2014

MAINTENANCE

SUPREME SUstainable PREdictive Maintenance for manufacturing Equipment 1

2

2

3

4

5

S.Sieg-Zieba , T. Adolf , D. Lucke , R. Haug , P. Boulet , J.García-Sedano 1

CETIM, France, [email protected] Fraunhofer IPA, Germany, [email protected], [email protected] 3 Loy & Hutz, Germany, [email protected] 4 Cofely Endel, France, [email protected] 5 OPTIMITIVE, Spain, [email protected] 2

Keywords (max. 5) – Predictive maintenance, Condition Monitoring, Energy efficiency, Paper industry

Abstract Productivity improvements have major impact on economy and competitiveness in European manufacturing industry. Industrial maintenance contributes largely to this competitiveness through reliability and availability of production equipment. Especially in continuous production industries (energy, chemical, food, cement or paper sectors) the ratio “maintenance costs/added value product” is even higher than 25%. Defect components or process failures can stop the whole production, therefore predictive maintenance is a critical issue. In this context, the SUPREME project will provide new tools to dynamically adapt the maintenance and operation strategies to the current condition of the critical components of production equipment. It also proposes to develop an integrated approach to optimize the production process and its energy consumption.

1. Introduction Started in September 2012, the SUPREME consortium integrates key technical players on maintenance added value chain. The consortium gathers ten partners (Picture 1). Three of them are SMEs (EC Systems, Loy & Hutz and Optimitive), with RTD capacities to make possible the development of three modules which will be integrated to set-up the complete approach (ECMS (Embedded Condition Monitoring System), Reliability and Maintainability Module and Intelligent Control Module). The research work is conducted by teams from Grenoble-INP, CETIM, Fraunhofer IPA and CVUT. The application case in the paper industry is driven by Orloga and Condat (Lecta group). While the experience of Cofely Endel in maintenance will ensure that the new developments are applicable for various industrial fields.

Picture 1 SUPREME Presentation The main objectives of the project are to: - develop and use most advanced signal and data processing dedicated to predictive maintenance and energy consumption reduction, - enhance and develop new maintenance tools, - implement all these tools in an industrial demonstrator (paper industry), During the first year of the project, several results have already been achieved. One of the first steps was to develop the SUPREME Reference Model, which is the foundation for the whole SUPREME System and the SUPREME Modules. It structures and explains the major concepts, semantics and relationships within the SUPREME System.

Furthermore, it is intended to be used as a foundation for implementations. In parallel, the test platform specification for the project was defined, based on the experience of the paper industry partners. It enabled to select the most critical equipment and the pertinent one to demonstrate the SUPREME concepts and tools. The whole instrumentation was defined and then set up at the paper industry to start gathering field data very early in the project. Dealing with dissemination, the first E-learning module has been released, to present the challenges of the predictive maintenance.

2. Predictive maintenance In the european standard EN 13306, predictive maintenance is defined as a ”condition based maintenance carried out following a forecast derived from repeated analysis or known characteristics and evaluation of the significant parameters of the degradation of the item (EN 13306)”. Compared to a corrective maintenance or a systematic maintenance it is based on observing, measuring the state of the machine and only planning maintenance when a failure is detected. It is not only creating alarms when values are above thresholds, but it also implies being able to do a correct ”forecast” on the speed at which the deterioration will progress.

3. The SUPREME Reference Model The SUPREME Reference Model offers an abstract framework explaining the major concepts, semantics and relationships of a so called SUPREME System. As it is in the nature of a reference model, the SUPREME Reference model is not explicitly tied to any standards or technologies. The SUPREME Reference Model consists of three partial models: • The SUPREME Method provides a generic process to set up and operate a SUPREME System. • The SUPREME Structural Model shows the different generalized functions of a SUPREME System. • The SUPREME Information Model shows the different information objects required to operate a SUPREME System. 3.1 The SUPREME Method The SUPREME Method provides the different steps of a generic process to set up and operate a SUPREME System in order to achieve a sustainable predictive maintenance (Picture 2). Compared to other methods the SUPREME

2014

MAINTENANCE

Method addresses not only maintenance related improvement but also process and energy consumption improvement.

