APPLICATION OF KNOWLEDGE-BASED FUZZY INFERENCE SYSTEM ON HIGH VOLTAGE TRANSMISSION LINE MAINTENANCE

A Thesis Submitted for the Degree of Master of Engineering

By Mohd Junaizee Mohd Noor, B.Sc. Elect. Eng. (Missouri)

School of Electrical & Electronic Systems Engineering Queensland University of Technology

2004

1

KEYWORDS

High voltage transmission lines; insulators; tower structures; foundations; conductors; maintenance management; visual inspection; artificial intelligence; fuzzy logic; fuzzy inference systems; knowledge-based systems

2

ABSTRACT

A majority of utilities conduct maintenance of transmission line components based on the results of routine visual inspection. The inspection is normally done by inspectors who detect defects by visually checking transmission line components either from the air (in helicopters), from the ground (by using high-powered binoculars) or from the top of the structure (by climbing the structure). The main problems with visual inspection of transmission lines are that the determination of the defects varies depending on the inspectors’ knowledge and experience and that the defects are often reported qualitatively using vague and linguistic terms such as “medium crack”, “heavy rust”, “small deflection”. As a result of these drawbacks, there is a large variance and inconsistency in defect reporting (which, in time, makes it difficult for the utility to monitor the condition of the components) leading to ineffective or wrong maintenance decisions. The use of inspection guides has not been able to fully address these uncertainties. This thesis reports on the application of a visual inspection methodology that is aimed at addressing the above-mentioned problems. A knowledge-based Fuzzy Inference System (FIS) is designed using Matlab’s Fuzzy Logic Toolbox as part of the methodology and its application is demonstrated on utility visual inspection practice of porcelain cap and pin insulators. The FIS consists of expert-specified input membership functions (representing various insulator defect levels), output membership functions (indicating the overall conditions of the insulator) and IF-THEN rules. Consistency in the inspection results is achieved because the condition of the insulator is inferred using the same knowledge-base in the FIS rather than by individual inspectors. The output of the FIS is also used in a mathematical model that is developed to suggest appropriate component replacement date. It is hoped that the methodology that is introduced in this research will help utilities achieve better maintenance management of transmission line assets.

3

CONTENTS Keywords ....................................................................................................................................................... i Abstract ......................................................................................................................................................... ii Contents ....................................................................................................................................................... iii List of Figures ............................................................................................................................................. vi List of Tables............................................................................................................................................... ix List of Abbreviations .................................................................................................................................. x Statement of Original Authorship.......................................................................................................... xii Acknowledgments ....................................................................................................................................xiii Chapter 1: Introduction ....................................................................................................... 1 1.1 Justification for and Introduction to the Research ......................................................................... 1 1.2 Aims and Objectives of the Research ............................................................................................... 4 1.2 Organization of the Thesis.................................................................................................................. 4 Chapter 2: Components of Transmission Lines and their Failure Modes ...................... 7 2.1 Chapter Overview................................................................................................................................. 7 2.2 Towers and Structures ......................................................................................................................... 7 2.2.1 Functions of Transmission Towers ....................................................................................................8 2.2.2 Failure Modes of Steel Transmission Towers ................................................................................ 12

2.3 Foundations .........................................................................................................................................13 2.3.1 Functions of Foundations .................................................................................................................. 13 2.3.2 Failure Modes of Foundations .......................................................................................................... 16

2.4 Conductors and Earth Wires ............................................................................................................18 2.4.1 Function of Conductors and Earth Wires ...................................................................................... 18 2.4.2 Failure Modes of Conductors............................................................................................................ 21 2.4.2.1 Conductor Corrosion .......................................................................................................... 22 2.4.2.2 Conductor Vibration ........................................................................................................... 24

4

2.4.2.3 Conductor Annealing .......................................................................................................... 27

2.5 Insulators ..............................................................................................................................................27 2.5.1 Functions of Insulators ....................................................................................................................... 27 2.5.2 Failure Modes of Insulators ............................................................................................................... 31 2.5.2.1 Mechanical Failures .............................................................................................................. 31 2.5.2.2 Electrical Failures ................................................................................................................. 34 2.5.2.3 Audible Noise and Radio Interference ............................................................................ 38

2.6 Chapter Summary ...............................................................................................................................39

Chapter 3: Inspection, Diagnosis and Maintenance of Transmission Line Components ...................................................................................................................... 41 3.1 Chapter Overview...............................................................................................................................41 3.2 Review of Component Diagnosis Methods ..................................................................................41 3.2.1 Test for Tower Structural Strength .................................................................................................. 42 3.2.2 Diagnostic Tests on Tower Foundations ....................................................................................... 43 3.2.3 Diagnostic Tests on Conductors ...................................................................................................... 45 3.2.4 Insulator Diagnostic Tests ................................................................................................................. 46

3.3 Review of Inspection and Maintenance Methods ........................................................................53 3.3.1 McMahon Survey of Inspection Practice of Australian and New Zealand Utilities .............. 54 3.3.2 CIGRE Survey of Utility Assessment of Existing Transmission Lines ................................... 55

3.4 Visual Inspection ................................................................................................................................56 3.4.1 Ground-level and Climbing Inspection........................................................................................... 56 3.4.2 Aerial Inspection .................................................................................................................................. 58 3.4.3 Drawbacks of Visual Inspection ....................................................................................................... 60

3.5 Chapter Summary ...............................................................................................................................61

Chapter 4: Fuzzy Logic and Fuzzy Inference System .....................................................63 4.1 Chapter Overview...............................................................................................................................63 4.2 Fuzzy Logic ..........................................................................................................................................63 4.2.1 Membership Functions ....................................................................................................................... 64 4.2.2 Fuzzy IF-THEN Rules ....................................................................................................................... 67

4.3 Fuzzy Inference System .....................................................................................................................68 4.3.1 Inference and Defuzzification Techniques..................................................................................... 69

5

4.3.2 Developing Fuzzy Inference Systems.............................................................................................. 72

4.4 Applications of Fuzzy Logic in Utility Environment ..................................................................73 4.5 Chapter Summary ...............................................................................................................................75

Chapter 5: Using Fuzzy Logic for Transmission Line Defect Assessment....................77 5.1 Chapter Overview...............................................................................................................................77 5.2 Application of the FIS on Porcelain Cap and Pin Insulators .....................................................78 5.2.1 Factors Affecting Pin Corrosion....................................................................................................... 79 5.2.2 Factors Affecting Chipped/Broken Porcelain Insulator Disc.................................................... 83 5.2.3 The Insulator Visual Inspection Process ........................................................................................ 89 5.2.4 The Fuzzy Inference System for Visual Inspection of Insulators ............................................. 91 5.2.4.1 Assessment of Single Insulators ........................................................................................ 91 5.2.4.2 Assessment of Multiple Insulators in a String .............................................................. 112 5.2.4.3 Assessment of Multiple Insulator Strings in a Transmission Line ........................... 114

5.3 Application of the FIS on Other Transmission Line Components ....................................... 118 5.4 Potential Savings due to Introduction of FIS in Tenaga Nasional Berhad’s Transmission Line Inspection and Maintenance Practice .............................................................. 120 5.5 Chapter Summary ............................................................................................................................ 122

Chapter 6: Conclusions and Future Work ...................................................................... 124 6.1 Summary of the Research .............................................................................................................. 124 6.2 Major Research Contributions ...................................................................................................... 129 6.3 Future Work ..................................................................................................................................... 130 References ....................................................................................................................... 131 Appendix.......................................................................................................................... 138

6

LIST OF FIGURES Figure 2-1: Example of a typical overhead transmission line.............................................................. 8 Figure 2-2: Monopole and steel lattice tower designs for 220 kV transmission lines ..................11 Figure 2-3: Structural failures of a transmission line due to an ice storm in Canada....................12 Figure 2-4: Uplift and compression forces on tower foundations...................................................14 Figure 2-5: Typical construction of transmission tower steel grillage foundation ........................15 Figure 2-6: Typical construction of transmission tower rock anchor foundation ........................16 Figure 2-7: Typical ACSR conductor stranding arrangements .........................................................18 Figure 2-8: Arrow showing evidence of rust on steel core of ACSR conductor ..........................22 Figure 2-9: Sectional view of ACSR conductor illustrating galvanic corrosion mechanism .......23 Figure 2-10: Arrow showing signs of corrosion at a conductor mid span joint............................24 Figure 2-11: Multiple fatigue cracks of outermost aluminum strands of an ACSR conductor ....................................................................................................................................................25 Figure 2-12: Cross-sectional view of a cap and pin insulator............................................................29 Figure 2-13: Porcelain cap and pin insulator string.............................................................................29 Figure 2-14: Typical polymeric insulator construction.......................................................................30 Figure 2-15: Pin corrosion is most severe at cement interface .........................................................32 Figure 2-16: Brittle fractures on composite insulators .......................................................................33 Figure 2-17: ‘Doughnut’-like separation of the porcelain shell from its cap and pin ...................36 Figure 3-1: Sequence photograph of a transmission tower collapse during a structural insitu test.........................................................................................................................................................43 Figure 3-2: Diagram of the half-cell measurement method ..............................................................44 Figure 3-3: Electric field distribution across an insulator string with a punctured unit in the middle ..........................................................................................................................................................49 Figure 3-4: Variation of electric field along an 18-unit insulator string which shows defective insulator units at insulator number 7, 11, 14 and 15 .........................................................50 Figure 3-5: Aerial visual inspection of transmission line components using the helicopter .......59 Figure 3-6: Broken conductor strands due to gunshot ......................................................................59 Figure 4-1: Triangular membership function (10, 20, 30)..................................................................65 Figure 4-2: Intersection of 2 membership functions (A AND B) ...................................................66 Figure 4-3: Union of 2 membership functions (A OR B) .................................................................67 Figure 4-4: Components of a fuzzy inference system ........................................................................68

7

Figure 4-5: Mamdani fuzzy inference method .....................................................................................70 Figure 4-6: Defuzzification schemes to derive a crisp output ..........................................................72 Figure 5-1: Voltage profile across a 132 kV insulator string .............................................................80 Figure 5-2: No rust insulator pin condition..........................................................................................81 Figure 5-3: Light rust insulator pin condition ......................................................................................82 Figure 5-4: Medium rust insulator pin condition ................................................................................82 Figure 5-5: Heavy rust insulator pin condition ....................................................................................82 Figure 5-6: P-F curve for rust condition of insulator pin ..................................................................83 Figure 5-7: Cutaway drawing of a normal type cap and pin insulator .............................................84 Figure 5-8: Cutaway drawing of an anti-fog type cap and pin insulator .........................................85 Figure 5-9: Chipped porcelain disc ........................................................................................................86 Figure 5-10: Small breakage of porcelain disc ......................................................................................86 Figure 5-11: Major radial breakage of porcelain disc ..........................................................................87 Figure 5-12: Total porcelain disc breakage ...........................................................................................87 Figure 5-13: Arrows show partially broken porcelain discs in an insulator string ........................87 Figure 5-14: Arrows show 2 totally broken porcelain discs in an insulator string ........................88 Figure 5-15: Structure of insulator inspection FIS ..............................................................................92 Figure 5-16: Input membership functions for pin rust conditions ..................................................93 Figure 5-17: Input membership functions for porcelain shell conditions ......................................94 Figure 5-18: Output membership functions for insulator condition...............................................94 Figure 5-19: A 3-dimensional plot of insulator inspection FIS showing the relationship between the two inputs and output .......................................................................................................98 Figure 5-20: Insulator Sample 1 ..............................................................................................................99 Figure 5-21: Pin rust condition for insulator Sample 1 ................................................................... 100 Figure 5-22: Porcelain shell condition for insulator Sample 1 ....................................................... 100 Figure 5-23: Resultant of invoking Rule 6 ......................................................................................... 101 Figure 5-24: Resultant of invoking Rule 11 ....................................................................................... 101 Figure 5-25: Aggregation of Rules 6 and 11 resultants ................................................................... 102 Figure 5-26: Insulator Sample 2 ........................................................................................................... 103 Figure 5-27: Porcelain shell condition for insulator Sample 2 ....................................................... 104 Figure 5-28: Pin rust condition for insulator Sample 2 ................................................................... 104 Figure 5-29: Resultant of firing Rule 2 ............................................................................................... 105 Figure 5-30: Resultant of invoking Rule 3 ......................................................................................... 106 Figure 5-31: Aggregation of Rules 2 and 3 resultants...................................................................... 106

8

Figure 5-32: Insulator Sample 3 ........................................................................................................... 107 Figure 5-33: Porcelain shell condition for insulator sample 3........................................................ 107 Figure 5-34: Pin rust condition for insulator sample 3.................................................................... 108 Figure 5-35: Resultant of invoking Rule 12 ....................................................................................... 109 Figure 5-36: Resultant of invoking Rule 17 ....................................................................................... 109 Figure 5-37: Aggregation of Rule 12 and Rule 17 resultants.......................................................... 110 Figure 5-38: Proposed structure of tower inspection FIS .............................................................. 119 Figure 5-39: TNB transmission line forced outages 1997-2003 .................................................... 121

9

LIST OF TABLES Table 2-1: Basic transmission structure types ........................................................................................ 9 Table 2-2: Requirement for zinc coating thickness as per BS EN ISO 1461 ................................11 Table 2-3: AN and RI limits for insulator at typical system voltages ..............................................39 Table 3-1: ASTM C867:1999 criteria for corrosion of steel in concrete.........................................44 Table 3-2: Emerging and available aerial inspection technologies ...................................................60 Table 4-1: Determination of membership function from α-cut sets ...............................................66 Table 5-1: Category of in-service porcelain insulator disc defects ...................................................88 Table 5-2: Pin corrosion description used by Powerlink during visual inspection .......................90 Table 5-3: IF-THEN rules used in the insulator inspection FIS......................................................97 Table 5-4: Suggested insulator maintenance decision ........................................................................98 Table 5-5: Introduction of environmental coding for areas based on IEC’s pollution severity classification .............................................................................................................................. 115 Table 5-6: Environmental coding for zinc loss ................................................................................ 116 Table 6-1: Summary of transmission line components, their functions and failure modes ..... 125 Table 6-2: Summary of TNB and Powerlink porcelain cap and pin insulator inspection practice ..................................................................................................................................................... 127

10

LIST OF ABBREVIATIONS

AAAC

All Aluminum Alloy Conductor

AC

Alternating Current

ACSR

Aluminum Conductor Steel Reinforced

ACSS

Aluminum Conductor Steel Supported

AGNIR

Advisory Group on Non-Ionizing Radiation

AN

Audible Noise

ANSI

American National Standards Institute

ASTM

American Society for Testing and Materials

BS

British Standard

CIGRE

International Council on Large Electric Systems

DC

Direct Current

EMF

Electromagnetic Field

EPDM

Ethylene Propylene Diane Monomer

EPR

Ethylene Propylene Rubber

EPRI

Electric Power Research Institute

ESDD

Equivalent Salt Deposit Density

FIS

Fuzzy Inference System

FRP

Fiber Reinforced Plastic

GTACSR

Gap Built-in Heat Resistant Aluminum Alloy Conductor

IEC

International Electrotechnical Commission

IEEE

Institution of Electrical and Electronic Engineers

ISO

International Standard Organization

NRPB

National Radiological Protection Board

OPGW

Optical Fiber Ground Wire

RBS

Rated Breaking Strength

RCM

Reliability-centered Maintenance

RF

Radio Frequency

RIV

Radio Interference Voltage

ROW

Right-of-Way

RTV

Room Temperature Vulcanization

11

SiR

Silicon Rubber

TNB

Tenaga Nasional Berhad

UV

Ultra Violet

ZTACIR

Heat Resistant Aluminum Alloy Conductor Invar Reinforced

12

STATEMENT OF ORIGINAL AUTHORSHIP

“The work contained in this thesis has not been previously submitted for a master degree at any other tertiary educational institution. To the best of my knowledge and belief, this thesis contains no material previously published or written by another person, except where due reference is made.”

