problem "structured low-rank approximation". Different model classes correspond to different types of structure and the notion of model complexity corresponds to the notion of rank. Apart from precise mathematical formulation of data modeling problems, the low-rank approximation setting makes a link to standard solution methods, algorithms, and software. In our experience, methods based on the variable projection principle are most effective when the model complexity is small compared to the number of data points. The presented ideas are illustrated on a system identification problem with missing data and a data-driven simulation problem.

6. Positioning, Synchronizing and other Amenities Paolo Carbone, University of Perugia, Italy, 23 November 2012 The topic of indoor positioning has gained increasing interest over the years in the scientific literature because of the involved technical challenges and of the demand in the application market. While it can be envisioned that portable electronic devices will feature seamless outdoor/indoor positioning capabilities in the future, current technology does not enable it yet at acceptable cost–performance trade–offs. Several approaches have been proposed over time that addressed the positioning problem from various perspectives but, even though requests for applications are growing at a fast pace, technology still seems to lay behind. The main challenges in the radio indoor positioning field are the high accuracy level required by many applications and the features of the indoor propagation channel. This talk describes the design and realization of a 5.6 GHz ultra–wide bandwidth based position measurement system. The system has been entirely made using off–the–shelf components and achieves centimeter level accuracy in an indoor environment. It is based on asynchronous modulated pulse round–trip–time measurements. Both system level and realization details will be illustrated along with experimental results. Because of the nature of the measuring principle, a mechanism to synchronize sensor nodes is also obtained as a benefit. Applications of other technique developed during the research process will also complement the discussion.

7. Control, Identification and Cheap Computing Brett Ninness, University of Newcastle, Australia, 26 November 2012

140

Bibliography

I'll give a short introduction to my home research base in Australia, and then mention the essentials of some projects I've been involved with in nonlinear system identification, Bayesian estimation via Markov-chain Monte-Carlo methods, and real time model predictive control using custom computing architectures. The work has a common theme of employing the full capacity of current cheaply available computing platforms for problems in systems and control.

8. The nonlinear LFR block-oriented model: potential benefits and improved, user-friendly identification method Laurent Vanbeylen, Department ELEC, 7 December 2012 Nowadays, there is a high need for accurate, parsimonious nonlinear dynamic models. Block-oriented nonlinear model structures are known to be excellent candidates for this task. The nonlinear LFR model, composed of a static nonlinearity (SNL) and a multiple-input-multiple-output (MIMO) linear time-invariant (LTI) part, is highly flexible since is creates an arbitrary MIMO-LTI interconnection between the model’s in- and output and the SNL’s in- and output. It can create nonlinear feedback (which is very important in oscillators and mechanical applications), incorporates e.g. the Wiener-Hammerstein model as a special case and does not postulate the SNL's location prior to the identification. Starting from 2 classical frequency response measurements of the system, the method generates the best possible MIMO-LTI configuration and estimates the SNL in an automated, user-friendly, and efficient way. The method will be illustrated via simulation experiments and Silverbox measurements.

9. System identification using output-only signals Erliang Zhang, Department ELEC, 18 December 2012 Operational modal analysis has been receiving great attention for its capability of identifying mechanical systems excited by ambient sources (e.g., wind, traffic noise). The presentation aims to deliver an image of the operational modal analysis techniques especially in the regard of their performance when dealing with system nonlinearity. The talk will be given around the following ingredients, a) the basic ideas to derive several powerful operational modal analysis algorithms b) their performance assessment in regard of system nonlinearity c) perspective on the operational modal analysis by system identification techniques. The results are illustrated on numerical and real nonlinear mechanical systems.

4.7

PATENTS

1. TDR Based Transfer Function Estimation of Local Loop Tom Bostoen, Patrick Boets, Leo Van Biesen, Thierry Pollet and Mohamed Zekri European Patent Office, Application No./Patent No. 01400832.0-1246 2. Method and Apparatus for Identification of an Access Network by Means of 1-Port Measurements Tom Bostoen, Thierry Pollet, Patrick Boets, Mohamed Zekri and Leo Van Biesen European Patent Application, Application No. EP 1 248 383 A1 3. Method and Apparatus for Identification of an Access Network by Means of 1-Port Measurements Tom Bostoen, Thierry Pollet, Patrick Boets, Mohamed Zekri and Leo Van Biesen United States Patent Application Publication, Pub. No. US 2002/0186760 A1 4. Method for Matching an Adaptive Hybrid to a Line Tom Bostoen, Patrick Boets, Leo Van Biesen and Thierry Pollet European Patent Office, Application No./Patent No. 03292891.3 5. Interpretation system for interpreting reflectometry information T. Vermeiren, Tom Bostoen, Leo Van Biesen, Frank Louage, Patrick Boets European Patent Application, Application No. EP 1 385 724 A1 6. Interpretation system for interpreting reflectometry information T. Vermeiren, Tom Bostoen, Leo Van Biesen, Frank Louage, Patrick Boets United States Patent Application Publication, Pub. No. US 2004/0019451 A1 7. Localisation of Customer Premises in a Local Loop Based on Reflectometry Measurements Tom Bostoen, Thierry Pollet, Patrick Boets, Leo Van Biesen United States Patent Application Publication, Pub. No. US 2004/0022368 A1

141

Annual report ELEC 2012

8. Localisation of Customer Premises in a Local Loop Based on Reflectometry Measurements Tom Bostoen, Thierry Pollet, Patrick Boets, Leo Van Biesen European Patent Application, Application No. EP 1 388 953 A1 9. Signal Pre Processing for Estimating attributes of transmission Line Tom Bostoen, Thierry Pollet, Patrick Boets, Leo Van Biesen European Patent Application, Application No. EP 1 411 361 A1 10. Signal Pre Processing for Estimating attributes of transmission Line Tom Bostoen, Thierry Pollet, Patrick Boets, Leo Van Biesen United States Patent Application Publication, Pub. No. US 2004/0080323 A1 11. A method for determining bit error rates Vandersteen Gerd, Verbeeck Jozef, Rolain Yves, Schoukens Johan, Wambacq Piet, Donnay Stephane VUB/IMEC US60171854/US 2001 0044915 - US6678844, granted 13 January 2004 12. A method for determining signals in mixed signal systems Wambacq Piet, Vandersteen Gerd, Rolain Yves, Dobrovolny Petr VUB/IMEC US60138642 - EP1059594, granted 13 October 2004

4.8 PhD1.

DOCTORAL DISSERTATIONS Etude de la Production et des Conditions de Propagation d'Ondes de Choc Crs par un Plasma de Dcharge Jean Renneboog Doctoral Dissertation, Universit Libre de Bruxelles, 1967 Promoter: P. Baudoux (ULB)

PhD2.