Production and maintenance system analysis Risk assessment and risk managment Focused component condition data acquisition Deterioration and failure prediction Maintenance related availability improvement of the production system

Focused process data acquisition

Process and energy parameter improvement

Picture 2 Key steps of the SUPREME Method The SUPREME Method starts with the analysis of the production and maintenance system. This step builds the foundation for the following process steps. This step is triggered after the first time by changes in management goals, products, production and production goals. For this analysis a wide range of so called high level production and maintenance data is required. The high level data comprise the data of the maintenance organization, the production system, the production processes, the products and the maintenance specific data of machines, equipment or components. Results are the structured as-is situation of the production system and maintenance system, the selected maintenance key performance indicators (KPI), energy and process parameters to be monitored and optimized. Next step in the maintenance improvement branch is the risk assessment of the production system. It identifies systematically the critical machines or equipment relevant for the value chain, considering holistic costs (direct and indirect failure costs). In the risk management these critical machines are structured into components as a foundation for an enhanced Failure Mode Cause & Effect Analysis (FMCEA) to determine the individual risk of each component. In workshops the failure possibilities of each component will be acquired, the consequences of the failure will be evaluated and the monetary risk will be calculated. The outcome of this risk analysis is the systematic identification of the critical components of equipment and the

definition of risk-reducing tasks for each component. Result of this step is the identification of highly critical and condition monitoring relevant components and the complete maintenance plan for the machine. It is the premise for the focused and cost effective development of deterioration models. In the following steps the measurement parameters and suitable techniques have to be selected. Then suitable sensors and data acquisition systems will be specified and installed, capable to detect the identified failures of the components. During the measurement a dataanalysis and processing is performed. It extracts features relevant for deterioration, degradation or process failures and ensures the quality of the measurement samples. After that, the deterioration and failure prediction models are configured in the case a deterioration or failure prediction model already exists. In all other cases a new deterioration or failure prediction model has to be developed. Then set up deterioration and failure prediction models delivering the current deterioration level or the residual useful lifetime (RUL) of the monitored component. In parallel in the process improvement related branch, the process failure model and energy optimisation models are selected and configured. In the case no suitable models exist, new models have to be developed. Results are machine process parameter recommendations for optimising process failures and energy consumption. These improvements have to be seen in context with the direct maintenance related availability improvements. The last step of the SUPREME Method is the maintenance related availability improvement of the production system. Based on the current deterioration level or the RUL, the goal of this step is to optimize the maintenance intervals and work content, considering the current situation in the factory. As a consequence the maintenance policy of a component will change several times during its lifetime. Furthermore, the upcoming maintenance tasks and production orders for the next weeks are filtered and simulated in a comprehensive model. The result is an improved maintenance work order and production order plan. Due to the complexity of this multi criteria optimisation task numerical production simulations (e. g. discrete event simulation, Mixed-Integer Linear Problem (MILP) Solver) are applied. Parallel to this activity maintenance key performance indicators are calculated.