Signed: Mohd Junaizee Mohd Noor (Author) Date:

13

ACKNOWLEDGEMENTS In the name of God ALLAH, the Most Gracious the Most Merciful First and foremost I would like to start by acknowledging, praising and thanking my Lord, Allah, the Glorified and the Creator of all things, for blessing me with the good health and wellbeing throughout my one-and-a-half year of studies and for making it a reality for me to complete this thesis. I would like to thank my academic supervisor, Associate Professor David Birtwhistle, for his guidance and support during all the phases of this research degree. Throughout the crests and troughs of the study, he was always there with invaluable technical advice and assistance. I would also like to state my sincere appreciation to my associate supervisor, Dr. Stewart C. Bell of Powerlink Queensland, for sharing his expertise and thoughts as well as acting as a sounding board for ideas, ensuring I knew exactly what I was talking about. In addition, I would like to thank Ian Nichols and Maurice Donnelly from Powerlink Queensland and Azmi Abdullah and Rafida Othman from TNB Malaysia for their assistance during the data gathering phase of this study. Thanks also to TNB Malaysia for providing me with the financial support and opportunity to further my studies in this beautiful country of Australia. Last and by no means the least, I would like to express my special gratitude to my wonderful wife and colleague Dr. Farah Inaz Syed Abdullah, my lovely daughter Esmeralda Noor, and my immediate family for their patience, compassion and encouragements while pursuing this study. Hopefully they will get to see more of me, now “it” is finished.

14

CHAPTER 1: INTRODUCTION

1.1 Justification for and Introduction to the Research High voltage transmission lines play a very important role in a power system. Apart from their primary function of transferring power at high voltages from generating stations to load centers, through various interconnections in the network, they provide the means for effective, safe, economic and reliable operation of a power system. A transmission line system consists of the following major components: 

Tower or structure



Foundations



Conductors and earth wires



Insulators

Each of the components has its own functions in order for the transmission line to operate. Failure of any of these components may render the transmission line inoperative. Transmission line failures are highly undesirable in a power system because such failures usually result in power interruption of a large area. In order to ensure the healthy operation of transmission lines, power utilities conduct periodic inspection and diagnosis of transmission line components. The main objective of conducting inspection and diagnosis is to locate component defects so that appropriate maintenance actions can be taken before the defects develop into catastrophic failures. During the literature review, it was found that, apart from manual visual inspection, there are other various methods of transmission line component inspection and diagnosis that are currently available. Depending on the component, some of these methods utilize equipments that detect characteristic parameters that are both electrical in nature, such as voltage, current, corona and magnetic fields, and non-electrical in nature, such as vibration and temperature. These equipments, however, were found to be either expensive, still in development stages or highly affected by environmental factors during field operation.

15

However, in a recent survey on 90 utilities around the world conducted by CIGRE [1] and a survey on Australian and New Zealand utilities conducted by McMahon [2], it was discovered that the most common utility practice for locating component defects on transmission lines is by visual inspection. Depending on which part of the tower that needs to be assessed, visual inspection is conducted by line inspectors either from the ground, from the top of the structure (by climbing the structure) or from the air (by observing from fixed wing aircrafts or helicopters) [3]. Field experience indicates that there are drawbacks associated with visual assessment of transmission line defects. These include: 

Defect levels are expressed qualitatively using vague and linguistic terms such as “medium crack”, “heavy rust”, “small deflection” and “large breakage”. These statements vary from one inspector to another depending on the inspector’s inherent knowledge, reasoning, experience, health and fitness levels, the environment wherefrom the inspection is made and whether visual aids are used.



The assessment of the defects is based solely on the inspector’s personal judgment making the inspection results highly subjective. Many a time the maintenance engineer requires a second inspection by a more experienced inspector to confirm the extent of the defect.



As a result, there is a large variance and inconsistency in defect level reporting which, in time, makes it difficult to monitor the condition of the components.



This leads to the utility making ineffective or wrong maintenance decisions and inefficient control of maintenance and repair expenditures.

The aim of this research is to develop a decision support tool that will primarily address the problem of inherent human bias and subjective judgment that prevail during transmission line inspection and assessment. The hypothesis that provides the motivation for the study is that if such a tool is able to reduce or remove the uncertainty associated to inspector visual assessment during field inspection, then the information gathered from field inspection can be used effectively by the utility to manage its maintenance actions. The tool is in the form of a standardized inference system utilizing fuzzy logic. Fuzzy logic, as introduced by Zadeh in 1965 [4], is a form of artificial intelligence technique that

16

was specifically developed to handle the intrinsically fuzzy human thinking, reasoning, cognition and perception process. This thesis presents the development, application and results of a knowledge-based fuzzy inference system (FIS) on utility inspection practice of assessing one of the most important transmission line components, porcelain cap and pin insulators. The FIS is designed using Matlab’s proprietary Fuzzy Logic module. Data regarding insulator inspection practice and sample defective insulators on which the system is tested are obtained from TNB Malaysia and Powerlink, the company responsible for the operation and maintenance of transmission network in Queensland, Australia. The results of applying the FIS show that not only can it reduce the subjectivity associated with visual inspection of insulators but it can also be used to assist engineers in making strategic decisions such as ‘replace immediately’, ‘flag for next maintenance cycle’ or ‘do nothing’. The success of applying the FIS on insulator visual inspection provides the impetus for applying the same design principle of the FIS on other components of the transmission line which are also normally subjected to visual inspection. The major contribution of the research is the presentation of a methodology that improves the visual inspection of transmission line components. The methodology uses a knowledge–based decision support tool. Such a tool can be programmed into mobile electronic devices such as laptop computers or handheld personal digital assistants (PDAs) for field use. The novel methodology can facilitate the visual inspection process by: 

Making available expert knowledge in the field



Reducing the inherent uncertainty faced by field inspectors



Providing a facility to store defect data in a computerized system for defect analysis and condition monitoring



Enabling effective use of field data for use with modern maintenance policies such as Condition-based Maintenance or Reliability-centered Maintenance

Another contribution of the research is the development of a mathematical model that utilizes the output of the inspection program to plan for bulk maintenance of transmission line components. To the best of the author’s knowledge, such a methodology has not

17

been investigated before and it is hoped that the research will provide utilities with a better method of maintaining its transmission line assets.

1.2 Aims and Objectives of the Research The research was conducted to achieve the following objectives: 

To identify the main components of transmission lines, their various designs, and their functions



To understand their failure modes that lead to loss of functions including factors that affect their operation



To review the inspection and maintenance practices of utilities around the world



To identify the problems associated with utility practice of manual visual inspection



To study the principles of fuzzy logic and fuzzy inference system



To design, apply and test a knowledge-based fuzzy inference system applied on the visual inspection of porcelain cap and pin insulators



To investigate the use of the system as a decision-support tool for maintenance management of transmission lines

1.3 Organization of the Thesis Chapter two provides an introductory account of the components of transmission lines, their functions and failure modes. Factors that can affect the operation of transmission lines in a power system are discussed. Mitigation steps that are currently taken by utilities to address these problems are also deliberated. Once the failure mechanisms are identified, it is necessary to appreciate the available methods of inspection and maintenance as practiced by the utilities. This is presented in the following chapter. Chapter three is divided into two parts. The first part of the chapter looks into the available methods of inspection and diagnosis of transmission line components that are reported in various literatures. The advantages and disadvantages of these methods, some of which utilize equipments that detect parameters such as voltage, current, noise, vibration and temperature, are discussed. The second part of the chapter reviews the

18

inspection practices and maintenance strategies as conducted by utilities worldwide. The results of recent industry surveys conducted by CIGRE and McMahon, which indicate that worldwide utility practice of locating transmission line defects is by visual inspection, are elaborated on. The chapter finally discusses the difficulties faced by the inspectors when making visual inspection and the problems faced by utilities that use visual inspection to manage its maintenance actions. This leads to the need for a decision support tool that can be used to address the uncertainties that are associated with visual inspection. This type of uncertainty can be best handled by the artificial intelligence technique of fuzzy logic. The next chapter provides background information on fuzzy logic which will be used later in the proposed decision support tool to assist visual inspection of transmission line components. The main elements of a FIS, namely membership functions, fuzzy IFTHEN rules, and defuzzification methods, are described here. Several example applications of fuzzy logic in the utility environment that are available in the literature are also discussed and commented on. Having understood the principles of fuzzy logic and its applications, chapter five details primarily the design, development and application of a knowledge-based FIS applied on utility visual inspection practice to locate defects on porcelain cap and pin insulators. A closer look at two types of porcelain insulator failure mechanisms – corrosion of the pin and breakage of porcelain shells – are firstly presented. Current insulator visual inspection practices of the two utilities (TNB and Powerlink) to locate the above defects are then discussed, highlighting the associated problems. The chapter then proceeds to explain the design of the insulator inspection FIS and its simulated application on actual in-service insulators taken from both the utilities. The results of this application are then discussed, indicating how the FIS assists in achieving consistency in assessing insulator defects. It is shown next how the output of the FIS can be used to assess multiple insulators in a string and also multiple insulator strings in a transmission line. It is also shown how the output of the FIS can be further used in a proposed mathematical model to estimate the appropriate date for bulk replacements of insulators. The chapter finally discusses the application of the knowledge-based FIS on the visual assessment exercise of other components of the transmission line.

19

Lastly, chapter six details the conclusions derived from the research contained in this thesis. Future work, which the author feels may prove particularly beneficial, is also elaborated on.

20

CHAPTER 2: COMPONENTS OF TRANSMISSION LINES AND THEIR FAILURE MODES

2.1 Chapter Overview This chapter presents an overview of the functions of transmission line components and their failure modes relevant to the research described later in this thesis. The major components of high voltage transmission lines as covered in this chapter are as follows: 

Towers and structures



Foundations



Conductors and earth wire (including joints)



Insulators

Each of these components has its own functions that enable power to be transmitted safely and reliably through transmission lines. Failure of a transmission line component as outlined in this thesis is defined by the loss of its functions. This chapter discusses the design of the components, their features and, most importantly, how they fail in operation. It is highlighted in this chapter that functional failures of transmission line components are generally gradual and can be of electrical, structural or mechanical in nature. It is also highlighted that these degradations are primarily influenced by environmental factors such as local climate, weather, elevation and ambience wherein the transmission line is built. Mitigation steps that utilities take to address these problems are also deliberated in this chapter. 2.2 Towers and structures Figure 2-1 shows a typical single circuit 765 kV extra high voltage transmission line and its components commonly used in North America. Notice that the transmission line has two earth wires (denoted by “shield conductors” in the figure) strung at the top of the structure and the bundle conductors are held by v-string insulators. 21

Figure 2-1: Example of a typical overhead transmission line (Source: [5])

2.2.1 Functions of transmission towers The main purpose of the transmission tower is to carry the overhead transmission line conductors and earth wires above the ground. In fulfilling this role, it has to: 

withstand all the variety of forces that it is exposed to with regards to the environment it is located. These forces include normal still air loads, extreme wind loads, ice loads, loads during erection and maintenance, and the changing of conductor sag when the conductor expands and contracts with normal daily current loads [5]. In certain instances, tower designs must withstand loads imposed by extreme conditions such as in earthquake, cyclone and tornado areas [6].



maintain electrical clearances between live conductors and any earthed body in the vicinity of the tower such that the energized lines do not induce any hazardous voltage that could render the operation of the transmission lines dangerous to the environment and the public. 22



provide a path to earth fault and/or lightning current so that any danger to the environment as a result of the two occurrences is reduced. Hence, the tower must also exhibit low resistance to ground during transient lightning over voltages.

Depending on network requirements of a power system, the tower may be designed to cater for a single 3-phase circuit, double 3-phase circuits, multiple voltage circuits, and, with direct current transmission, either monopolar or bipolar construction. In certain countries, due to land constraints, new transmission lines are always built on double circuit towers and old single circuit lines are upgraded to double circuits to optimize the use of land easements. There are basically six different types of transmission line structures in common use worldwide with many design variants based upon them. Table 2-1 lists the basic structure types [7]: Structure Type

Description

Self-supporting

Most common design style, easiest to work on but

lattice steel tower

makes the largest visual impact. Relative medium cost.

Lattice guyed steel Simple foundation but requires large easement width to tower

accommodate guys. Lowest capital cost.

Steel/concrete

Slim appearance, low maintenance for galvanized steel

monopole

and concrete. Small easement width. Applicable in urban areas where easement acquisition is expensive and aesthetic visual impact is required. Highest capital cost.

Compact

Special adaptations of the foregoing types to meet

structures

specific space or environmental limitations

H-frame in either Traditional rectilinear design, best suited to single wood

pole

or circuit use. Requires wide easement.

concrete Table 2-1: Basic transmission structure types The selection of the type of structures to be used in a transmission line is mainly affected by the environment where the structures are located. These factors are [7]:

23



Pollution level: whether the route of the transmission line is going to cross highly polluted and corrosive area or environmentally sensitive country which requires specific designs or surface treatment of the structures on part or all of the route



Terrain: whether there is difficult terrain where the line trespasses which makes the structures susceptible to problems of land slides, floods or tidal inundations, high soil resistivity, ‘bottomless’ sand, exposure to wind, lightning or other natural phenomena



Aesthetic: whether the visual impact of the transmission line can be improved to make it more acceptable to the public eye by using either a compact design or a monopole



Climatic: whether the transmission line has to operate in extreme climatic conditions such as regular high winds, extremes of ambient temperature (including snow and ice) or marine salt sprays



Current load: whether the transmission line is meant to be operated during emergency loading conditions within the transmission network



Maintenance requirements: whether maintenance on the transmission line can be done live or de-energized, with climbing, using elevated platform vehicles or helicopter

Figure 2-2 shows an example of 2 types of tower structures which carry 220 kV transmission lines located side-by-side.