Bijdrage tot het Fazeverschuivingen

Verwekken

en

Meten

van

Nauwkeurig

Bepaalde

Alain Barel Doctoral Dissertation, Vrije Universiteit Brussel, April 1976 Judging-committee: G. Maggetto (VUB), Verlinden (VUB), Hoffman (RUG), C. Eugène (UCL) Promoter: J. Renneboog (VUB) PhD3.

Maximum Informatie Extractie door middel van een Optimaal Frequentie Domein Experiment Guy Vilain Doctoral Dissertation, Vrije Universiteit Brussel, March 1983 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), J. Cornelis (VUB), C. Eugène (UCL), Kerkhof, J. Vereecken (VUB), G. Vansteenkiste (VUB) Promoter: J. Renneboog (VUB)

PhD4.

Foutdetectie op Electrische Lijnen met behulp van een Digitale Behandeling van het Reflectogram Leo Van Biesen Doctoral Dissertation, Vrije Universiteit Brussel, April 1983 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), Baert (RUG), C. Eugène (UCL), Goossens, Kirschvinck, J. Tiberghien (VUB) Promoter: J. Renneboog (VUB)

PhD5.

Parameterestimatie in Lineaire en Niet-Lineaire Systemen met Behulp van Digitale Tijdsdomein Metingen Johan Schoukens Doctoral Dissertation, Vrije Universiteit Brussel, February 1985 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), P. Eykhoff (TU Eindhoven), C. Eugène (UCL), Hoffman (RUG), Spriet (RUG), O. Steenhaut (VUB), G. Vansteenkiste (VUB)

142

Bibliography

Promoter: J. Renneboog (VUB) PhD6.

Active Microstrip Antennas Russell Dearnly Doctoral Dissertation, Vrije Universiteit Brussel, June 1987 Judging-committee: G. Maggetto (VUB), L.P. Ligthart (TU Delft), O. Steenhaut (VUB), G. Szymanski (Tech. University of Poznan), J. Tiberghien (VUB), A. Van De Capelle (KUL) Promoters: J. Renneboog (VUB), A. Barel (VUB)

PhD7.

Analysis and Application of a Maximum Likelihood Estimator for linear Systems Rik Pintelon Doctoral Dissertation, Vrije Universiteit Brussel, January 1988 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), P. Eykhoff (TU Eindhoven), A. van den Bos (TU Delft), J. Vandewalle (KUL) Promoters: J. Renneboog (VUB), J. Schoukens (VUB)

PhD8.

Time Division Multiplexing in Optical Fiber Networks Danny Sevenhans Doctoral Dissertation, Vrije Universiteit Brussel, June 1988 Judging-committee: G. Maggetto (VUB), P. Kool (VUB), E. Stijns (VUB), R. Blondel (Universit de Mons), P. Bulteel (Atea), C. Eugène (UCL), Baert (RUG) Promoters: J. Renneboog (VUB), A. Barel (VUB)

PhD9.

Design of Optimal Input Signals with Minimal Crest Factor Edwin Van der Ouderaa Doctoral Dissertation, Vrije Universiteit Brussel, Januari 1989 Judging-committee: G. Maggetto (VUB), F. Delbaen (VUB) , P. Eykhoff (TU Eindhoven), R. Pintelon (VUB), J. Renneboog (VUB), A. van den Bos (TU Delft), J. Vandewalle (KUL) Promoter: J. Schoukens (VUB)

PhD10.

Channel Multiple Access Protocols for a Hydrological Multihop Packet Radio Network Thomas J. Odhiambo Afullo Doctoral Dissertation, Vrije Universiteit Brussel, June 1989 Judging-committee: G. Maggetto (VUB), L. Van Biesen (VUB), J. Tiberghien (VUB), P. Van Binst (VUB), A. Van Der Beken (VUB) Promoter: A. Barel (VUB)

PhD11.

Steady-State Analysis of Strongly Nonlinear Circuits Eli Van Den Eijnde Doctoral dissertation, Vrije Universiteit Brussel, December 1989 Judging-committee: G. Maggetto (VUB), R. Pintelon (VUB), J. Renneboog (VUB), A. Barel (VUB), R. Van Geen (VUB), P. Eykhoff (TU Eindhoven), J. Vandewalle (KUL), R. Pollard (T.U. Leeds) Promoter: J. Schoukens (VUB)

PhD12.

Knowledge-Based Spectral Estimation James Ambani Kulubi Doctoral Dissertation, Vrije Universiteit Brussel, December 1989 Judging-committee: G. Maggetto (VUB), R. Pintelon (VUB), A. Barel (VUB), W. Verhelst (VUB), O. Steenhaut (VUB), Hoffman (RUG) Promoter: L. Van Biesen (VUB)

PhD13.

Radar Cross Section Reduction using Multiple - Layer Strip Gratings Gert Van Der Plas Doctoral Dissertation, Vrije Universiteit Brussel, February 1990 Judging-committee: G. Maggetto (VUB), R. Van Loon (VUB), A. Van de Capelle (KUL), D. De Zutter (RUG), P. Delogne (UCL) Promoters: A. Barel (VUB), E. Schweicher (KMS)

PhD14.

Automated Diagnosis for Arbitrary Digital Circuits Patrick Bakx Doctoral dissertation, Vrije Universiteit Brussel, May, 1990 Judging-committee: G. Maggetto (VUB), M. Goossens (VUB), A. Barel (VUB), M. Verlinden (VUB), V. Jonckers (VUB), P. Vandeloo (UIA)

143

Annual report ELEC 2012

Promoter: L. Van Biesen (VUB) PhD15.