2014

MAINTENANCE

3.2 The SUPREME Structure Model The SUPREME Structure Model shows the functions required for the operation of a SUPREME System, its relations and its application focus within the factory (Picture 3). Therefore, the following factory levels will be defined: • The production system (also manufacture and assembly system) level comprises the integrated systems of the production composed of material processing systems (machines), material handling systems and information systems. The functions operating on this level focus on several machines up to the whole production. • The machine level spans a single machine. Today machines are often mechatronic systems, which also include information systems to control the mechanics and exchange data with other information systems. In the context of maintenance a machine is categorized into a production machine and auxiliary machine. A production machine is defined as a machine which produces directly a product. In difference to that, an auxiliary machine provides for one or several production machine(s) a function required for the production machine to work. Functions operating on this level process data only from one and for one machine including the related auxiliary machines such as machine control and process data. • The component level reaches from one specific component, assembly up to a functional unit or subsystem of one machine. Functions operating on this level process data only from one and for one component such as sensor measurement data. The functions are clustered in so-called functional entities. Categories of functional entities are for • acquisition of data of the physical world (e.g. vibration sensors) and digital information systems (e.g. form Enterprise Resource Planning systems), • analysis and processing (signal processing) of accruing measurement, process and machine control data, • improvement to find the best solution of a given problem from available alternatives, under consideration of defined criteria and constraints and finally • data management to manage the accruing data within the SUPRME System ranging from measurement data up to production system data.

Production System Level

Equipment and Machine Component Level Risk Assessment

Machine Process and Control Data Acquisition

Sensor Data Processing

Machine Process and Control Data Management

Sensor Data Processing

Maintenance Sensor Data Management

Component Level Online Remaining

Wired Sensor

Wireless Sensor Network

Deterioration Level

Categories of functional entities: Data Acquisition

Analyses and Processing

Improvement

Data Management

Picture 3: Cut out of the SUPREME Structure Model 3.3 The SUPREME information model An overview of the different SUPREME packages, clustering similar information object classes is given in Picture 4. The SUPREME Information Model follows the modelling principle, that static information (e. g. master data, sensor description) is modelled separately from dynamic information (e.g. sensor values).

2014

MAINTENANCE

provides classes for are “machines”, “sensors” and “facility objects”. Machines are further distinguished into production machines and auxiliary machines, according to the definition in the SUPREME Structure Model. Each of these objects comprise at least one component. The link to the factory structure is established between the machine class and the production cell class. The package “Product Structure” covers all information object classes related to products and the production structure, such as the product master data, Bill of materials (BOM), items, part assemblies and single parts. The product can be the final product for selling to customers or an intermediate product that is further processed or assembled in a downstream production process. The package “Maintenance Work Order” includes all information object classes related to maintenance tasks, which is also one of the main classes. The maintenance task class is linked to the maintenance asset class. For event and maintenance handling the package provides also classes for maintenance plans, maintenance task templates, work orders and demand messages. The package “Production Order” comprises all information object classes related to the production process for the products. It provides the classes to model production order, work plans, work tasks and the supporting classes storage task. The package “Base Classes” contains all information object classes, which are used by the classes of the other packages. Among others, it provides classes for documents, sensors, stock, location, key performance indicators (KPI), cost centres, shift models or customer master data.

4. Test Platform specifications

Picture 4: SUPREME Information model Packages The package “Factory Structure” comprises object classes related to the logical factory structure. It provides the classes “site”, “production segment”, “production system” and “production cell”. The package “Maintenance Assets” includes all information object classes related to real physical objects which are subject to maintenance activities. The main objects for which the package

The tools which will be developed in the SUPREME project are generic to be applied to any industrial field, to both improve the maintenance and decrease the energy consumption. In order to demonstrate the potential of the new tools, the method will be implemented on a test platform, a coated paper mill, namely Condat, in France (Picture 5). Orloga and Condat worked together with all the partners to identify the pertinent application case and the critical equipment to focus on. For the energy efficiency purpose a set of interesting process parameters have been selected from available control data. This database will be analysed to perform operation optimization under normal and abnormal working conditions, while developing new tools for this function.

2014

MAINTENANCE

components like rotating machinery, bearings or gears. But correct failure prediction is not yet a reality. New techniques have to be developed to be able to take into account varying operating conditions (e.g. speed, load) and the noisy industrial environment, and also to enhance the capacities for automatic analysis of the signal.