24

Figure 2-2: Monopole and steel lattice tower designs for 220 kV transmission lines (Source: [8]) Under normal circumstances, utilities worldwide use steel lattice transmission tower as the structure of choice due to its cheap installation cost and relatively low operational maintenance cost compared to the other types. To make the steel tower resistant to the corrosive effects of the environment, the steel lattice members are treated in a process called ‘hot-dip galvanisation’ which is to provide a layer of zinc by dipping the members into a zinc bath. The zinc coating acts as a sacrificial element to protect the steel from being oxidized or developing rust. One of the international reference standards for coating thickness is based on BS EN ISO 1461:1999 [9] which requires coating thickness of the steel members as per Table 2 below: Steel thickness

Minimum average coating

Coating thickness,

weight, G/M2

ΜM

1 mm up to 2 mm

335

47

2 mm up to 5 mm

460

65

Table 2-2: Requirement for zinc coating thickness as per BS EN ISO 1461 [9]

25

2.2.2 Failure modes of steel transmission towers Steel transmission towers are typically designed and built to last for 60-70 years of operation. They are considered to have failed when they can no longer serve their function which is primarily to withstand all the forces that they are subjected to within the environment they are located. Tower buckling and/or collapse are a sign of the tower losing its function. Tower failures can be attributed to either: 

aging



natural disasters (such as earthquakes, wind/ice/snow storms, floods, landslides and foundation erosion)



vandalism or sabotage

Failures due to natural disasters and vandalism are relatively rare and difficult to control, although the damage they cause to the transmission line can be quite extensive. Figure 2-3 shows a series of tower failures in Canada due to an ice storm.

Figure 2-3: Structural failures of a transmission line due to an ice storm in Canada (Source: [10])

26

The single factor that affects the integrity of steel transmission towers is steel corrosion [11]. Corrosion creates weak points on the tower members. When the effects of rust on the tower members reach a point whereby the members cannot withstand the mechanical strength they are designed for, the tower can buckle and, at worst, collapse. What is more serious is that the cascading effect of one tower collapse can bring down several other interconnected towers thus rendering the whole transmission line circuit unserviceable. Factors that influence the rate of corrosion include: 

Relative humidity



Atmospheric presence of corrosive agents (e.g. sulphur dioxide or nitrogen oxide)



Age of tower

Corrosion is usually more evident at the base of the transmission tower where there is contact with ground. 2.3 Foundations 2.3.1 Functions of foundations The purpose of tower foundation is to transmit to ground the mechanical loads that are due to the tower and conductors so that the loads do not cause any settlement or movement which would affect the stability of the tower [5]. The loads are the resultants of the dead weight of the tower and the wind forces that act upon the tower and conductors. These forces result in uplift and compression forces which is transmitted to the tower legs to the foundation. Therefore it is important that the foundation is designed correctly to resist these forces in order to ensure the stability of the tower structure. Figure 2-4 shows the direction of uplift and compression forces imposed on tower foundation based on wind direction acting upon the tower.

27

Wind direction

Tower

Ground line Uplift

Foundations

Compression

Figure 2-4: Uplift and compression forces on tower foundations Tower foundation design needs to consider the properties of the soil on which the tower is standing. These are determined by conducting soil investigations which are performed in situ and the soil properties are analysed in the laboratory. The soil property classifications are based on measurements of grain size, density, shear strength, compressibility, chemical composition and moisture content [12]. There are basically four types of foundation design used for tower structures: 

Steel grillage This type of foundation is used where the upper layers of soil have relatively low bearing capacities. It serves to evenly distribute the load in order to avoid bearing capacity failure and to ensure any ground settlement is evenly catered for. Steel reinforcement bars are built into the concrete cast to increase its strength. It is also commonly referred to as the ‘pad and chimney’ foundation due to the shape of its design. Figure 2-5 shows a typical steel grillage foundation:

28

Steel tower leg member Ground level

Steel grillage

Concrete chimney

Concrete pad

Figure 2-5: Typical construction of transmission tower steel grillage foundation 

Pile foundation This type of foundation is necessary for very poor soils. It transmits the load to the lower ground layers which are more stable. Several piles may be required for each tower foundation depending on the load capacity of the piles. A few types of piles are normally used which include steel sections, precast concrete sections, and steel screw anchor [12]. There is also the bored, cast in situ concrete pile in which steel casings are removed after the concrete is cured.



Strip footing [12] These are used in good denser soils with high bearing capacity so that the weight of the structure is transferred adequately to the ground. Its construction is similar to the steel grillage foundation but with a much smaller pad.



Anchor foundation In areas where there is very hard or rocky ground, it may not be economically feasible to excavate and construct concrete foundations. Instead, the stability of the rock ground is used as a base for the tower leg. This is done by grouting the steel reinforcement bars into predrilled holes in the rock. Figure 2-6 below shows a typical tower rock anchor foundation.

29

Steel tower leg member Ground level

Soil layer

Concrete cast

Steel grillage

Rock layer

Figure 2-6: Typical construction of transmission tower rock anchor foundation 2.3.2 Failure modes of foundations The foundations of a transmission tower are deemed to fail when they can no longer transfer the tower loads to ground. The effects of the foundation failure could bring the tower down as the foundations can no longer withstand the compression and/or uplift forces they are subjected to. It has been identified that there are two major failure modes of transmission tower steel grillage concrete foundations: 

failure of the foundation itself due to corrosion of steel reinforced bar



excessive earth movement due to natural disasters such as earthquakes, floods, landslides and erosions

Steel corrosion in concrete is a specialized subject in civil engineering. The mechanism of steel corrosion in concrete is generally associated with the loss of the protective action provided by the cement [13]. Concrete exhibits alkaline properties and fresh concrete mixture typically registers pH 13 [14]. Through micro cracks in the concrete, this alkaline property can turn to acidic (pH < 9) either through the action of atmospheric carbon dioxide (“carbonation”) or by penetration of aggressive ions such as chloride. In the

30

presence of oxygen and moisture, these ions react with the steel bar to form bulky rust, which can exert sufficient forces to form cracks in the concrete. This leads to further ingress of moisture and the reaction is accelerated until the concrete ultimately crumbles leading to failure of the foundation. When the foundation can no longer bear the uplift and compression forces exerted by the tower, the tower then collapses. Corrosion, fortunately, is a slow-moving process. It is important to detect signs of steel corrosion in foundation early and to take preventive measures before it reaches a state of imminent foundation failure. Foundation defects due to earth settlement can be remedied if the settlement process is gradual and defects are detected early. Remedial actions to contain earth movement such as building a retaining wall, creating proper drainage and adding vegetation to top soil would help alleviate this problem. However, foundation failures as a result of excessive earth movements due to natural disasters like floods, earthquakes or land slides would normally be catastrophic. Some examples of the extent of such occurrences in several countries are given below: 

In January 1994, an earthquake measuring 6.6 on the Richter scale hit Los Angeles, California. This earthquake resulted in damage to a section of a fourcircuit 230kV line supplying 1500MW to the city of Los Angeles. Due to this incident, electricity supply to the whole city was interrupted [15].



In 1991, Mount Pinatubo in the Philippines erupted after lying dormant for 600 years. Lahar (mixture of volcanic ash and water) flow from the mountain destroyed a section of a 220kV transmission lines that were supplying 1200MW power to the city of Manila [16].



In October 1999, a massive flood hit the state of Orissa in India [17]. The flood swept a few towers of the transmission line that was supplying power to its capital city, Bhubaneswar.

31

2.4 Conductors and earth wires 2.4.1 Functions of conductors and earth wires The main function of transmission line conductors is to carry rated current up to their design temperature within mechanical design limits. In satisfying this purpose, conductors also need to maintain even sags throughout the line route so that the clearance between the ground and the conductors are within statutory limits. There are 2 conductor types that are currently in use by transmission utilities worldwide. They are: 

ACSR conductor (Aluminium Conductor Steel Reinforced) This conductor type gets its name from its construction which is galvanized steel strands surrounded by aluminium steel strands. The steel strands provide strength whilst current is conducted through the aluminium strands. A layer of grease is normally applied on the steel strands to provide extra corrosion protection. Figure 2-7 shows typical ACSR stranding arrangements.

Figure 2-7: Typical ACSR conductor stranding arrangements (Source: [18]) 32



AAAC conductor (All Aluminium Alloy Conductor) This type of conductor is made up of strands of aluminium alloys. These alloys, often containing silicon and/or magnesium, provide high breaking strength and good electrical conductivity [5]. 2 alloys that are commonly used to manufacture AAAC conductors are alloy 1120 and alloy 6201 [19]

The ACSR conductor is very widely used because of its mechanical strength, widespread manufacturing capacity and cost effectiveness [12]. However, the AAAC conductor is increasingly being used in place of ACSR due the advantages which are outlined below [19]: 

Higher conductor and line ratings: for the same outer diameter, AAAC conductors can have from 9% to 14 % higher current ratings compared to ACSR conductors.



Higher strength-to-weight ratio: AAAC conductors have up to 35% higher strength-to-weight ratio compared to ACSR conductors. Hence, AAAC conductors provides for smaller sags and lower tower heights [12].



Less capital cost: for the same stranding configuration, AAAC conductor costs 15% lower per meter compared to ACSR .



Lower I2R losses: the lower AC resistance of AAAC gives savings in operating cost over ACSR .



Better corrosion resistance: ACSR conductors are more prone to corrosion related problems due to the presence of steel in its stranding.



Easier jointing procedure: jointing of ACSR conductors requires compression procedures on 2 sleeves whereas for AAAC only 1 jointing sleeve is needed.

The disadvantages of using AAAC conductor are that it has lower breaking load capacity and limited self-damping properties compared to ACSR. One of the challenges faced by utilities today is to transmit more power from generating stations to existing load centres in the most cost-effective way. One way to achieve this objective is to retrofit old towers with higher capacity conductors rather than constructing new transmission lines. The alternative conductor types that are available today are [20]:

33



ACSS conductor (Aluminum Conductor Steel Supported)



ZTACIR conductor (Heat Resistant Aluminum Alloy Conductor Invar Reinforced)



GTACSR conductor (Gap Built-in Heat Resistant Aluminum Alloy Conductor)

These conductors are electrically and dimensionally very similar to ACSR conductors except that they have lower heat expansion coefficient than that of ACSR conductors. This results in them being able to operate at higher temperature but with no change in sag, which in turn brings about higher current carrying capacity. When the conductor is energized at high voltages, the air surrounding the conductor becomes ionized. When this ionization process exceeds the air’s dielectric breakdown strength, electrical discharges at various intensities along the conductor occurs. This phenomenon is termed ‘corona’ [21] and can be visually characterized by a luminous glow given the right atmospheric conditions. Corona is the source of radio interference (RI) signals (unwanted electrical signals that may affect radio transmission or navigation equipment [5]) and audible noise (AN) – the crackling or buzzing sound often heard when someone stands in the vicinity of transmission lines. Energized conductors are also the source of electric and magnetic fields (EMF) under the line. There have been public concerns over long-term health effects of prolonged exposure to EMF. Many scientific studies have been done to look into the possibility of developing cancer but, to date they have not conclusively proven any causal link between EMF and any health effect. In 2001, a report [22] was released by the Advisory Group on Non-Ionizing Radiation (AGNIR) working for the National Radiological Protection Board (NRPB) in the UK. The report found some evidence to suggest that prolonged exposure to EMF at relatively high average levels (4 milliGauss (mG) magnetic field or greater) may be associated to an increased risk of childhood leukaemia. However the report also said that the evidence is “not strong enough to justify a firm conclusion that such fields cause leukaemia in children”. Furthermore, measurements near power lines have found [23] that the highest magnetic field reading is only around 2 mG.

34

Earth wires (sometimes referred to as shield conductors) are strung at the topmost part of the transmission tower. Under normal operations, they do not carry any current but under fault conditions, they must be able to withstand the fault current levels for which the system is designed. Earth wires are designed to intercept lightning strokes so as to restrict lightning surge voltage rise on phase wires and reduce the risk of insulator flashovers. Earth wires are normally smaller in size compared to phase conductors. They can be made of the same material as the phase conductors i.e. ACSR or AAAC. Optical Ground Wire (OPGW) [7] is a specially constructed earth wire with a protective tube into which optical fibre strands are layered. The optical fibre is used for communication, control and information transmission purposes. Conductors have limited length, particularly for manufacturing convenience and transport purposes. Therefore, many segments of conductors are connected to make up a transmission line. Connections between lengths of conductors are made manually at site by using a sleeve that is compressed on the conductor with a hydraulic press. The function of the joint is primarily to achieve the electrical and mechanical continuity of the conductor. Hence, the operational performance of a transmission line is affected by factors that influence both the conductor and joints. 2.4.2 Failure modes of conductors Conductor failures are highly undesirable because when conductors break, they affect system operations, supply reliability and public safety. The common failure modes for conductors are due to: 

Corrosion



Vibration fatigue



Annealing

The factors affecting the condition of conductors which can result to the failure modes above include: 

Conductor loading



Manufacturing and installation quality 35



Material stress (due to tension or vibration)



Protection (application of grease on steel core of ACSR conductors)



Weather



Levels and types of environmental pollution



Age

2.4.2.1 Conductor corrosion Shannon [24] explains that for ACSR conductors, there are two predominant mediums that result in corrosion problems: 

Industrial pollution



Salt spray from the sea

Corrosive industrial pollution is carried to the conductor in the form of rain, fog or mist. The corrosive effluence then reacts with the galvanising layer of the steel strand which is eventually depleted. When this happens, oxidation of the steel core takes place and rust forms on the steel core. The presence of rust eventually leads to the steel core losing its mechanical strength. Typically, there is little or no loss of aluminium in this process and therefore no external indication of conductor deterioration is visible until the conductor fails. The arrow in Figure 2-8 shows evidence of rust on the steel core of an ACSR conductor which can be characterised by the rough and pitting appearance.