Measuring Nonlinear Implementation

Systems

-

A

Black

Box

Approach

for

Instrument

Marc Vanden Bossche Doctoral dissertation, Vrije Universiteit Brussel, May, 1990 Judging-committee: G. Maggetto (VUB),R. Pollard (T.U. Leeds), P. Eykhoff (TU Eindhoven), D. De Zutter (RUG), Rik Pintelon (VUB), D. Ritting (Hewlett Packard - USA) Promoters: A. Barel (VUB), J. Schoukens (VUB) PhD16.

Identification Models

of Multi-Input

Multi-Output

Sytems

using Frequency-Domain

Patrick Guillaume Doctoral dissertation, Vrije Universiteit Brussel, June, 1992 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), M. Van Overmeire (VUB), P. Eykhoff (TU Eindhoven), M. Gevers (UCL), J. Vandewalle (KUL) Promoters: J. Schoukens (VUB), R. Pintelon (VUB) PhD17.

Identification of Linear Systems from Time- or Frequency-Domain Measurements: Part I: Parameter Estimation in Regression Models Applied to the Identification of Continuous-Time and Discrete-Time Systems Part II: Parameter Estimation of Superimposed Cisoids Applied to the Identification of Distributed Systems Hugo Van hamme Doctoral dissertation, Vrije Universiteit Brussel, June, 1992 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), F. Delbaen (VUB), M. Vanden Bossche (Hewlett Packard Belgium), A. van den Bos (TU Delft), B. De Moor (KUL), L. Ljung (University of Linköping) Promoters: R. Pintelon (VUB), J. Schoukens (VUB)

PhD18.

Identification of Linear Systems from Amplitude Information only Yves Rolain Doctoral dissertation, Vrije Universiteit Brussel, March, 1993 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), F. Delbaen (VUB), P. Eykhoff (TU Eindhoven), A. van den Bos (TU Delft), K. Godfrey (Univ. of Warwick, UK), J. Vandewalle (KUL) Promoters: J. Schoukens (VUB), R. Pintelon (VUB)

PhD19.

The Use of the Method of Moments in Designing NMR Antennas Guido Annaert Doctoral dissertation, Vrije Universiteit Brussel, December, 1993 Judging-committee: G. Maggetto (VUB), R. Van Loon (VUB), R. Luypaert (VUB-AZ), R. Turner, C. De Wagter (RUG), P. Van Hecke (AGFA-GEVAERT NV.), M. Lumori (VECO) Promoters: A. Barel (VUB), M. Osteaux (VUB-AZ)

PhD20.

Identification of Parametric Plane Wave Propagation Models for Underwater Acoustic Reflection and Transmission Experiments Luc Peirlinckx Doctoral dissertation, Vrije Universiteit Brussel, June 1994 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), L. Bjørnø (TU Denmark), P. De Wilde (VUB), H. Leroy (KULAK), J. Van Campenhout (UG), J.P. Sessarego (CNRS-LMA, France) Promoters: L. Van Biesen (VUB), R. Pintelon (VUB)

PhD21.

Radar Cross Section Calculations of Three-Dimensional Objects, Modelled by CAD Isabelle De Leeneer Doctoral Dissertation, Vrije Universiteit Brussel, March 1995 Judging-committee: G. Maggetto (VUB), V. Stein, D. De Zutter (UG), A. Van De Capelle (KUL), R. Van Loon (VUB) Promoters: A. Barel (VUB), E. Schweicher (KMS)

PhD22.

Design and Realization of Low Crest Factor Broadband Microwave Excitation Signals Tom Van den Broeck Doctoral Dissertation, Vrije Universiteit Brussel, September 1995 Judging-committee: G. Maggetto (VUB), J. Tiberghien (VUB), L. Martens (UG), R. Pollard (University of Leeds), M. Vanden Bossche (Hewlett Packard Belgium) Promoters: A. Barel (VUB), J. Schoukens (VUB)

144

Bibliography

PhD23.

Accurate Experimental Modelling of Bounded Wave Propagation in Viscoelastic Materials Dayu Zhou Doctoral Dissertation, Vrije Universiteit Brussel, October 1995 Judging-committee: G. Maggetto (VUB), P. Guillaume (VUB), L. Bjørnø (TU Denmark), H. Leroy (KULAK), M. Lumori (VECO), J.P. Sessarego (CNRS-LMA, France), I. Veretennicoff (VUB). Promoters: L. Van Biesen (VUB), L. Peirlinckx (VUB)

PhD24.

Calibration of a Measurement System for High Frequency Nonlinear Devices Jan Verspecht Doctoral Dissertation, Vrije Universiteit Brussel, November 1995 Judging-committee: G. Maggetto (VUB), J. Schoukens (VUB), A. Cardon (VUB), M. Vanden Bossche (Hewlett Packard, Belgium), L. Martens (RUG), B. Nauwelaers (KUL), U. Lott, A. Roddie, R. Pintelon (VUB), I. Veretennicoff (VUB) Promoter: A. Barel (VUB)

PhD25.

Performance with dielectric resonators at microwave frequencies for studying the pairing state in high-Tc supraconductors Andrei Mourachkine Doctoral Dissertation, Vrije Universiteit Brussel, January 1996 Judging-committee: W. Van Rensbergen (VUB), G. Van Tendeloo (VUB), J. Drowart (VUB), N. Klein (IFF, Julich), V. Gasumyants (St. Petersburg) Promoters: A. Barel (VUB), S. Tavernier (VUB), R. Deltour (ULB)

PhD26.

Identification of Linear and Nonlinear Systems in an Errors-in-Variables Least Squares and Total Least Squares Framework Gerd Vandersteen Doctoral dissertation, Vrije Universiteit Brussel, April 1997 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), M. Gevers (UCL), L. Ljung (University of Linköping), A. van den Bos (TU Delft), J. Vandewalle (KUL) Promoters: R. Pintelon (VUB), J. Schoukens (VUB)

PhD27.

Frequency Domain Identification of Transmission Lines from Time Domain Measurements Patrick Boets Doctoral dissertation, Vrije Universiteit Brussel, June 1997 Judging-committee: G. Maggetto (VUB), A. Cardon (VUB), A. Barel (VUB), M. Goossens (VUB), R. Pintelon (VUB), D. Baert (RUG), C. Eugène (UCL), J. Capon (Belgacom), J. Verspecht (Hewlett Packard, Belgium) Promoter: L. Van Biesen (VUB)

PhD28.