Picture 5: General View of the Paper Machine The available database will also be used for failure prediction, based on data fusion and feature extraction and matching. A detailed analysis of mechanical failures that could happen on the critical components has been done (bearings, gearbox, bending of the shaft, misalignment…).Then a complete instrumentation has been set to be able to monitor on-line the evolution of all these failures. This instrumentation includes vibration and acoustic emission sensors, torque meters, speed sensors, as well as electrical parameters measured on the drives (Picture 6). In addition several process parameters are acquired simultaneously. A first acquisition system was installed at a very early stage in the project to start building a database of all these signals recorded for different operating conditions and during any problem / failure which would happen in production.

So this industrial database will be used by the different partners for their research work: - to develop new signal processing tools to automatically analyse the vibration signals and correlate with the kinematic of the machine and also to take into account the varying conditions (through a cyclo-stationary approach), - to set-up a condition monitoring (CM) method based on motor current analysis and acoustic emission (AE), - to compare the capacities of the different signals for CM and merge the information coming from different sensors, - to develop a data mining based approach for failure prediction, - to develop new deterioration models for residual life prediction. The state of the art on all these methods has been studied and presented in a deliverable (SUPREME, 2013). So as a summary, this database will enable to build the most pertinent feature for deterioration and failure prediction. These results will be the input to dynamically adapt the maintenance strategy and plan. In parallel an ECMS is being developed which will embed and integrate the new signal processing tools issued from the project.

5. Intelligent Control and Energy Optimization An Artificial Intelligence based module is being developed as well. Data Mining is being performed on failure modes starting from historical data, which will be exploited for failure prediction. Energy optimization is being carried out by means of enhanced operation of the plant which considers both process state data and process condition data in Real Time. The advantage of this approach is on enabling to save energy both in normal and failure conditions and on anticipating to failures before they happen. Picture 6: Ex of Vibration & Temperature sensors

6. E-learning tools

On the market today, a large number of condition monitoring systems, mainly based on vibration analysis, are available focusing on specific

As a dissemination tool, five e-learning modules will be developed in the project. Four of them will address the different technical backgrounds involved in SUPREME. These modules are

2014

MAINTENANCE

intended to be understandable by a large number of people. The last module will be specific to the project to present in detail the SUPREME methods, the tools developed, the set up in the industry and the results in the application case in Condat. The four first modules will last about 20 minutes and the specific module about SUPREME will last several hours. The modules are structured in different themes (Picture 7):

Picture 8 : A View of the first E-learning module

7. Conclusions

Picture 7 : E-learning modules The first E-learning module has been finalized (Picture 8). It deals with Predictive Maintenance, introducing the concept of “Predictive Maintenance” compared to other types of maintenance strategies and showing the strength of this approach. It is divided in four parts: - definitions (following the standard EN 13306) to get an overall view of all different types of maintenance. - an economical approach of this type of maintenance. This part helps the reader to understand what type of maintenance should be used, how to elaborate a maintenance budget, the impact on the maintenance cost and the losses. - a description of the different methods and tools. - a quiz to evaluate the understanding of the module. A link to get access to the e-learning module developed so far is available on the SUPREME website (www.supreme-fof.eu). The four other E-learning modules are currently being developed. The second module about maintenance tools should be published on our website during 2014.

This paper presented the partners, the objectives, the challenges and the first outputs of the SUPREME project. During the first year, several results have already been achieved, through the development of the SUPREME Reference Model or the specification of the test plaform in the paper industry and its instrumentation. The project will go on for two more years and will now focus on developing and testing three modules that will interact closely (ECMS, maintenance and reliability module, intelligent control and data mining module), and that will integrate all the new alogrithms and methods issued from the project.

8. Acknowledgements The authors thank all the partners of the project who contribute to this research work. The SUPREME project (“SUstainable PREdictive Maintenance for manufacturing Equipment”) is funded by the European Commission in FP7 Programme, under the Factories of the Future PPP.

9. References EN 13306 (2010) Maintenance – Maintenance terminology SUPREME Project (2013): D2.12 “State of the Art and beyond the State of the Art”.