Figure 2-8: Arrow showing evidence of rust on steel core of ACSR conductor (Source: [10])

36

Corrosion due to salt spray usually occurs where the transmission lines are close to the coastal area. In this case, corrosion of the conductors mainly occurs as a result of galvanic corrosion. Figure 2-9 shows a diagram of galvanic corrosion mechanism in an ACSR conductor:

Figure 2-9: Sectional view of ACSR conductor illustrating galvanic corrosion mechanism (Source: [24]) In Figure 2-9, salt spray from the sea combined with precipitation collected on the conductor creates an electrolytic circuit between the aluminium and the steel strands. The galvanizing layer then becomes depleted thereby creating direct contact of steel and aluminium. In the presence of the electrolyte, an electrolytic cell is then set up between the steel and aluminium strands, with the aluminium becoming the sacrificial anode. This results in loss of aluminium and creates high electrical resistance of the conductor at the affected location. Failure of the conductor may result from loss of aluminium strands which can melt because of high temperature due to increase in resistance to current flow. In this process, evidence of aluminium loss can be detected by the presence of bulky white powder (AlCOH) on the surface of the conductor. At the same time, conductors develop hotspots at locations where there is high resistance. Conductor mid span joints are given particular attention by many utilities because corrosive effluence as mentioned above tends to concentrate at the joints [24]. Marshall [25] pointed out that although joints appear homogeneous, they actually have micro-voids which collect water, corrosive contaminants and oxygen that can accelerate the corrosion process. Corrosion on joints can be of three types: 37



Pitting corrosion: localised corrosion which forms holes on the surface of the joint



Crevice corrosion: corrosion in confined crevices where acidic build up occurs



Corrosion (rusting/oxidising) of the steel sleeve and strands inside the joint

Marshall also suggested that the effect of corrosion at the joints leads to an increase in joint resistance long before they develop high temperature and eventually fail. The probability of the conductors or joints developing corrosion problems is higher at the lowest point of the span since it takes the longest for the conductor to dry out at this position and it is also easier for the corrosive effluence to collect at this point [24]. The arrow in Figure 2-10 shows signs of corrosion at a conductor mid span joint:

Figure 2-10: Arrow showing signs of corrosion at a conductor mid span joint (Source: [26]) 2.4.2.2 Conductor vibration Conductor failures can also occur due to the generation of vibration in the conductors as a result of wind action. There are three main forms of conductor vibration: 

Aeolian vibration



Galloping

38



Wake-induced sub-conductor oscillation

Aeolian vibration is generally caused by light steady wind (between 1 to 8 m/sec) blowing between 0o to 75o to the conductor axis [5] which results in the formation of vortices on the other side of the conductor. These vortices shed alternately from the conductor top and bottom and when this happens synchronously along the conductor length, it vibrates. The frequency of aeolian vibration is usually in the range of 5 to 100 Hz and with amplitudes of 0.5 to 2 conductor diameters. If uncontrolled, aeolian vibration can lead to early fatigue of the conductor by causing fretting between strands of the conductor adjacent to clamps or fittings. This can ultimately lead to fatigue crack initiation and conductor breakage. Figure 2-11 below shows multiple fatigue cracks on the outer most strands of an ACSR conductor.

Figure 2-11: Multiple fatigue cracks of outermost aluminum strands of an ACSR conductor (Source: [10]) Aeolian vibration is more likely to occur on transmission lines located in open terrains where there is no shielding from winds. There are several methods that can be used to mitigate aeolian vibration problem. The use of Stockbridge vibration dampers fixed at points where antinodes form during conductor vibration is one of them. Another method is to string the conductor at a relatively low tension as it is found [5] that if the conductor tension is maintained at less than 20% of the ultimate tensile strength, the likelihood of vibration is significantly reduced. A recent CIGRE Task Force [27] suggested a new and more precise approach of conductor tensioning based on the ratio of H (horizontal tensile load derived at average temperature of the coldest month where the line is built) and W (conductor mass per meter). The Task Force recommends values of H/W of about 1000 m in flat open grounds and up to 1425 m for areas with vegetation shielding.

39

Galloping is a low frequency phenomenon in which large amplitude mechanical oscillations occur on conductors with a frequency of between 0.15 Hz and 1 Hz. Galloping occurs when ice has formed on the conductors and where there are moderately strong crosswinds of up to 45 km/h. The power transmitted by the wind is therefore higher than for aeolian vibration. With the right angle of incidence, galloping can cause the amplitudes to reach or exceed the sag of the conductor. In severe cases, phase-tophase conductor clashes can result. Galloping is also the cause of loosening or fatigue of tower members. Galloping can be controlled by [28]: 

Using interphase ties (restricts conductor clashing)



Using conductors with smooth body profiles (increase aerodynamics)



Heating the conductors by increasing electrical loading



Increasing conductor spacing

Wake-induced sub-conductor oscillation [12] (also called sub-span oscillation) occurs only on bundled conductors and is produced by forces from the shielding effect of windward sub-conductor on the leeward sub-conductor. Any vortices that are formed by the windward conductor immediately expose the leeward conductor to a complex and variable set of forces. These forces can either suppress motion or cause conductors to move in various formations. Fortunately, problems associated with wake-induced subconductor oscillation are not widespread. However, it can cause sub-conductor clashing, fatigue of dampers or spacer dampers and damage to hardware fittings. These problems can be mitigated by [29]: 

Installing spacer dampers



Setting spacers at unequal intervals



Tilting the bundles at angles greater than 20o so that both sub-conductors are not on the same horizontal plane



Increasing sub-conductor spacing

40

2.4.2.3 Conductor annealing Annealing is caused by heating of a material and followed by a cooling period. During the annealing process, the material experiences a change in its structure and for metals, this result in a loss of tensile strength. This was proven by an experiment [30] conducted by di Troia. In the study, a length of ACSR conductor was heated up to 175 oC in 100 cycles and a series of tests conducted after the heating periods are over. The results showed that the ACSR conductor experienced a reduction of material strength of up to 90% rated breaking strength (RBS) (ANSI Class 1 standard requires withstand of up to 95% RBS). Di Troia’s experiment showed that conductor problems associated with annealing occur when the conductor is operated at beyond its design operating temperature. For ACSR conductors, the aluminium strands stretch and when the conductor gets too hot (more than 100 oC), the strands fail to retain their original shape and lose tensile strength. As a result, transmission lines sag. Excessive sag due to reduced tensile strength, together with high ice and/or wind loading events, can eventually lead to conductor tensile failure. To operate conductors at higher temperatures without risking the problems associated to annealing, special low-sag, high-temperature rated conductors should be used. One such conductor is the ACSS (Aluminium Conductor Steel Supported) conductor. For the ACSS, the aluminium strands are fully annealed at the factory and therefore will not change its shape at high temperature operations. This prohibits the conductor from further sagging and allows the conductor to be operated at high current loadings. 2.5 Insulators 2.5.1 Functions of Insulators Insulators are used to attach the conductors to the cross arms of transmission towers. They are the medium by which the live parts of the transmission line are separated from the earthed parts. In doing so, they serve three purposes: 

To provide mechanical support for the conductors under the worst loading conditions

41



To withstand electrical stresses as a result of lightning, switching or temporary over voltage surges that could initiate voltage flashovers under the worst weather and pollution situations



To operate within interference (RIV and Audible Noise) limits

Insulators are therefore made of components consisting of dielectric material that gives them the insulating properties and materials that provide mechanical strength. Insulators are mostly recognized by their dielectric material, which is either of these three [31]: 

Porcelain



Toughened glass



Polymers or composites

Porcelain and toughened glass insulators have been used for more than 100 years worldwide while polymeric insulators were introduced as an alternative in the 1960’s [32]. Porcelain insulators are made of a mixture of clay materials which forms into porcelain after being fired in a kiln. Glass insulators are usually made from soda-lime-silica glass which is formed in a furnace and made toughened by rapid cooling. Composite insulators are made from polymeric materials such as ethylene propylene dimethyl monomer (EPDM), ethylene propylene rubber (EPR) or silicone rubber (SiR). The most common type of porcelain and glass insulators is the cap and pin type where one unit is attached to another with galvanized steel fittings to form a string. Figure 2-12 shows a typical cross-sectional view of a cap and pin insulator unit and Figure 2-13 shows a complete porcelain cap and pin insulator string.

42

Figure 2-12: Cross-sectional view of a cap and pin insulator (Source: [8])

Figure 2-13: Porcelain cap and pin insulator string (Source: [8]) Polymeric insulators are always manufactured in one piece (commonly called long rod insulators) with a fibreglass reinforced core to provide mechanical strength and polymeric material cover attached to galvanized metal end fittings. Figure 2-14 shows a typical polymeric insulator construction.

43

Figure 2-14: Typical polymeric insulator construction (Source: [8]) Two measurement terms always applied to insulators are: creepage distance, which is the distance measured along insulating surfaces between the conductive parts, and arcing distance, which is the distance in air between the conductive parts that normally have operating voltage between them [33]. Selection of creepage distance on insulators is determined by environmental factors (insulators with longer creepage distance are used in highly polluted environments) whereas arcing distance is determined by line voltage and, hence, the higher the voltage, the longer the insulator string. Depending on the design of the transmission tower, insulators are strung in either horizontal, vertical or v orientation. The voltage across an insulator string is not uniform. Voltage at the line end of the string may constitute up to 15% [5] of the transmission line voltage and progressively decreases towards the earth end of the string. Therefore, the electrical stress is higher at the line end of the insulator string. On extra high voltage lines, grading rings [31] are used at the live end of the insulator to mitigate this phenomenon. Between the porcelain cap and pin insulators, utilities normally have equal preference to use on transmission lines as they tend to have the same operational quality and cost [5]. 44

However, polymeric insulators have been increasingly used by utilities due to the following advantages according to surveys conducted by Schneider et al [34], Kikuchi et al [35] and Philips [36]: 

90% weight reduction



Lower installation cost



Aesthetically more pleasing



Resistance to vandalism



Improved handling of shock loads



Good contamination performance

The disadvantages of polymeric insulators are [31]: 

More susceptible to ageing



More prone to handling and storage damage



Attract pecking birds



Contamination performance can change with time

2.5.2 Failure modes of Insulator When put in service on transmission lines, insulators are considered to have failed when they can no longer serve the purposes mentioned in section 2.1.4. Depending on factors such as temperature, humidity, contamination, rain, mechanical loading, and voltage, the failure modes can either be of mechanical or electrical in nature. 2.5.2.1 Mechanical failures For cap and pin insulators, mechanical failure can result from failure of the pins due to either corrosion or metal fatigue. The result of this phenomenon is pin fracture leading to a conductor drop. The consequences of a conductor drop can be serious as, in addition to loss of supply, if the transmission line is constructed across populated areas, damage to public properties and/or death or injury to the public may be imminent. Pin corrosion, according to tests carried out by NGK Insulators [37], is caused by electrolytic action of the steel material with environmental contaminants catalysed by 45

leakage current. Gorur [31] further explains that the small mass and slender design of the pin can lead to high levels of leakage current density on the pin surface which, in time, can cause considerable amount of corrosion. This explains why evidence of rust is normally found at the point where the pin is closest to the cement, the place where leakage current is highest, as shown in Figure 2-15:

Figure 2-15: Pin corrosion is most severe at cement interface (Source: Author’s personal collection) It can be observed from the author’s field experience in high voltage transmission line in TNB Malaysia that: 

Pin corrosion is more evident on suspension strings compared to tension strings. This is due to the washing effect of direct rainfall on to the pin on tension strings.



Corrosion rate is not constant with time



Pin corrosion is more severe on insulators situated in the middle of the string compared to the ones at the extreme ends



Rate of pin corrosion is higher in highly polluted environments such as in coastal or industrial areas

To counter the effect of corrosion, most utilities use cap and pin insulators with sacrificial zinc sleeves around the pin to inhibit the growth of corrosion a few years and thus prolonging their useful life. The zinc sleeves work as sacrificial anodes.

46

Corrosion on the cap, although noticeable, has minor effect on the mechanical integrity of the insulator due to its larger cross sectional area and lower leakage current density. The insulator pin can also be damaged by metal fatigue. Aeolian vibration on the conductor which is induced by incidental winds can be transmitted to the line hardware and the insulators. Under these conditions, the pins, being the smallest part of the insulators in a string, are subjected to high alternating mechanical stresses. This coupled with the high tensions that pins are subjected to in a string, can cause eventual pin fracture. Fractures of the cap do not usually occur because of its bigger volume and higher mass. Mechanical failures of polymeric insulators are always related to damage of the core. It can be recalled that the core of polymeric insulators is made of fibreglass reinforced plastic (FRP). Studies and experiments by many researchers have suggested that the reason for this failure is due to the chemical degradation of the FRP as a result of reaction with environmentally-induced acids [36] [38] [39] , or atmospheric humidity [40]. An IEEE Task Force was formed to study this phenomenon and found out that brittle fracture can be avoided if a good covering or seal is provided to prevent moisture from penetrating the core [41]. Figure 2-16 shows the results of a brittle fracture on 2 polymeric insulators:

Figure 2-16: Brittle fractures on composite insulators (Source: [42]) Mechanical failure of the fibre glass rod of composite insulators can be due to poor rod manufacture, mishandling during shipping, storage and installation, and operational overloading [36]. 47

2.5.2.2 Electrical failures The electrical performance of insulators on transmission lines is determined by their ability to prevent flashovers which can result in tripping of the lines. Flashovers occur when the insulating medium, which is air surrounding the insulator string, breaks down under electrical stress between the live and earthed parts of the transmission tower. Under normal operating conditions, the insulator provides sufficient air-gap distance to prevent flashover from occurring. However, several factors can effectively make this distance ‘shorter’ and initiate the flashover. For cap and pin insulators, these include: 

Loss of insulator shells due to: o Vandalism [43]: This problem is particularly troublesome for glass insulators. During production of the glass shells, the hot and molten glass mixture is forcedcooled in its mould by blowing cool air followed by dipping in cold water baths. This action results in strong compressive forces at the surface of the glass in balance with equally strong internal tension forces making the glass shell toughened. If this balance is disturbed by a large impact, the compressive forces on the surface will be released and the glass shell will shatter to pieces. To some people, this shattering is a spectacular sight and the glass insulators therefore often become targets for shooting practice. This action reduces the number of glass insulator shells on the string thereby reducing leakage distance which can lead to a line flashover. The problem is not as serious with porcelain shells as a bullet will usually chip or, sometimes, crack the shell. A small chip will not affect the insulator electrically or mechanically but a large crack can initiate shell breakages giving the same effect as glass insulators. o Self shattering of glass shells [31]:

48

Self-shattering of glass insulators is a wear-in type of failure: the failure incidents will reduce with time after installation, usually in 3 to 5 years. The glass shells can shatter by itself due to impurities and air bubbles trapped during the manufacturing process. Air bubbles result from not enough fining agents used in the glass mixture. Impurities in terms of stones and cords may also find their way into the glass mixture. Although there are precautions that can be taken during manufacturing to contain this problem, a number of imperfect insulators may be eventually used on a transmission line. The glass shells can also shatter due to erosion of the glass shells. Sodium ions (Na+) which are contained within the glass shell will migrate under DC voltage and heat. This ionic conduction may give rise to shattering of the shells, particularly in highly polluted environments. o Cement growth: Cement growth is a term given to the phenomenon when dried Portland cement, which is used to cement the porcelain shells to the cap and pin, expands with time. Gorur [31] explains that there are three mechanisms of Portland cement expansion which can occur in the cement mixture:  Delayed hydration of uncombined CaO to Ca(OH)2  Reaction of excess gypsum with tricalcium aluminate  Delayed hydration of periclase (MgO) to form brucite (Mg(OH)2) The after-effect of cement growth is cracked porcelain shells and once this happens, their usefulness as an electrical insulator is terminated. To counter the problem, mortar is normally added to the cement mixture to reduce cement expansion. o Voltage stress [31]: As mentioned in section 2.1.4, under normal conditions, the voltage distribution along a string of insulators is not uniform. It is generally higher at the line end and gets lower towards the earthed end. As a result,

49

the insulator units that are close to the line end experience higher potential stresses. At high voltages, this leads to the formation of corona around the cement near the pin and/or between the bottom of the cap and the shell. Continuous corona discharge on a porcelain insulator will result in separation of the porcelain shell from the cap and pin in the form of a ‘doughnut’. This phenomenon is shown in Figure 2-17.