Nonparametric Identification of Nonlinear Mechanical Systems Stefaan Duym Doctoral dissertation, Vrije Universiteit Brussel, January 1998 Judging-committee: G. Maggetto (VUB), R. Pintelon (VUB), A. Barel (VUB), J. Vandewalle (KUL), J. Swevers (KUL), K. Worden (Univ. of Sheffield) Promoters: J. Schoukens (VUB), M. Van Overmeire (VUB)

PhD29.

Design of Digital Chebyshev Filters in the Complex Domain Rudi Vuerinckx Doctoral dissertation, Vrije Universiteit Brussel, February 1998 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), F. Grenez (ULB), I. Kollár (TU Budapest), McClellan (Georgia Institute of Technology), R. Pintelon (VUB), W. Verhelst (VUB) Promoters: J. Schoukens (VUB), Y. Rolain (VUB)

PhD30.

Caching in Dataflow-Based Instrumentation & Measurement Environments Eli Steenput Doctoral dissertation, Vrije Universiteit Brussel, October 1999 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), M. Goossens (VUB), E. Dierickx (VUB), H. Spoelder (Vrije Universiteit Amsterdam), E. Petriu (SITE-Ottawa) Promoters: Y. Rolain (VUB), J. Schoukens (VUB)

PhD31.

Standstill Frequency Response Measurement and Identification Methods for Synchronous Machines Jef Verbeeck

145

Annual report ELEC 2012

Doctoral dissertation, Vrije Universiteit Brussel, January 2000 Judging-committee: G. Maggetto (VUB), J. Vereecken (VUB), A. Barel (VUB), J. Deuse (Tractebel), I. Kamwa (Institut de Recherche d’Hydro-Qubec), J.C. Maun (ULB), J. Schoukens (VUB) Promoters: R. Pintelon (VUB), Ph. Lataire (VUB) PhD32.

Nonlinear Identification with Neural Networks and Fuzzy Logic Jürgen Van Gorp Doctoral dissertation, Vrije Universiteit Brussel, August 2000 Judging-committee: G. Maggetto (VUB), A. Barel (VUB), G. Horváth (Budapest University of Technology and Economics), P. Kool (VUB), Y. Rolain (VUB), J. Sjöberg (Chalmers University of Technology, Göteborg), J. Suykens (KUL) Promoters: J. Schoukens (VUB), R. Pintelon (VUB)

PhD33.

Patient and staff dosimetry in diagnostic radiology Jessica Pages Pulido Doctoral dissertation, Vrije Universiteit Brussel, September 2000 Judging-committee: A. Hermanne (AZ-VUB), J. Vereecken (VUB), P. Van den Winkel (VUB-cyclotron), M. Osteaux (AZ-VUB), H. Thierens (Universiteit Gent), M.A.O. Thijssen (St. Radboud Ziekenhuis Nijmegen), H. Mol (AZ-VUB), M. Sonck (VUB-cyclotron) Promoters: R. Van Loon (VUB)

PhD34.

Experimental Study of the Wave Propagation Through Sediments and the Characterization of its Acoustical Properties by Means of High-Frequency Acoustics Steve Vandenplas Doctoral Dissertation, Vrije Universiteit Brussel, May 2001 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), A. Barel (VUB), L. Bjørnø (Technical University of Denmark), J.P. Sessarego (CNRS-Marseille), O. Leroy (KULAK), L. Peirlinckx (Phonetics Topographics) Promoter: L. Van Biesen (VUB)

PhD35.

High Spatial Resolution Experimental Modal Analysis Steve Vanlanduit Doctoral Dissertation, Vrije Universiteit Brussel, May 2001 Judging-commitee: G. Maggetto (VUB), R. Arruda (Univ. Est. de Campinas, Brazil), A. Barel (VUB), R. Pintelon (VUB), J. Swevers (PMA - KULeuven), M. Van Overmeire (VUB), H. Van der Auweraer (LMS International)) Promoters: J. Schoukens (VUB), P. Guillaume (VUB)

PhD36.

Development of New Measuring and Modelling Techniques for RFICs and their Nonlinear Behaviour Wendy Van Moer Doctoral Dissertation, Vrije Universiteit Brussel, June 2001 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), R. Pollard (Univ. of Leeds, UK), D. Van Hoenacker (Unv. Catholique de Louvain), J. Schoukens (VUB) Promoters: Yves Rolain (VUB), Alain Barel (VUB)

PhD37.

Spectral and Kinetic Analysis of Radiation Induced Optical Attenuation in Silica: Towards Instrinsic Fibre Optic Dosimetry? Borgermans Paul Doctoral Dissertation, Vrije Universiteit Brussel, September 2001 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), J. Schoukens (VUB), I. Veretennicoff (VUB), B. Neerdael (SCK-CEN), M. Decrton (SCK-CEN), David Griscom (Naval Research Laboratory, USA)) Promoter: Alain Barel (VUB)

PhD38.

Multi-Carrier Modulation with Reduced Peak to Average Power Rati Zekri Mohamed Doctoral Dissertation, Vrije Universiteit Brussel, February 2002 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), A. Barel (VUB), J. Tiberghien (VUB), P. Boets (VUB), G. Vanhoutte (Belgacom), S. Popescu (Polytechnic Univ. of Bucharest) Promoter: L. Van Biesen

PhD39.

Parametric modeling and estimation of ultrasonic bounded beam propagation in viscoelastic media Bey Temsamani Abdellatif

146

Bibliography

Doctoral Dissertation, Vrije Universiteit Brussel, February 2002 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), A. Barel (VUB), D. Van Hemelrijck (VUB), L. Peirlinckx (Phonetic-Topographics, Ieper), O. Leroy (KULAK), Leif Bjørnø (Technical University of Denmark, Lyngby (DK)), JeanPierre Sessarego (Laboratoirre d'Acoustique et de Mecanique, CNRS, Marseille (F)) Promoter: L. Van Biesen PhD40.