Figure 2-17: ‘Doughnut’-like separation of the porcelain shell from its cap and pin (Source: Author’s personal collection) For glass insulators, corona can lead to the shattering of the glass shell. Corona gradient rings are normally installed at the line end of an insulator string of extra high voltage transmission lines to mitigate this problem. 

Pollution deposits on insulator shells: Pollution effluence from the environment which deposits on either glass or porcelain insulator surface gives rise to leakage current which leads to the formation of dry-band arcing in high humidity and wet conditions and, eventually, flashovers [44]. Pollution accumulation also reduces surface resistance, which effectively reduces the arcing distance of the insulator string. Typical types of contamination which give rise to this problem include dust of carbon or metal oxides, soluble salts, bird droppings, and agricultural spray. Because certain pollution deposits on the insulator discs can be spotted by visual inspection, utilities normally conduct routine washing of the insulators. Another method to solve the problem is to apply room temperature vulcanized (RTV) silicone rubber

50

coating on the insulator surface [45] to increase its hydrophobicity which maintains high surface resistance. Some utilities use creepage or leakage distance extenders which are attached to the insulators using special adhesives [31]. 

Internal puncture of porcelain cap and pin insulators [31]: Lightning strokes with steep wavefronts greater than about 2500kV/µs can penetrate the corner areas within the cap which is filled by porcelain if there are defects such as small air pockets or sand band material presence in the areas. Effectively, the puncture creates a path for the lightning current to conduct thereby internally ‘shorting’ the insulator. When this happens, the insulator then loses its insulating property.

For polymeric insulators, the electrical failure modes are due to: 

Surface damage The sheds and rod covering of the polymeric insulator is made up of the same material. The material not only gives protection to the dielectric FRP rod but also provides a hydrophobic surface (ability to prevent water filming on the surface) to the insulator. Certain polymeric materials are prone to damage which can result from [31]: o degradation due to effects of solar ultraviolet (UV) radiation which leads to chalking, cracking and crazing of the insulator surface o tracking and erosion of the polymeric material due to improperly mixed and incorrect amount of fillers o cuts and splices on the polymeric cover and sheds due to corona activity o attack by birds which like to chew the rubbery polymeric material [7] Degradation, tracking and erosion of the polymeric material can result in the insulator losing its hydrophobic properties. When this happens, wetting of contamination deposits on the insulator can initiate dry band arcing [31] which can lead to flashover across the insulator. Prolonged corona activity, particularly at the line end of the insulator, can cause radial splitting of sheds and rod cover which can also lead to tracking and flashover. As for the bird attack, the reduced

51

number of sheds results in reduced creepage distance which can also result in flashover. 

Flash under [36] Flash under is the term given to the phenomenon the FRP rod fails due to moisture ingress. A conductive path is essentially formed within the FRP rod as a result of the moisture ingress which leads to tracking and eventually flashover within the rod. The power arc developed within the rod is usually strong enough to effect a puncture through the polymeric covering, sometimes, fracturing the rod altogether.



Pollution accumulation The hydrophobic (water repellence) property of polymeric material makes it beneficial to use the polymeric insulator in contaminated environments. This is because water will not accumulate on the sheds and rod cover and thus the insulator remains dry, even if there is pollution effluence accumulated on the cover. This prevents dry-band arcing from occurring. However, the insulator can lose hydrophobicity with age or under extreme polluted and wet conditions [31]. When this happens, dry-band arcing can occur and this can lead to eventual flashover along the external part of the insulator.

2.5.2.3 Audible noise (AN) and radio interference (RI) It needs to be acknowledged that, apart from conductors, insulators are also the source for audible noise (AN) and radio interference (RI) on a transmission line. AN is generated by electrical discharges and microsparks due to corona activity. It can be detected by the sounds of crackling noise when standing near transmission lines. RI, on the other hand, is electromagnetic wave radiation from the conductors as a result of high-frequency current injection due to discharges on insulators [46]. Both AN and RI levels are higher during wet weather due to the higher discharge activities. Pollution accumulation on the insulator surface may also give rise to the noise levels. While there are no failures to the component of transmission lines which are attributed to AN or RI, the operation of the insulator is deemed failed if it emits AN and RI beyond the approved standard. Table 2-3 below gives

52

typical requirements for AN inception level and RI limits for insulators based on system voltage [46]: System

RI (KV)

AN Inception level (DB)

Voltage (KV) Single unit

String

Single unit

String

132

16

110

24

12

275

26

240

32

18

400

30

320

34

22

Table 2-3: AN and RI limits for insulator at typical system voltages(Source: [46]) Mitigation steps normally taken to solve this problem includes using corona gradient ring to distribute the high voltage stresses and introducing a hydrophobic layer on the insulator surface (such as RTV silicone coating or resistive glazing) so that corona discharge does not occur. 2.6 Chapter Summary In this chapter, we have discussed the major components of transmission lines, their respective functions and corresponding failure modes which are referenced throughout this thesis. It has firstly covered the functions and failure modes of transmission towers and structures. Then, it addressed the functions and failure modes of foundations, followed by conductors and, lastly, insulators. Some mitigation steps that utilities take to counter these failures have also been discussed. It has been highlighted in this chapter that most of the faults are electrical, mechanical or structural in nature. It has also been highlighted in this chapter that the major contributing factor to transmission line failures is the environmental conditions wherein the transmission line traverses. Most importantly, this chapter has provided the necessary insight into the failure mechanisms of transmission lines components. Failure of these components normally starts with gradual degradation in the form of defects. There are diagnostic methods that are available for utilities to detect these defects. These methods and techniques are covered in the following chapter.

53

54

CHAPTER 3: INSPECTION, DIAGNOSIS AND MAINTENANCE OF TRANSMISSION LINE COMPONENTS

3.1 Chapter Overview Having understood the failure mechanisms of transmission line components from the previous chapter, in this chapter we examine the available inspection and diagnosis techniques as reported in various sources of literature. This chapter is organized into two parts. The first part of this chapter provides a review of the available methods of diagnosis of the transmission line components. The methods discussed are either currently put in practice, in experimental stages or on trial by some utilities. Any shortcomings of these diagnosis methods are addressed. In order to appreciate the application of test and diagnosis methods, the second part of the chapter looks into inspection practices and maintenance strategies as currently conducted by utilities worldwide. A review of survey reports of inspection and maintenance practices conducted on utilities worldwide is presented here. Finally, an account of the most common method of inspection, i.e. visual inspection, is provided towards the end of this chapter. 3.2 Review of Component Diagnosis Methods Transmission lines in operation are exposed to continuous mechanical and electrical stresses as well as aging. Under normal circumstances, damage to transmission line components due to these factors is initially minor and insignificant. Over the course of time, it can lead to serious damage which, if not rectified in time, can result in forced interruptions. Forced interruptions of transmission lines due to any type of failure are highly undesirable for a number of reasons. These include:

55



The interruption may result in loss of supply to a large number of consumers or important consumers



Depending on system connectivity and operations, the interruption may initiate system-wide collapse of a power system



The interruption may incur costs to the utility company



The interruption may create a bad impression to the management of the utility company



The interruption may result in the utility company not complying to statutory regulations

Many utilities conduct diagnostic tests on the components to determine their condition and to predict their useful life. These tests are either done on the components at site sometimes during transmission lines inspection - or in laboratories after the components have been removed. 3.2.1 Test for Tower Structural Strength The integrity of transmission towers can be accurately determined mathematically using finite element analysis. Although the computations can be complex, computer programs are presently available to model the structure and analyse its behaviour under different simulated stress points.

To compare with the results obtained from the computer

simulation model, some utilities conduct actual in-situ tests on tower structures [47]. These tests are usually done on aged structures to assess its integrity and remaining life [48]. The tests are particularly useful if the utility needs to decide whether to replace the tower or refurbish it with new conductors. In these tests, sample towers are chosen from a group of aged tower structures. Test rigs are used to connect the different parts of the tower at one end and the other end to a high-powered bulldozer. The bulldozer then pulls the tower from a certain angle to apply the forces that the towers are subjected to as simulated in the computer program. The behaviour of the tower under these stresses and how it eventually fails provide information for the utility to determine the integrity of the tower. This information is also

56

used by the utility to decide what maintenance strategy to use in order to achieve the desired reliability. Figure 3-1 shows a sequence photograph of a tower site test.

Figure 3-1: Sequence photograph of a transmission tower collapse during a structural in-situ test (Source: [49]) 3.2.2 Diagnostic Tests on Tower Foundations As discussed in section 2.3.2, one of the failure modes of tower foundations is due to corrosion of steel-reinforced bar in concrete. Unless the degree of corrosion is at the end stage which is denoted by spalling of the concrete, the status of steel-reinforced bar corrosion in concrete cannot be determined by visual checks. One of the methods to determine the probability of steel-reinforced bar corrosion is by using the half-cell potential measurement [50]. The half-cell is basically comprised of a piece of metal in a solution of its own ions (e.g. copper bar in a copper sulphate solution). If this configuration is connected to another metal in its own solution (e.g. iron in ferrous hydroxide solution), there exists a potential difference between the two half-cells. This difference of potential results in dc current to flow between the two metals. If a dc voltmeter is connected between the two metals, a voltage reading can be obtained. The voltage reading indicates the dissolving of the metal in its own solution. The more negative the voltage reading on the voltmeter, the more active the metal is dissolving. Figure 3-2 shows the diagram of a half-cell potential reading of steel-reinforced bar in concrete.

57

Voltmeter + ferrous solution

Half cell

Concrete

Steel-reinforced bar

Figure 3-2: Diagram of the half-cell measurement method In the above diagram, if the steel-reinforced bar is not dissolving internally due to corrosive reaction, the potential measured is small (typically 0 to -200 mV will register on the voltmeter). If there is reaction and increasing amounts of steel is dissolving, the potential moves towards -350 mV. At more negative than -350 mV, the steel bar is corroding rapidly. The ASTM C876:1991 (American Society for Testing and Materials) [51] provides a guide to interpret half cell readings in the field. This is given in the Table 3-1: Half-cell potential relative to

Percentage chance of

copper/copper sulphate reference

active corrosion

electrode > -200 mV

Low (10% risk)

-200 mV to -350 mV

Intermediate (50% risk)

< -350 mV

High (90% risk)

< -500 mV

Severe corrosion

Table 3-1: ASTM C867:1999 criteria for corrosion of steel in concrete (Source: [51]) Together with visual inspection of the condition of concrete foundation, half-cell potential measurement provides an indicator of the integrity of the foundation.

58

3.2.3 Diagnostic Tests on Conductors As discussed in section 2.4.2, the major failure modes of conductors are due to corrosion, vibration and annealing. For an ACSR conductor, corrosion will commence when there is loss of galvanizing of the steel conductor. The loss of galvanization can be determined non-destructively by using the CORMON detector, which utilises an eddy current sensor incorporated into a motorized trolley that travels along the conductor to determine the level of galvanization loss [52]. The information is then relayed to a ground-based receiver where the data is stored for analysis. This method gives a signature of the condition of the conductor/joints and is quite reliable as the reading is not affected by humidity or ambient conditions. However, the CORMON equipment is very expensive and is rarely used in utility practices. The effect of corrosion on the conductor is to cause loss of surface contact between strands of the conductor and also within the joints. When current flows through the conductor, the area where there is loss of contact due to corrosion exhibits high resistance. This results in high temperature at that particular location. Therefore, the presence of corrosion along the conductor can be determined by using a thermographic camera or by measuring its resistance. Both methods may be used together for increased effectiveness. The live conductor and joint are firstly scanned using infrared equipment. This can either be done from the ground or from the air. A good conductor/joint should have a uniform temperature reading throughout. From the infrared equipment, hotspots can be easily determined by light/bright areas whereas cold/normal areas exhibit dark regions. Careful interpretation of the infrared indication is required since measurement can be affected by convective cooling of the conductor, electrical loads, component emissivity, thermal gradient and ambient conditions [53]. The user then notes the location of the hotspot and flags it to the maintenance crew who will then perform resistance measurement at the location. Current technology makes it possible to make conductor resistance measurement live from the air [25]. Measurement of the resistance of the conductor at the location will

59

confirm whether it was a true hotspot caused by deteriorating joint/conductor or just a solar reflection. For example, Marshall [25] explains that in Transpower New Zealand, the known ACSR ‘Goat’ conductor resistance of 89.1 microohm/m is compared against the resistance of the joint in question. A good joint should have lower resistances (between 45-65 microohm/m). If the joint or conductor exhibits resistances higher than 89.1 microohm/m, the joint is deemed to need replacing. Tests for conductor strength are normally done in the laboratories. Samples of conductor lengths that are taken from a particular site are subjected to tensile and torsional tests. They are normally subjected to a tensile load of 1% of the maximum strength and the number of turns it takes for the conductors to break [52] is recorded. The results of the test provide information about the condition of the conductor and they can also be used to predict its useful life. Experiments [54] conducted by Harvard et al. indicate that in general, aged conductors that have experienced prolonged vibration have lower tensile strength and ductility compared to those conductors that are situated in less aggressive environments. 3.2.4 Insulator Diagnostic Tests There are several in-service inspection and diagnostics methods that are being employed by transmission utilities to determine the condition of insulators: 

Visual inspection: This method is employed by almost all utilities in the world. Although is relatively easy to implement, the success of visual inspection of the insulators is associated with the level of detail that can be seen from the observation point and the acquired experience of the observer. Many utilities use the helicopter for this purpose. The use of high-powered binoculars is also quite common while doing the visual inspection. The following conditions can be detected by doing a visual check of the insulators [31, 36, 55]: o Surface pollution o Flashover burn marks o Chipped/cracked porcelain

60

o Broken glass o Pin corrosion o Polymeric shed damage due to gunshot o Exposed FRP rod o Chalking, tracking and splitting of polymeric cover and shed Abnormal observations are normally recorded for further observations, rectification or corrective actions depending on the severity of the condition. 