Measurement and modelling of the noise behaviour of high-frequency nonlinear active systems Geens Alain Doctoral Dissertation, Vrije Universiteit Brussel, May 2002 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), R. Pintelon (VUB), R. Pollard (University of Leeds, UK), D. Van Hoenacker (UCL), J.C. Pedro (Universidade de Aveiro, Portugal), A. Barel (VUB) Promoter: Y. Rolain

PhD41.

Model Based Calibration of D/A Converters Vargha Balázs Doctoral Dissertation, Vrije Universiteit Brussel, June 2002 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), A. Barel (VUB), B. Bell (NIST, USA), I. Kollar (TUB, Hungary), Y. Rolain (VUB), G. Vandersteen (VUB-IMEC) Promoters: Johan Schoukens, István Zoltan (TUB)

PhD42.

Frequency Response Function Measurements in the Presence of Non-Linear Distortions Kenneth Vanhoenacker Doctoral Dissertation, Vrije Universiteit Brussel, June 2003 Judging-commitee: G. Maggetto (VUB), J. Vereecken (VUB), A. Barel (VUB), P. Guillaume (VUB), H. Sol (VUB), J. Swevers (KUL), H. Van der Auweraer (LMS International) Promoter: Johan Schoukens

PhD43.

Identification of Nonlinear Systems using Interpolated Volterra Models József G. Nmeth Doctoral Dissertation, Vrije Universiteit Brussel, June 2003 Judging-commitee: A. Barel (VUB), T. Dobrowiecki (Budapest University of Technology and Economics), M.P. Kennedy (University College Cork), R. Pintelon (VUB) Promoters: Johan Schoukens, István Kollár (TUB)

PhD44.

Identification of block-oriented nonlinear models Philippe Crama Doctoral Dissertation, Vrije Universiteit Brussel, June 2004 Judging-commitee: G. Maggetto (VYB), J. Vereecken (VUB), P. Guillaume (VUB), L. Ljung (Linköping Universitet, Sweden), M. Verhaegen (TUD, The Netherlands), J. Vandewalle (KUL), A. Barel (VUB), R. Pintelon(VUB) Promoters: Johan Schoukens, Y. Rolain

PhD45.

Identification of the Time Base in Environmental Archives Fjo De Ridder Doctoral Dissertation, Vrije Universiteit Brussel, December 2004 Judging-commitee: W. Baeyens (VUB), A. Barel (VUB), A. Berger (UCL), Ph. Lataire (VUB), G. Munhoven (Universit de Liège), D. Paillard (Centre d’Etudes de Saclay, Orme des Merisiers, France), J. Schoukens (VUB), J. Vandewalle (KUL), J. Vereecken (VUB) Promoters: R. Pintelon, Frank Dehairs

PhD46.

Optimisatie van patiënt dosissen, gekoppeld aan beeldkwaliteit, in de vasculaire radiologie Lara Struelens Doctoral Dissertation, Vrije Universiteit Brussel, January 2005 Judging-commitee: Promoter: R. Van Loon, co-promotors: H. Bosmans (KUL), F. Vanhavere (SCK-CEN)

PhD47.

Modelling, Designing and Developing a Multidisciplinary Geodatabase GIS with the Implementation of RDBMS in conjuction with CAD and different GIS applications for the development of Coastal/Marine Environment Tesfazghi Ghebre Egziabeher Doctoral Dissertation, Vrije Universiteit Brussel, September 2005

147

Annual report ELEC 2012

Judging-commitee: E. Vandijck (VUB), F. Canters (VUB), A. Barel (VUB), S. Wartel (Royal Belgian Institute of Natural Sciences), L. Peirlinckx (Phonetics Topographics, Belgium) Promoters: L. Van Biesen and Marc Van Molle (Geography dept., fac. of Sciences) PhD48.

Modeling of Substrate Noise Impact on CMOS VCOS on a Lightly-Doped Substrate Charlotte Soens Doctoral Dissertation, Vrije Universiteit Brussel, February 2006 Promoters: M. Kuijk, Y. Rolain, P. Wambacq, Judging-commitee: A. Barel, G. Gielen, Ph. Lataire, G. Maggetto, M. Tiebout, G. Van der Plas, J. Vereecken, P. Wambacq

PhD49.

Evaluation of deep-sub-quarter micron CMOS technology: low noise amplifiers, oscillators and ESD reliability Dimitri Linten Doctoral Dissertation, Vrije Universiteit Brussel, February 2006 Promoters: Y. Rolain, P. Wambacq and M. Kuijk Judging-commitee: G. Maggetto, J. Vereecken, A. Barel, M.I. Natarajan (Mentor IMEC), I. Smedes (Philips Semiconductors), M. Tiebout (Infineon)

PhD50.

Verification and Correction of Test Signals with a Spectrum Analyzer Daan Rabijns Doctoral Dissertation, Vrije Universiteit Brussel, March 2006 Promoters: G. Vandersteen, J. Schoukens Judging-commitee: P. Lataire, J. Vereecken, I. Kollar (Budapest Univ. of Techn. & Economics), D. DeGroot (CCNi Measurement Service), P. Guillaume, W. Van Moer

PhD51.

Impact and Mitigation of Analog Impairments in Multiple Antenna Wireless Communications Jian Liu Doctoral Dissertation, Vrije Universiteit Brussel, May 2006 Promoters: A. Barel, J. Stiens, G. Vandersteen Judging-commitee: P. Lataire, J. Vereecken, L. Van der Perre (IMEC), V. Öwall (Lund University, Sweden), W. Van Moer

PhD52.

Development and evaluation of a numerical method for the identification of a physical system described by a partial differential equation: a case study Kathleen De Belder Doctoral Dissertation, Vrije Universiteit Brussel, May 2006 Promoters: R. Pintelon, J. Schoukens Judging-commitee: Ph. Lataire, J. Vereecken, J. Swevers (KUL), P. Guillaume, H. Van der Auweraer (LMS International), H. Sol, P. Roose (Cytec)

PhD53.

Contributions to Large-Signal Network Analysis Frans Verbeyst Doctoral Dissertation, Vrije Universiteit Brussel, September 2006 Promoter: Y. Rolain Judging-commitee: J. De Ruyck, Jean Vereecken, Alain Barel, Don DeGroot (CCNi Measurement Services, Andrews University, Michigan, USA), Rik Pintelon, Roger Pollard (University of Leeds, UK), Johan Schoukens, Steve Vnlanduit

PhD54.

Contribution to severe weather and multimodel ensemble forecasting in Belgium David Dehenauw Doctoral Dissertation, Vrije Universiteit Brussel, November 2006 Promoters: A. Barel, H. Decleir Judging-commitee: Ph. Lataire, J. Vereecken, J. Schoukens, R. Willem, H. Malcoprs (KMI), A. Tijm (Kon. Ned. Metrologisch Inst.), F. De Meyer (KMI)

PhD55.

A system identification view on two aquatic topics: phytoplankton dynamics and water mass mixing Anouk de Brauwere Doctoral Dissertation, Vrije Universiteit Brussel, April 2007 Promoters: Willy Baeyens, Johan Schoukens Judging-commitee: Robert Finsy, Frank Dehairs, Rik Pintelon, An Smeyers-Verbeke, Joos Vandewalle (KUL), Eric Deleersnijder (UCL), Karline Soetaert (NIOO-KNAW), Johannes Karstensen (University of Kiel)

148

Bibliography

PhD56.

Ultra-Wideband transceiver for low-power low data rate applications Julien Ryckaert Doctoral Dissertation, Vrije Universiteit Brussel, May 2007 Promoters: Yves Rolain, Piet Wambacq (IMEC) Judging-commitee: Annick Hubin, Rik Pintelon, Gerd Vandersteen, J. Rabaey (University of Berkley, USA), M. Tiebout (Infeneon Germany), Christof Debaes, C. Desset (IMEC)

PhD57.

Measuring, modeling and realization of high-frequency amplifiers Ludwig De Locht Doctoral Dissertation, Vrije Universiteit Brussel, November 2007 Promoters: Yves Rolain,Gerd Vandersteen Judging-commitee: Annick Hubin, Rik Pintelon, Wendy Van Moer, Danielle VAnhoenacker (UCL), Andrea Ferrero (Politechnico di Torino), Christof Debaes (VUB), Marc Vanden Bossche (NMDG Engineering)

PhD58.

Body Area Communications: Channel system-level approach for low power

characterization

and

ultra-wideband

Andrew Fort Doctoral Dissertation, Vrije Universiteit Brussel, November 2007 Promoters: Leo Van Biesen, Piet Wambacq (IMEC) Judging-commitee: Annick Hubin, R. Pintelon, G. Vandersteen, Y. Hao (Univ. of London), C. Desset (IMEC) PhD59.

Algorithms for identifying guaranteed stable and passive models from noisy data Tom D’Haene Doctoral Dissertation, Vrije Universiteit Brussel, January 2008 Promoter: Rik Pintelon Judging-commitee: Gert Desmet, Jean Vereecken, Patrick Guillaume, Paul Van Dooren (UCL), Tom Dhaene (UGent), Martine Olivi (INRIA), Gerd Vandersteen

PhD60.

Identification of Nonlinear Systems Using Polynomial Nonlinear State Space Models Johan Paduart Doctoral Dissertation, Vrije Universiteit Brussel, January 2008 Promoters: Johan Schoukens, Rik Pintelon Judging-commitee: Annick Hubin, Jean Vereecken, Steve Vanlanduit, Lennart Ljung (Linköping University), Jan Swevers (KUL), Yves Rolain

PhD61.

GSM-based Positioning: Techniques and Application Nico Deblauwe Doctoral Dissertation, Vrije Universiteit Brussel, June 2008 Promoters: Leo Van Biesen, Prof. Dr. Claudia Linnhoff-Popien (Ludwig-Maximilians-Univ. Munchen) Judging-commitee: Dirk Lefeber, Rik Pintelon, Peter Schelkens, Wendy Van Moer, Luc Vandendorpe (Universit Catholique de Louvain), Luc Martens (Universiteit Gent), Fredrik Gustafsson (Linköping University

PhD62.

Resource-Aware Design of Smart Measurement Systems: A Learning-fromexamples Approach Anna Marconato Doctoral Dissertation, Vrije Universiteit Brussel - Università degli Studi di Trento, March 2009 Promoters: Prof. Dario Petri, Johan Schoukens, Bruno Caprile Judging-commitee: Annick Hubin, Gerd Vandersteen, Michel Verleysen (UCL), Davide Anguita (University of Genova), Anne Nowé

PhD63.

A framework for the analysis and modelling of substrate noise Stephane Bronckers Doctoral Dissertation, Vrije Universiteit Brussel, June 2009 Promoters: G. Van der Plas, G. Vandersteen Judging-commitee: A. Hubin, voorzitter, R. Pintelon, P. Wambacq, M. Nagat, (Kobe University, Japan), F.J. Clment, (Coupling Wave Solutions, France), W. Schoenmaker (Magwel, Belgium)

PhD64.

Identification and use of nonparametric noise models extracted from overlapping subrecords Kurt Barbé

149

Annual report ELEC 2012

Doctoral Dissertation, Vrije Universiteit Brussel, September 2009 Promoters: Rik Pintelon, Johan Schoukens Judging-commitee: Annick Hubin, Patrick Guillaume, Gerd Vandersteen, Lennart Ljung (Linköping University), Jrôme Antoni (Univ. de Technologie de Compiègne), Joos Vandewalle (KULeuven), Steve Vanlanduit PhD65.

Model Fitting in Frequency Domain Imposing Stability of the Model László Balogh Doctoral Dissertation, Vrije Universiteit Brussel, October 2009 Promoters: Rik Pintelon, István Kollár (TUBudapest) Judging-commitee: Johan Schoukens, Patrick Guillaume, Joos Vandewalle (KULeuven), Steve Vanlanduit, Barnabás Garay (TUBudapest)

PhD66.

CMOS building blocks for 60 GHz Phased-Array receivers Karen Scheir Doctoral Dissertation, Vrije Universiteit Brussel, December 2009 Promoters: Piet Wambacq, Yves Rolain) Judging-commitee: A. Hubin, R. Pintelon, G. Vandersteen, J. Long (TUDelft, Nederland), K. Halonen (Helsinki University of Technology, Finland), C. Debaes

PhD67.