The ‘Buzz’ Method [31, 56] This method is mainly done on porcelain insulators to determine whether there is any internal short. The ‘Buzz’ method works on the principal that when two metallic parts with different potentials are short-circuited, arcing occurs. This is done by using an instrument with two metallic prongs at the end of a hot stick. The test is conducted on live lines whereby each of the insulator units on a string is tested individually by placing the prongs across each cap. For healthy insulator units, there is a potential difference across the units that are enough to cause sparks in the small air-gap at the tip of the prongs. The sparking activity comes with a buzzing sound - hence it is called the ‘Buzz’ method. No buzzing means no potential difference and therefore the insulator is deemed defective. This method is simple, requires little operator experience, relatively inexpensive, and provides clear cut results. However, it is very time consuming because each insulator on the tower needs to be tested. The work is particularly laborious at higher transmission voltages due to the large number of insulators per string.



Resistance Measurement Method [31, 56] This method uses an instrument which has a DC source for resistance measurement. The instrument, which has two prongs, is located at one end of a hot stick and a meter that indicates the DC resistance of the insulator unit at the other end. The two prongs are inserted across each cap and the meter provides resistance values that indicate either a defective or good insulator. High DC resistance readings of up to 5 mega ohms suggest healthy insulators whereas lower resistance readings indicate faulty ones.

61

This method is simple and provides clear cut results. However, the readings can be affected by relative humidity as they are only reliable if the measurements are taken in fairly dry weather conditions. Like the buzz method, this method is laborious and time consuming as each insulator on a string needs to be tested. 

Infrared Thermography This method is employed on cap and pin as well as polymeric insulators. Infrared equipment is used from the ground or in the air to scan the temperature distribution along the insulator string. Healthy insulators on a string exhibit high temperatures at the pin area whereas defective (punctured) units are cold as they do not support any voltage. From the thermovision equipment, light/bright areas denote healthy units whereas dark areas indicate faulty units. Although the thermovision equipment is expensive, thermographic inspection allows snapshot evaluation of a string that facilitates fast detection of faulty units and can prove to be cost-effective for long high voltage lines.



Electric Field Measurement Method [57] [58] Voltage across an insulator produces an electric field across the insulator surface and the air surrounding the insulator. For an insulator string, the distribution of the electric field smoothly varies from its lowest value at the earth end to the highest value at the live end of the string. If there is a defective insulator unit in the string, there exists an abrupt change in the magnitude of the electric field and the field distribution across the string is not smooth anymore. The field distribution is shown in Figure 3-3 [56]:

62

Figure 3-3: Electric field distribution across an insulator string with a punctured unit in the middle (Source: [56]) Measurement of electric field across insulator strings is done by means of an instrument that is fixed at the end of a hot stick. It has a built-in data logger that stores electric field reading at each insulator in the string. Starting from the first insulator unit, the instrument is placed adjacent to it and the user then slides the instrument along the string all the way to the last unit. The user then slides the instrument back to the starting point. The data is then downloaded to a computer which gives a display of the electric field along the insulator string. A graph of the electric field distribution along the string versus the insulator units provides information regarding which units in the string are faulty. The graph for an 18unit insulator string is shown in Figure 3-4 [56]:

63

Figure 3-4: Variation of electric field along an 18-unit insulator string which shows defective insulator units at insulator number 7, 11, 14 and 15 (Source: [56]) It was found that this method is more reliable in detecting defects in porcelain cap and pin insulators than the buzz or resistance methods [31]. Attempts have been made to use this method on polymeric insulators [58] but the test results showed that ambient humidity and contamination deposits greatly affect readings and produce confusing results. 

Corona Detection Method Defective insulator units can cause the electric field to be concentrated in and around the defective units. Under high voltage application, the place which is experiencing highly concentrated electric field can experience ionic breakdown of air. The ionic breakdown of air is called corona and its by-products are audible crackling noise (AN) and radio frequency (RF) interference. A device called the corona phone [31] which consists of a parabolic dish, very sensitive microphone and high gain amplifier are used to detect the sound of corona. This device is normally used to detect corona on polymeric insulators. Once there is evidence of audible noise, further investigation is done to detect the source of the noise. However, the measurement can be unreliable as it can be affected by background noise. Until quite recently, visual detection of corona with the naked eye was not possible due to its emission in the ultraviolet range. The development of daytime

64

corona viewing technology allows utilities to observe corona during the day and indicate the source, position and magnitude of such activity [59]. This special camera is being used by a number of utilities in the US and is being put into trial use by utilities in Japan, Hong Kong and the UK. Detection can be done either from the ground or an aerial platform [60]. This method may prove to be effective due to its straightforward spotting, high sensitivity and immunity from solar reflection. But there is a problem of determining whether the corona spotted implies serious damage which requires immediate fixing. Further investigation still needs to be carried out to determine the right course of action after corona scanning. The cost of such a camera is still high and the benefits of the technology have yet to be demonstrated. 

Other methods A review of current literature shows there are other emerging techniques that can be used to detect in-service faulty insulators. These include: o Detection by using radio frequency signatures and signal processing techniques [61] This technique uses the radio frequency (RF) emission from the faulty insulator as the input to a system that consists of an antenna, an amplifier, an oscilloscope and a computer. Radio frequency, which is transmitted in the frequency range of 30 MHz to 300 MHz, is not affected by background noise which typically exists in the 10 kHz to 40 kHz range. The radio frequency pulses are captured by the antenna and stored in the computer. The data is compared with signature characteristics of good insulators by using the oscilloscope and digital signal processing techniques. Faulty insulators can be detected if the captured signal waveform is different from signal signatures of healthy insulators. However, further investigation needs to be carried out to locate the faulty insulators in the string.

65

o Analysis of polymeric insulator samples [62] This method was introduced by researchers at Queensland University of Technology. In this method, samples of in-service polymeric insulator material are taken from energized transmission lines by using a novel hotline tool. The samples are then analysed in the laboratory by using the scanning electron microscope and Fourier Transform Infrared (FTIR) spectroscope to assess their conditions which are represented by numerical indices of oxidation, chalking and ester/ketone ratio. The results of the analysis indicate the degradation condition of the insulators and this can provide information for utilities to take necessary maintenance actions. However, the practicality of the method has yet to be proven. o Detection of corona current pulse [63, 64] This technique uses corona current pulses that emanates from faulty insulator units as inputs to a system that consists of a coupling antenna, a wide-band current sensor and a computer. The system is operated from the ground and picks up corona current signals from insulators of all the three phases. Captured signals that are distributed around 20 MHz denote faulty insulator strings. However, there is still a need to carry out further inspection to pin-point the faulty insulator units. The method is still being investigated for its applicability. o Measurement of Surface Pollution [65] In this technique, pollution deposits on the insulator surface are used as an indicator to conduct maintenance. It starts with the collection of pollution deposits and converting it to equivalent salt deposit density (ESDD) values. Tests on similar insulators in laboratories have shown that there exists relationships between leakage current, surface resistance, flashover voltage and ESDD values in that:  The lower the ESDD value, the lower the leakage current levels  The lower the ESDD value, the higher the surface resistance  The lower the ESDD value, the higher the flashover voltage value

66

The laboratory tests also determine the maximum ESDD values that are permissible before flashover occurs for each insulator types. The ESDD values of pollution deposit samples from the site are then used to compare with the values obtained from the laboratory. Maintenance on the insulators is only carried out when the ESDD values reach their critical levels. This method is useful for predicting insulator flashovers due to pollution but it does not give any indication about the condition of the insulator itself. In addition, the method can be expensive as the insulators need to be continuously monitored. Apart from electrical tests, utilities also conduct mechanical tests to determine the mechanical strength of the insulator. This is conducted at different intervals to check whether there is any reduction of mechanical strength compared to the specified mechanical load of new insulators and to determine mechanical strength limits of corroded pins for end-of-life criteria.

3.3 Review of Inspection and Maintenance Methods The main driver for maintenance is to achieve high circuit availability, reliability and safety. In order to achieve these objectives, transmission lines inspection and maintenance strategies are generally based on periodic (time-based) maintenance and condition-based maintenance. Many utilities are considering reliability-centred maintenance (RCM) techniques to further improve their maintenance quality and reduce maintenance cost. Periodic (time-based) maintenance is based on inspections that utilities conduct at fixed intervals. Periodic ground and aerial patrols provide information about the condition of ROW and access tracks, particularly whether there are any encroachments, human activities and tree growth. Maintenance actions such as cutting undergrowth and clearing access tracks are some of the activities that utilities do at fixed periods. Condition-based maintenance is usually carried out on components such as conductors, insulators, spacers and fittings so that preventive replacements can be implemented to

67

avoid failures. Increased surveillance is initially carried out on components that exhibit abnormal conditions and this is followed by sample testing and selective replacement of the component with similar defect. When it is economically justified, system-wide replacements of the component are made. RCM is focussed on maintaining the equipment to achieve the required availability and reliability whilst optimizing maintenance expenditures [66]. Its methodology is structured on establishing the required maintenance based on the consequences of failure. Failure of components that results in the highest cost and economic impact in a rank is given top inspection and replacement priority. In order to gain insight into the type of transmission line maintenance practice of utilities around the world, an overview of two international surveys conducted on utilities worldwide follows. 3.3.1 McMahon Survey on Inspection Practice of Australian and New Zealand Utilities In 1995, McMahon conducted a survey [2] of electricity supply utilities operating high voltage grids throughout Australia and New Zealand. The results of the survey are: 

Periodic visual inspections via ground and/or aerial patrols are conducted by all the utilities in the survey at various intervals. Most authorities conduct ground and/or aerial patrols twice per year. All authorities in the region maintain records of defects obtained from line patrols, some using electronic data loggers.



Several authorities use diagnostic techniques such as the CORMON device to detect conductor corrosion, radiography to detect conductor damage due to vibration, half-cell potential method to check for grillage corrosion and thermographic camera to check for hotspots at conductor joints.



Not many utilities carry out efforts to establish end-of-life criteria of the components and residual strength of ageing structures or components.

68



All authorities have some measure of transmission lines reliability, which is expressed in terms of availability, outage rates or loss of supply.



Live-line techniques for corrective and preventive maintenance are increasingly being used by all utilities as system operators have shown a reluctance to take out important lines.

The most common activity in live-line work is changing

insulators, installing aircraft warning devices, installing conductor repair rods, and removing foreign objects. All live-line works are done using the helicopter. 

There is an increasing interest among the authorities towards using computerised maintenance management systems to assist maintenance planning.

3.3.2 CIGRE Survey of Utility Assessment of Existing Transmission Lines In 2001, CIGRE formed a task force to collect and compare worldwide information about practices and experiences on inspection methods, diagnostic tools and defect assessment of existing transmission lines. The survey was responded by 61 utilities from 30 countries in Africa, Asia, Europe, North America, Oceania and South America. The findings are incorporated in a report [1]. Some of the findings include: 

All of the survey respondents indicate that inspection is limited to visual inspection. For lines above 150 kV, 74% of the respondents indicate that they use aerial visual inspection using helicopter. For lines up to 150 kV, 63% of the respondents point out that visual inspection by climbing towers is their main practice. Inspection on foot is becoming less popular.



74% of the respondents use formatted checklists for support decision; many of them use inspection guides that contain different categories (between 2 to 4 levels) to indicate different defect levels to suggest urgency of repairs. Some utilities do not use any guides at all.



75% of the respondents point out that the most typical form of defect is corrosion attack on steel components; many of them categorise corrosion attack

69

based on surface extent, location and depth. They also indicate that corrosion caused by normal weathering is more frequent than by industrial pollution. 

The most important cause of tower collapse is structural failure due to wind loading. This is followed by combined wind and ice loading and ice loading only.



Most utilities take precautions against vandalism and/or terrorism that could jeopardise the operation of the transmission line. This includes fencing, covering anchor bolts in concrete, using polymeric insulators and installing anti-climbing devices.

3.4 Visual Inspection It has been found from the surveys above that despite the advent of modern instruments to diagnose defects, a majority of utilities worldwide rely on visual inspection to determine the condition of transmission line components. Strategic maintenance decisions such as to replace/repair, flag for next maintenance cycle, or do nothing are all based on information gathered from visual inspection. Based on the surveys too, utilities conduct visual inspection either from the ground, from the top of the tower (by climbing), or from the air (in a helicopter). 3.4.1 Ground-level and Climbing Inspection In this method, the transmission lines are visually checked by a team of inspectors (also called linesmen) who walk or drive along the line route – if there is good access. Defects that can be detected visually during ground level inspection include: 

Ground line corrosion of steel tower legs



Cracks on concrete foundations



Corrosion, looseness or arcing marks at earthing joints



Verticality of the tower structure



Missing tower bracings, step bolts, danger plates or tower number/circuit plates



Corrosion of tower bracings



Bent tower bracings

70



Looseness of bracing nuts or step bolts



Reduction of clearance between conductor and ground or objects such as trees, telecommunication lines, low voltage lines, roads, railways and rivers



Ground erosion or earth movements at tower base



Unauthorized man-made structures and/or activities in the ROW



Broken glass insulators or severely chipped porcelain insulators



Missing conductor spacers and/or vibration dampers



Presence of bird nests on tower



Swelling of conductor strands near line terminations

Sometimes it is necessary for the inspectors to climb the transmission tower in order to get a good view of the components on the upper part of the tower i.e. the cross arms, insulators, fittings, conductors and earth wires. For instance, the presence of pollution deposits on insulator surface can be better detected when the inspectors are at the same level as the insulators. Some utilities conduct climbing inspections after taking the line out of service while others do it live. For live-line inspections, safety precautions must be taken into consideration when climbing so that safety clearance limits are not compromised. During ground level inspection, the inspectors normally make corrective actions that are minor in nature such as tightening of loose nuts, cutting danger trees, replacing lost step bolts and clearing tower base of shrubs. Other defects are normally recorded for further review and action by the maintenance engineer. Activities during climbing inspection are limited to observing whether there are any defects on the components and recording them. Since the component defects are observed qualitatively, the reports are often described using linguistic terms such as “medium cracks on porcelain insulator shells”, “heavy rust on tower stubs”, “slight deflection of tower steel bracings” etc. In many aspects, this is probably the best way to indicate information about a defect level. Ground-line and climbing inspection is a very labour intensive, time consuming and, to some extent, limited method since it very much depends on whether there is access to the transmission tower. It also incurs high cost as in certain areas the utility needs to use

71

special four-wheel-drive vehicles crossing over rugged terrains to get to the towers. For long lines, it takes a long time to complete a cycle of inspection. Furthermore, inspectors have to be sufficiently fit to carry out the work and they must possess a certain level of knowledge to assess the degree of defects. In addition, experiences have shown that no matter how careful or conscientious the inspection is performed, there is always a possibility of the inspector missing less pronounced defects that could lead to eventual failure [3], often resulting in forced interruptions. This may lead to reduced public confidence in the utility company which, in the end, may affect the company’s bottom line. 3.4.2 Aerial Inspection Aerial visual checks are now quite common throughout the world as a means to inspect the physical condition of transmission lines. The major advantage of airborne inspection is it provides access to transmission towers which is otherwise impossible to reach using ground transportation. For this reason too, airborne inspection is also used to conduct asset mapping and emergency line patrols [67]. Helicopters are normally used to carry inspectors for this purpose as they are able to hover over the lines, although several utilities in the US use fixed wing aircraft [68]. It enables location of defects from elevated positions and could provide a better field of vision to the inspector compared to observing from the ground. It is particularly beneficial in locating easily observed large scale defects such as broken cross arms, dropped conductors and severe ground erosion at tower base. Aerial inspection of transmission lines can be done at a much faster rate compared to ground line inspection. Figure 3-5 shows aerial transmission lines inspection using the helicopter.