Localization in wireless networks and co-existence of broadband services Mussa Bshara Doctoral Dissertation, Vrije Universiteit Brussel, June 2010 Promoter: Leo Van Biesen Judging-commitee: J. Tiberghien, R. Pintelon, P. Schelkens (IBBT), F. Gustafsson (linkoping Universitet), G. Vandersteen, P. Boets (Alcatel-Lucent-bell), L. Vandendorpe (UCL)

PhD68.

Advanced calibration and Instrumentation setups for nonlinear RF devices Liesbeth Gommé Doctoral Dissertation, Vrije Universiteit Brussel, August 2010 Promoter: Yves Rolain Judging-commitee: A. Hubin, R. Pintelon, G. Vandersteen, K. Godfrey (University of Warwick), D. Barataud (University of Limoges), M. Vanden Bossche (NMDG Engineering)

PhD69.

A Bayesian Model To Construct A Knowledge Based Spatial Decision Support System For The Chaguana River Basin Indira Nolivos Alvarez Doctoral Dissertation, Vrije Universiteit Brussel, October 2010 Promoters: Leo Van Biesen, Pilar Cornejo (ESPOL, Ecuador) Judging-commitee: J. Tiberghien, R. Pintelon, W. Bauwens, Pedro Girao (Universidade Tcnica de Lisboa), Rony Swennen (KUL), Ann Now

PhD70.

Best Linearized models for RF systems Koen Vandermot Doctoral Dissertation, Vrije Universiteit Brussel, October 2010 Promoter: Yves Rolain Judging-commitee: W. Bauwens, R. Pintelon, G. Vandersteen, D. Vanhoenacker (UCL), T. Dhaene (Universiteit Gent), M. Vanden Bossche (NMDG Engineering)

PhD71.

Time series reconstruction of environmental proxy records Veerle Beelaerts Doctoral Dissertation, Vrije Universiteit Brussel, January 2011 Promoter: Rik Pintelon, Frank Dehairs Judging-commitee: W. Bauwens, H. Terryn, J. Schoukens, J. Vandewalle (KUL), G. Munhoven (Université de Liège), D. Paillard (Lab. des Sciences du Climat et de l’environnement, Centre de Saclay, France), M. Elskens

PhD72.

Multirate Cascaded

S Converters for Wireless Applications

Lynn Bos Doctoral Dissertation, Vrije Universiteit Brussel, January 2011 Promoters: G. Vandersteen, Dr, ir. J. Ryckaert Judging-commitee: P. Guillaume, H. Terryn, P. Wambacq, P. Rombouts (Universiteit Gent), K. Makinwa (Delft University of Technology), B. Murmann (Stanford University)

150

Bibliography

PhD73.

Use and modeling of overtone resonances in FBAR resonators operating at RF frequencies Mohamed Reda Amin El-Barkouky Doctoral Dissertation, Vrije Universiteit Brussel, January 2011 Promoters: Y. Rolain, P. Wambacq Judging-commitee: D. Lefeber, H. Terryn, G. Vandersteen, B. Otis (University of Washington, Seattle, USA), J. Vandewalle (KUL)

PhD74.

Nonlinear and Dynamical Models for Temperature Reconstructions from Multi Proxy Data In Bivalve Shells Maite Bauwens Doctoral Dissertation, Vrije Universiteit Brussel, March 2011 Promoters: Johan Schoukens and Frank Dehairs Judging-commitee: Alan Wanamaker (Iowa State University, USA), Luc André (ULB-MRAC), Fjo De Ridder, Rik Pintelon, Willy Baeyens, Mark Elskens

PhD75.

Reflectometric Analysis of Transmission Line Networks Carine Neus Doctoral Dissertation, Vrije Universiteit Brussel, March 2011 Promoters: Leo Van Biesen, Yves Rolain Judging-commitee: Annick Hubin, Ludwig De Locht, Patrick Boets (Alcatel-Lucent, Belgium), Luc Martens (U-Gent), Tomas Nordström (FTW Telecommunication Research Center Vienna, Austria)

PhD76.

Frequency Domain Measurement and identification of Linear, Time-varying Systems John Lataire Doctoral Dissertation, Vrije Universiteit Brussel, March 2011 Promoter: Rik Pintelon Judging-commitee: Jérome Antoni (Université Compiègne, France), Lennart Ljung (University of Linköping, Sweden), Paul Van den Hof (Delft University of Technology, The Netherlands), Johan Schoukens, Patrick Guillaume, Herman Terryn, Steve Vanlanduit.

PhD77.

Nonlinear dynamic systems: blind identification of block-oriented models, and instability under random inputs Vanbeylen Laurent Doctoral Dissertation, Vrije Universiteit Brussel, May 2011 Promoters: Rik Pintelon, Johan Schoukens Judging-commitee: Hugo Sol, Herman Terryn, Gerd Vandersteen, Adrian Wills (University of Newcastle, Australia), Johan Suykens (KULeuven), Rodolphe Sepulchre (Université de Liège), Patrick Guillaume

PhD78.

Study of 3D position determination of the interaction point in monolithic scintillator blocks for PET Zhi Li Doctoral Dissertation, Vrije Universiteit Brussel, May 2011 Promoters: Stefaan Tavernier, Gerd Vandersteen Judging-commitee: Johan Schoukens, Michel Defrise, Karl Ziemons (University of Aachen, Germany), Jose Perez (Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas, Spain)

PhD79.

Some practical applications of the best linear approximation in nonlinear blockoriented modelling Lieve Lauwers Doctoral Dissertation, Vrije Universiteit Brussel, May 2011 Promoter: Johan Schoukens Judging-commitee: Gert Desmet, Herman Terryn, Kurt Barbé, Keith Godfrey (University of Warwick), Joos Vandewalle (KULeuven), Steve Vanlanduit

PhD80.

Design and evaluation of channel models for DSL applications Wim Foubert Doctoral Dissertation, Vrije Universiteit Brussel, November 2011 Promoters: Leo Van Biesen and Yves Rolain Judging-commitee: Willy Bauwens, Herman Terryn, Gerd Vandersteen, Luis Correia (Instituto Superior Técnico Lisboa), Christian Eugène (UCLouvain), Patrick Boets, (Alcates-Lucent).

151

Annual report ELEC 2012

PhD81.

Digital Basedband Modeling and Correction of Radio Frequency Power Amplifiers Per Landin Doctoral Dissertation, KTH School of Electrical Engineering, Stockholm, Sweden - Vrije Universiteit Brussel (cotutelle), June 2012 Promoters: Wendy Van Moer Judging-commitee:

PhD82.

Frequency Domain Based Performance Optimization of Systems with Static Nonlinearities David Rijlaarsdam Doctoral Dissertation, Cotutelle Technische Universiteit Eindhoven - Vrije Universiteit Brussel, June, 2012 Promoters: Maarten Steinbuch (TUE) - Johan Schoukens, P.W.J.M. Nuij (co-promoter TUE)

PhD83.

Tackling two drawbacks of polynomial nonlinear state-space models Van Mulders Anne Doctoral Dissertation, Vrije Universiteit Brussel, June 2012 Promoter: Johan Schoukens Jury: Hugo Sol, Herman Terryn, Gerd Vandersteen, Håkan Hjalmarsson (KTH), Thomas Schön (Linköping Sweden), Jan Swevers (KUL), Patrick Guillaume

PhD84.

Signal Shaping and Sampling-based Measurement Techniques for Improved Radio Frequency Systems Charles Nader Doctoral Dissertation, University of Gävle, Sweden - Vrije Universiteit Brussel (cotutelle), August 2012 Promoters: Prof. Nilas Björsel (University of Gävle); Prof. Wendy Van Moer (Vrije Universiteit Brussel) Jury: Prof. Heidi Ottevaere, Prof. Herman Terryn, Prof. Kurt Barbé, Dr. Lee Barford (Agilent Technologies); Prof. Håkan Hjalmarsson (KTH Royal Institute of Tehnology); Prof. Gurvinder Virk Singh (University of Gävle); Dr. Marc Vanden Bossche (NMDG)

PhD85.

Modelling and Optimization of Algorithms for Multiuser Multicarrier systems Cordova Junco Hernan Xavier Doctoral Dissertation, Vrije Universiteit Brussel, October 2012 Promoter: Leo Van Biesen Jury: J. Tieberghien, J. Deconinck, K. Steenhaut, Philippe De Doncker, Patrick Boets (Alcatel-Lucent-Bell), Lous M. Correia (IST-TV Lisbon, Portugal)

4.9

THESIS TOT HET BEHALEN VAN HET AGGREGAAT VAN HET HOGER ONDERWIJS

1. Sytem Identification. A Frequency Domain Modeling Approach Johan Schoukens Geaggregeerde van het hoger onderwijs, Vrije Universiteit Brussel, 1991 Judging-committee: G. Maggetto (VUB), G. Baron (VUB), G. Vansteenkiste (VUB), P. Eykhoff (TU Eindhoven), A. van den Bos (TU Delft), J. Vandewalle (KUL), M. Gevers (UCL) Promoters: J. Renneboog (VUB), A. Barel (VUB)

2. Frequency Domain Identification of Linear Time Invariant Systems Rik Pintelon Geaggregeerde van het hoger onderwijs, Vrije Universiteit Brussel, 1994 Judging-committee: G. Maggetto (VUB), A. Cardon (VUB), G. Baron (VUB), P. Eykhoff (TU Eindhoven), M. Gevers (UCL), A. van den Bos (TU Delft), J. Vandewalle (KUL), G. Van Steenkiste (RUG) Promoter: A. Barel (VUB)

152

Location of the university (VUB) and the dept. ELEC

5. Location of the university (VUB) and the dept. ELEC Getting to the department ELEC of the “Vrije Universiteit Brussel”

5.1

ARRIVAL BY CAR: Take the “Ring” and exit at the

crossing

motorway

E411

with

the

direction

centre of Brussels. At the end of the motorway, take the viaduct (go straight on), and at the second traffic lights, turn on right (“Triomflaan”), the VUB is situated on the left side of this road, starting from entrance 6 (see next map).

5.2

FROM THE BRUSSELS NATIONAL AIRPORT AT ZAVENTEM:

Brussels International Airport is at Zaventem, 14 km from the city centre. Information can be obtained by phone: Tel +32 2 753 42 21 / +32 2 723 31 11 Flight information: Tel +32 900 70 000 (7 a.m. - 10 p.m.) www.brusselsairport.be

153

Annual report ELEC 2012

From the airport, every 20 minutes the rail shuttle quickly takes you to the North Station in the centre of Brussels. At the North Station (“Bruxelles Nord”), you take the train to Etterbeek (direction Etterbeek or Louvain La Neuve) and get off the train in Etterbeek, which is 10 min. walking distance from the VUB. You only pay € 6.00 for a standard jump ticket (a taxi or cab from airport to the University is about € 50.00). More information and timetables of the Belgian railways: www.b-rail

5.3

FROM BRUSSELS SOUTH AIRPORT (CHARLEROI)

Situated to the south of Brussels, approximately 60 km away, Brussels-South Charleroi airport mainly houses low-cost airlines. www.charleroi-airport.com A bus links Charleroi Brussels-South and the Gare du Midi railway station in Brussels more than 20 times a day. The timetables are organised to coincide with Ryanair airline flights. Brussels to Charleroi: The shuttle departure point is situated at the junction of rue de France and rue de l'Instruction (follow "Thalys" exit at the Gare du Midi station). Charleroi to Brussels: shuttle departs 30 minutes after the Ryanair airline flight arrives at the airport. One-way ticket fare: 10.00 € (tickets are sold inside the shuttle)

5.4

ARRIVAL BY TRAIN:

Change in “Bruxelles Nord” and take the train to Etterbeek (direction Etterbeek or Louvain La Neuve) www.b-rail.be

154

Location of the university (VUB) and the dept. ELEC

5.5

ARRIVAL BY SUBWAY (€ 2,00/JUMP-TICKET):

Take line 5 direction “Hermann-Debroux” and get off at “Petillon”, which is also 10 min. walking distance from the VUB. 1 single fare JUMP (purchased outside vehicle) 2,00 € 1 single fare JUMP (purchased inside vehicle) 2,50 € More information about Brussels subway: www.stib.be The dept. ELEC is located in building K, 6th floor.

155