72

Figure 3-5: Aerial visual inspection of transmission line components using the helicopter (Source: [69]) When conducting the inspection from the helicopter by using still or video cameras, the observer may be able to locate defects along the conductor span which may otherwise be impossible to detect with the naked eye. One example is damage of conductor due to gunshot, shown in Figure 3-6, which was taken from the helicopter:

Figure 3-6: Broken conductor strands due to gunshot (Source: [70]) The disadvantages of airborne visual inspection include [67]: 

Observer’s level of alertness may deteriorate over the course of inspection



Noisy environment of the aircraft may affect the observer’s concentration



Rapid rate of passing the structures may affect accuracy of detection



Inspection can only be done in fair and flight-friendly weather



High cost

73

Many technological innovations have surfaced in recent years in an effort to make aerial inspection exercise more effective. This includes automating some of the activities that are handled by the human observer. A survey [67] was conducted by EPRI to determine the emerging technologies that are being experimented for use in airborne inspection of transmission lines. A resume is given in Table 3-2. Technology

Description

Airborne Vibrometry Predicts component condition using laser doppler Inspection

Analyzed vibrometer

by Neural Network (AVIANN) Robotic Inspection of Unmanned aerial vehicle which is remote controllable Power Lines (RIPL)

while in flight and directed to various viewpoints to identify potential problems

AVCAN

Tracker Helicopter mounted cameras complete with GPS and

System

digital video that capture video photos and temperature data

MediaMapper

Collects digital still images and video that are indexed with GPS data

Ohmstick

Directly contacts energized lines and reads resistance in micro-ohms of joints, connectors and splices

Ultraprobe

Audio recording system used to detect ultrasonic frequencies emitted from corona discharges

Korona

Detection of electric field surrounding transmission lines and converting it to graphical representations Table 3-2: Emerging and available aerial inspection technologies

3.4.3 Drawbacks of Visual Inspection The results of a visual inspection exercise are largely dependent on the information that is gathered by the inspector. It is therefore important that the inspector provide quality information so that maintenance planners can confidently use the information to develop maintenance plans and strategies.

74

Unfortunately, inspectors are normal human beings and not machines or robots that can be programmed to produce accurate results all the time. Defect information that is based on human visual observation is largely influenced by the inspector’s perceptions. The interpretation of the defect level which, in this case, can only be described qualitatively, is successively based on the inherent knowledge, reasoning and experience of the inspector. One inspector’s interpretation of a certain defect level is different from another. Other factors that can influence the inspector include the environment wherefrom the observation is made, the inspector’s health and fitness level, and whether visual aids are used. These problems result in a high level of subjectivity and uncertainty during visual inspection. This often leads to a large variance in defect reporting, which makes it difficult for the utility management to progressively monitor the condition of the component. As a result, ineffective or wrong maintenance decisions are taken which extends to ineffective control of maintenance or repair costs. Therefore, it is conceivable that a tool is required that can be used by utilities to narrow down the level of subjectivity and uncertainty associated to visual inspection. Such a tool must be able to use the linguistic descriptions, infer the condition of the components and suggest the appropriate maintenance actions. In this way, the interpretation of a certain defect condition can be standardized and objective decisions can be made to improve maintenance quality. 3.5 Chapter Summary This chapter has provided an insight to the various methods of inspection and diagnostics of transmission line components that have been reported in various literatures. Some of these methods have been adopted by utilities whereas others are in the experimental stages. The advantages and disadvantages of the different methods are identified and discussed.

75

Two of the current survey reports on utility inspection and maintenance practice are discussed. It was found from the surveys reports that visual inspection is by far the main means of locating defects performed by many utilities in the world. Finally, shortcomings associated with visual inspection practice of transmission line components are further commented on. The next chapter introduces an artificial intelligence technique that can be used to improve the reliability of visual inspection. The proposed technique utilizes principles of fuzzy logic.

76

CHAPTER 4: FUZZY LOGIC AND FUZZY INFERENCE SYSTEM

4.1 Chapter Overview It was indicated in the previous chapter that there are inherent uncertainties in the data collected by visual inspection of transmission lines which are due in part to the inspectors’ limited knowledge. Hence a tool is needed to reduce these underlying uncertainties associated with visual inspection of transmission line components. Such a tool should be able to process information based on natural language descriptions which is used by inspectors when making defect assessments. Fuzzy logic was introduced in the 1960’s to deal with such fuzziness of human perception and decision making. The application of fuzzy logic in real world utilization is represented in knowledge-based fuzzy inference systems. The following chapter introduces the fundamental principles of fuzzy logic and fuzzy inference system which the research discussed in this thesis is applied. First, the fundamental concept of fuzzy logic is elucidated by describing a simple example. Membership functions and fuzzy rules within the context of their use in fuzzy logic are then explained. The chapter next proceeds to explain the application of fuzzy logic in a fuzzy inference system. The processing steps and techniques of inferencing used in a fuzzy inference system are further deliberated. Steps taken to design a fuzzy inference system are then described. Some examples of application of fuzzy logic and fuzzy inference systems in power utility are discussed at the end of this chapter. 4.2 Fuzzy Logic The concept of fuzzy logic was developed by Lotfi Zadeh, a professor at University of California, Berkeley, in the mid 1960’s as a way of processing data based on linguistic descriptions [4]. Unlike Boolean logic or classical logic, which assumes that every fact is

77

either entirely true or false, fuzzy logic extends Boolean logic to handle vague and imprecise expressions. According to Zadeh [4], the essential characteristics of fuzzy logic are: 

Exact reasoning is viewed as a limiting case of approximate reasoning



Everything is a matter of degree



Any logic system can be fuzzified



Knowledge is interpreted as a collection of equivalent and fuzzy constraints on a collection of variables



Inference is viewed as a process of propagation of fuzzy constraints

The idea is best explained by using an example. Suppose that Boolean logic is used to identify whether a room temperature is “hot” or “cold”. Most people would agree that 40oC is a “hot” room temperature and 10oC is a “cold” room temperature. However, if the room temperature falls to 25oC, it becomes much harder to classify the temperature as “hot” or “cold”. The concept in reality allows imprecision to be expressed in a quantitative fashion. This is done by introducing a set membership function, represented by μA(x), which maps element x to real values between 0 and 1; the value indicates the degree to which an element belongs to set A. A membership value of 0 (μA(x) = 0) indicates the element x is entirely outside the set, whereas a μA(x) = 1 indicates the element x lies entirely inside the given set A. Consider then the previous example: if fuzzy logic is used to represent the “hotness” of a room, 40 oC would have a membership value of 1 and 10 oC would have a membership value of 0. 25 oC, on the other hand, would have a “hotness” membership value of, say, 0.6 and a “coldness” membership value of, say, 0.3. 4.2.1 Membership Functions Membership values can be expressed in different types of fuzzy membership functions. The most common representations are the triangular and the trapezoidal membership functions [71]. Figure 4-1 shows an example of a triangular membership function, which

78

is expressed by (a1, a2, a3); it should be noted that the closer the number is to 20, the higher the degree of membership to number 20:

Degree of membership, μA A

1 0.75 0.5 0.25 0

Number (x) 10

20

30

Figure 4-1: Triangular membership function (10, 20, 30) To account for all values of the numbers between 10 and 30, the membership function above can be represented mathematically by (Eq. 4-1):    μ A (x)    

0, for x  10 x  10 / 10, for 10  x  20

30  x  / 10, for 0,

20  x  30 for x  30

(Eq. 4-1)

The membership function in Figure 4-1 can also be represented by a technique called the resolution identity of membership functions [72].

The technique is based on the

decomposition of the membership function into non-fuzzy level-sets or intervals by slicing the membership function at different membership levels (also called alpha-cuts or α-cuts) between 0 and 1. An α-cut set, denoted by Aα, consists of all elements that have a membership grade in the membership function greater than or equal to the value of α [72]. This definition can be written as: A α   x  X μ A (x)  α

(Eq. 4-2)

79

Therefore, the membership function A can be represented by the summation of all its αcut sets: 1

A   A

(Eq. 4-3)

 0

Thus the membership function shown in Figure 4-1 can also be represented by the values in Table 4-1 below: Α-CUT

Lower boundary

Upper boundary

0

10

30

0.25

12.5

27.5

0.5

15

25

0.75

17.5

22.5

1

20

20

Table 4-1: Determination of membership function from α-cut sets Other membership functions that are in use include Gaussian, Bell, and Sigmoid [73]. The intersection and union of 2 membership functions, A and B, are represented by (Eq. 4-4) and (Eq. 4-5) and Figures 4-2 and 4-3 respectively: Intersection (AND operator): ( A  B )( x)  minA( x), B( x)

A

B

A∩B

Figure 4-2: Intersection of 2 membership functions (A AND B)

80

(Eq. 4-4)

Union (OR operator): ( A  B)  maxA( x), B( x)

A

(Eq. 4-5)

B

AUB

Figure 4-3: Union of 2 membership functions (A OR B) 4.2.2 Fuzzy IF-THEN Rules The application of fuzzy logic in real world systems is mainly used with fuzzy IF-THEN rules. In this application, conditional statements take the form of: IF < premise > THEN < consequence > whereby both premise and consequence are characterized by fuzzy or linguistic elements respectively. Due to their straightforward forms, fuzzy if-then rules are often employed to process information captured by human reasoning in order to make decisions that are based on linguistic inputs. An example that illustrates this relationship is: IF < food is excellent > THEN < tip is high > where ‘excellent’ and ‘high’ are linguistic variables that can be characterized by membership functions. Through the use of linguistic inputs and membership functions, fuzzy IF-THEN rules can easily capture the spirit of “rule of thumb” used by humans. This is done by incorporating them in fuzzy inference systems.

81

4.3 Fuzzy Inference System Fuzzy inference systems [71] are essentially knowledge-based systems that utilize all of the concepts that have been described in the previous sections: fuzzy logic, fuzzy IF-THEN rules and membership functions. Figure 4-4 shows the components of a fuzzy inference system:

Fuzzy Inference system

Knowledge base

input

output Database

Rule base Defuzzification interface

Fuzzification interface

Aggregator

Figure 4-4: Components of a fuzzy inference system As shown in the above figure, a fuzzy inference system is made up of five functional blocks. They are: 

A rule base containing a number of fuzzy IF-THEN rules



A database which defines the membership functions



An aggregator which performs the inference operation based on the rules



A fuzzification interface which transforms the crisp inputs into degrees of match with linguistic values



A defuzzification interface which converts the fuzzy results into crisp outputs.

The steps performed by fuzzy inference systems when processing inputs are:

82

1. compare the input variables with the membership functions of the premise part to obtain the membership values of each linguistic terms – this step is called fuzzification 2. combine the membership values of the premise part to deduce firing strength of each rule using the selected operator 3. generate the consequence or results of each rule 4. aggregate the results or consequences to produce a crisp output – this step is called defuzzification 4.3.1 Inference and Defuzzification Techniques There are a number of inference techniques used in a fuzzy inference system. One of the techniques that is used mostly in various practical applications is the Mamdani inference technique [73]. The Mamdani technique was introduced by Professor Ebrahim Mamdani of London University in 1975 to control kiln temperature in a cement factory [71]. To come up with the technique, he applied a set of fuzzy rules supplied by experienced human operators rather than using theoretical formula. The technique takes the form of: IF < x is A > AND < y is B > THEN < z is C > where A, B, are fuzzy sets in the premise and C is a fuzzy set in the consequence. Figure 4-5 illustrates the Mamdani inference technique for a 2-input 1-output system.

83

Figure 4-5: Mamdani fuzzy inference method (Source: [74]) In the above figure, A1 and A2 represent the degree of memberships for fuzzy input x and B1 and B2 represent the degree of memberships for fuzzy input y. Use of the min operator indicates that the minimum of these two values are mapped to fuzzy output z, indicated by areas under C’1 and C’2 respectively. The resultant, which is the area under C’, is an aggregation (union) of areas under C1’ and C2’. In order to indicate an appropriate representative value for the final output, the aggregated output membership function has to be converted into a crisp form. The conversion is done in the defuzzification step of the inference system and can be achieved using any one of the five computational schemes: 

Centroid or Centre-of-Area (COA), zCOA: This method calculates the point which is central to the area under the aggregated output membership function. It is calculated using the following equation

z COA 

  ( z ) zdz   ( z )dz A

(Eq. 4-6)

A

where μA(z) is the aggregated output membership function. This is the most commonly used defuzzification technique and is very accurate. The only disadvantage of this technique is that it can be computationally difficult for complex membership functions.

84



Bisector of Area (BOA), zBOA: This method calculates the area under the aggregated output membership function and divides it into two equal areas. The point of division is returned as the bisector of the area. It is calculated using the following equation:

z BOA



 A (z)dz 

a

b



A

( z)dz

(Eq. 4-7)

z BOA

where a  minz z  Z , b  maxz z  Z  and zBOA is the value at which the area is divided equally by 2. This technique can also be computationally difficult if the out membership functions are complex. 

Mean-of-Maximum (MOM), zMOM: This method calculates the arithmetic mean of all the maximum values of the aggregated output membership function. It is calculated using the following equation:

z MOM 



 zdz  dz

(Eq. 4-8)

Smallest-of-Maximum (SOM), zSOM: This method returns the smallest of the maximum values on the aggregated output membership function. This defuzzification technique is very fast as it doesn’t require any calculations.



Largest-of-Maximum (LOM), zLOM: This method returns the largest of the maximum values on the aggregated output membership function. For the same reason as the Smallest-of-Maximum defuzzification technique, this technique is also very fast.

The above methods are graphically represented in Figure 4-6:

85

Figure 4-6: Defuzzification schemes to derive a crisp output (Source: [74]) 4.3.2 Developing Fuzzy Inference Systems There are essentially 5 steps that one can follow in developing fuzzy inference models [71]: 1. Specify the problem and define linguistic variables used 2. Determine fuzzy membership functions Membership functions can take a variety of shapes. However, trapezoidal and triangular membership functions can often provide adequate representation of the expert knowledge, while at the same time, simplifies the process of computation 3. Elicit and construct fuzzy rules At this point, the expert might be asked to describe how the problem can be solved using the fuzzy linguistic variables defined previously. The required knowledge can also be obtained from other sources such as books, manuals, specifications or observed human behaviour. 4. Encode the fuzzy membership functions, fuzzy rules and procedures to perform fuzzy inference in an expert system This can be done in either one of two methods: to write the codes and algorithms using high-end computer programming languages such as C/C++, FORTRAN or PASCAL, or to use fuzzy logic development tools such as MATLAB Fuzzy Logic Toolbox or other commercially available fuzzy knowledge builder software.

86

5. Evaluate and tune the system Finally, the system is tested to check whether it conforms to the required settings specified in the beginning. This step may prove to be laborious as various levels of input data needs to be fed into the system and the outputs verified to the satisfaction of the expert. Tuning the system may be required and this can be done by: 

Reviewing the model input and output variables



Reviewing the membership functions. If necessary, additional membership functions may be required to be defined.



Providing sufficient overlaps between neighbouring membership functions. It is suggested in [71] that triangle-to-triangle and trapezoidal-to-triangle membership functions should overlap around 25% to 50% of their bases.



Reviewing the rules. If necessary, additional rules may be required to be defined.



Revising the shapes of the membership functions.

4.4 Applications of Fuzzy Logic in Utility Environment Several published references have shown the benefits and successful application of fuzzy logic in the utility environment. These applications provide the motivation to implement fuzzy logic in this research. Below is a resume of some of these works: 

Kumar et. al. [75] designed a fuzzy inference system to determine the type of faults on a transmission line. The inputs to the fuzzy inference system are times required for the relay to trip (“very small” to “very large”) and the characteristic impedances of the line (“low”, “medium” and ‘high”). The output variables for the system are estimated distance of fault (“very close” to “very far”) and the type of fault (phase-phase fault, phase-ground fault and three-phase fault). Inference is by means of IF-THEN rules that relate the input membership functions and the out membership functions.

87

The system shows that linguistic representations of a range of crisp values (such as relay operating time and line impedances) can be used to infer the type of fault on the line. Further action can be taken once the fault is identified by the inference system. 

Hathout [76] introduced a method which he termed as “soft reliability assessment” to infer the reliability of transmission steel structures taking into consideration the structural design safety factors and damage consequences of deteriorating transmission steel structures. The level of deterioration is gathered from field inspection of structures and is represented by linguistic terms such as “very poor”, “fair” and “very good”. He used complex mathematical manipulations to relate the structure’s failure probability with cumulative distribution function of its standardized safety margin and its coefficient of variation of safety margin. Together with the fuzzy indicators of structural deterioration, he then used a computer program to determine the remaining structural integrity. Depending on the output of the program, further steps can be taken to improve the reliability of the structure. Hathout’s method shows that fuzzy logic principles need not be used in a fuzzy inference system to process information. The method shown is rather complicated but the use of computer program greatly simplifies calculation.



Marannino et. al. [77] developed a rule-based fuzzy logic decision support system to mitigate the risks of a power system voltage collapse. The input to the support system is a set of numerical variables that represent a snapshot of the operating system. This input is converted to symbolic and linguistic quantities (“low”, “average” and “high”) which are processed by a set of IF-THEN rules in the fuzzy logic system. The rules were devised by human experts. The output of the system provides a measure of the security level degradation of the power system with respect to the voltage collapse risk (“stable network”, “stressed network” or “collapse”). The system was successfully tested on the Italian extra high voltage network.

88

The system shows that the rules used in the inference system can be designed to model the reasoning of human expertise. These rules are used in an inference system that helps power network control engineers manage risks of voltage collapse. 

Islam and Ledwich [78] used fuzzy logic principles to relate results of transformer dissolved gas analysis (DGA) to types of incipient faults within the transformer. Input data of different levels of dissolved gases in transformer oil such as hydrogen, ethylene, methane and ethane, were converted to linguistic variables (“low”, “medium” and “high”) and different types of incipient faults (such as “thermal fault of high temperature”, “partial discharge of low energy intensity” and “electrical discharges of high energy”) were inferred using a fuzzy logic system. The authors used the semi-Cauchy type of membership function shape which “provides smooth transition of overlapping ascending and descending membership functions”. Tests on three power transformers showed that the system was capable of providing good results.

The applications mentioned above show that fuzzy logic is suitable for use in a decision support tool to solve power system problems. The use of appropriate membership functions and IF-THEN rules that represent expert knowledge in a fuzzy inference system allows information based on natural language be processed to suggest an action. These principles are used to design a knowledge-based fuzzy inference system for the assessment of transmission lines components. In the next chapter, the design and application of a fuzzy inference system to determine the condition of one of the most common and important components of transmission lines – porcelain cap and pin insulators – is shown. 4.5 Chapter Summary This chapter has described the principles of fuzzy logic and its application in fuzzy inference systems which are used in this research. The basic components of a fuzzy inference system i.e. membership functions, IF-THEN rules and defuzzification techniques have also been described. Some example applications of fuzzy logic and fuzzy inference system in use at power utilities have been presented. However, this is not an

89

inexhaustible list of sample usages of fuzzy logic in power utilities as there are various other applications. The reader should be led by the bibliography for further reading on the subject.

90

CHAPTER 5: USING FUZZY LOGIC FOR TRANSMISSION LINE DEFECT ASSESSMENT

5.1 Chapter Overview In the preceding chapters, we have reviewed the failure modes of transmission line components and the methods that are available to detect component defects. From the review, we have found that most utilities still use visual inspection as the main method of detecting component defects. As highlighted towards the end of chapter 3, visual inspection is a process that is inherently fuzzy. We therefore would like to examine whether fuzzy logic techniques can be applied to the improvement of transmission line visual inspection. The components that are normally inspected during routine transmission line visual inspection include towers/structures, conductors, fittings and insulators. In this chapter, we primarily present the design and implementation of a knowledge-based fuzzy inference system (FIS) applied to the visual inspection of porcelain cap and pin insulators. Firstly, a closer look at the mechanisms for insulator pin corrosion deterioration and broken porcelain shells are presented. Then, the chapter proceeds to explain the design and development of the insulator inspection fuzzy inference system. Reference is made in the chapter to the inspection practice of two utilities - Tenaga Nasional Berhad (Malaysia) and Powerlink Queensland (Australia) – to illustrate current utility practice in locating insulator defects. The fuzzy inference system is then applied to actual insulator data from both the utilities to simulate the use of the system during an insulator inspection exercise. Discussions regarding the experience of using the insulator inspection FIS are also presented. We then proceed to show that the output of the insulator inspection FIS can be used to assess the condition of a string of insulators and further, multiple strings in a transmission line. With the insulator string condition rating and a proposed numerical rating for severity of environment, a mathematical model is then developed to estimate the

91

recommended year for bulk replacement of corroded insulators. Simulated insulator data for different types of environment are used to show the applicability of the model. The chapter further explains how the FIS can be applied to the inspection of other components of the transmission line. Finally, an account of the potential savings in inspection and maintenance expenditure that can be achieved due to the introduction of FIS in Tenaga Nasional Berhad is deliberated at the end of the chapter. 5.2 Application of the FIS on Porcelain Cap and Pin Insulators Electric utilities around the world are currently using millions of porcelain cap and pin insulators. The fact that these insulators are still being used, despite the advent of ‘modern’ insulators such as those made of polymers, show that utilities have traditionally been satisfied with the performance of porcelain cap and pin insulators. In many utilities, however, these insulators are either operated close to or beyond their intended service life (30 to 50 years), many of which have recently resulted in an increase of observed problems [55]. As discussed in Section 3.2.4, the main method of in-service porcelain cap and pin insulator assessment is by visual assessment. The reason for this is because the type of defects associated with porcelain cap and pin insulators can be detected by assessing their physical appearance. The two major types of defects identified in this thesis are corrosion of the pin and breakage of the porcelain shell. In doing the insulator assessment to determine these types of defects, the inspector makes a visual observation either from the ground (often using visual aids such as high-powered binoculars or cameras), from the top of the tower (by climbing the tower structure) or from the air (in helicopters). Maintenance actions such as ‘do nothing’, ‘flag for next maintenance’ or ‘replace immediately’ are often taken as a result of these observations. Thus, it is imperative that the assessment exercise returns reliable report regarding the condition of the insulators so that the correct maintenance action is taken.

92

However, as discussed in Section 3.4.3, there exist the inherent problems associated with human visual inspection. One is that the observer makes his own judgment with regards to how he interprets the level of defect. As a result, defect reports from visual inspections vary from one observer to another. Another problem is that descriptions of the defects are usually explained by qualitative indicators that use linguistic terms. For example, the different levels of rusty insulator pins are best represented by statements such as “light rust”, “medium rust” or “very rusty”. Such statements are vague and subjective. As a result, in time, it is difficult to monitor insulator condition and to decide the appropriate maintenance action. Therefore, it is conceivable that a tool or method is made available to address these inherent problems. 5.2.1 Factors Affecting Pin Corrosion Insulator pin corrosion is primarily caused by the electrolytic action of leakage current. Leakage current is dependent on the voltage stress and the level of contamination. Other factors that affect the rate of corrosion are: 

Thickness of the pin



Amount of galvanising of the pin



Orientation of the insulator string (suspension or tension)



Rainfall



Presence of zinc sleeve

If we assume that the voltage profile along an insulator string is linear, then the voltage across each insulator unit on a 10 unit string at 132 kV would be around 7.5-8 kV and on a 16 unit string at 275 kV about 10 kV. But because of strong capacitance, the voltage profile along insulator string is not linear and typically takes the form shown in Figure 5-1 [7].

93

% Line voltage

120 100 80 60 40 20 0 1

2

3

4

5

6

7

8

9 10

Insulator Units (Earth end to line end)

Figure 5-1: Voltage profile across a 132 kV insulator string [7] The above figure shows that the highest voltage stress is at the line end where up to 20% of the line-to-ground voltage can apply. A high voltage stress increases the likelihood of electrical discharges. The stress is even higher if the insulators are contaminated, and with moisture present, corona activity can arise. One of the by-products of corona is ozone which can accelerate the corrosion process [29]. This explains why in some areas, corrosion is prevalent at the last insulator in the string. This is also a reason why some utilities give particular attention to the insulator unit at the line end of the string. Corrosion is more likely to occur in areas of high pollution, particularly where there is salt spray from the sea and very few rainfall washings. The rate of corrosion attack is also higher in these areas due to electrolytic reaction between salt deposits and atmospheric humidity. A CIGRE survey [79] on the serviceable life of line components found a range from around 25 to 80 years with a mean of 50 to 60 years for porcelain cap and pin insulators. Good quality porcelain insulators in clean environments are therefore expected to give an average serviceable life of 50 years. Insulators used in TNB for the 132 kV and 275 kV transmission lines have pin sizes of 16 mm (70 kN SML) and 20 mm (120 kN SML) respectively. The size of the pin will affect the leakage current density and hence the rate of corrosion. Most of the pin corrosion problems in TNB occur on the 132 kV towers.

94

The orientation of the insulator string on the tower, either in suspension or tension, affects the rate of corrosion. Most corrosion problems occur on suspension strings. This is because the undersides of the insulators do not get washed properly with rain and pollution level can build up in that location. Furthermore, moisture does not escape easily on the undersides of suspension insulators and this leads to an increase in corrosion rates in areas of high rainfall and humidity. To decrease the rate of pin corrosion, TNB uses insulators with zinc sleeves around the pin. These sleeves are specified to be not less than 3 mm and its purity must not be less than 99.9% in order to provide good galvanic performance. The zinc sleeve is fused on the pin shank to cover the area around the cement line of the insulator. It prolongs the rate of corrosion by acting as a sacrificial element before being fully consumed to expose the steel pin. The failure mechanism of insulator pins due to corrosion is as follows: 

The galvanising layer corrodes and is lost, exposing the steel pin



Corrosion of steel occurs, reducing the pin diameter and increasing tensile stress in the pin



The load stress exceeds the ultimate tensile strength of the pin, causing the pin to fail, and the conductor is dropped.

Based on this failure mechanism, several levels of corrosion can be detected visually to indicate the approximate criteria for insulator pin failure. Four distinct categories of pin corrosion levels are shown in Figures 5-2, 5-3, 5-4 and 5-5 below: 

Category 1 (No rust):

Figure 5-2: No rust insulator pin condition 95



Category 2 (Light Rust):

Figure 5-3: Light rust insulator pin condition



Category 3 (Medium Rust):

Figure 5-4: Medium rust insulator pin condition 

Category 4 (Heavy Rust):

Figure 5-5: Heavy rust insulator pin condition

96

The insulators shown in Figures 5-2 to 5-5 have no zinc sleeve around the pin and in Figure 5-3, most of the pin galvanising layer has been lost. Moubray explains that the term “potential failure” is the failure process at which it is possible to detect that the failure is occurring or about to occur [66]. The rust conditions of the pin can indicate its potential to fail i.e. break. And because rust conditions on an insulator pin deteriorate with time, the relationship between the pin’s rust condition and the time the insulator is put in service can be represented by the P-F curve; the P being the point in time when we can determine that the pin is failing (“potential failure”) and the F is the point at which it fails. This curve is shown in Figure 5-6.

Rust Condition

P-F Interval: Remaining time before imminent failure

No rust

Light Rust P Medium Rust Heavy Rust Failure

Year F End of life

Commission

Figure 5-6: Schematic representation of P-F curve for rust condition of insulator pin The P-F curve shown above has a big influence on the insulator inspection guide that is used in the field. This is because data from the field inspection should reflect the actual condition of the insulator and the state it is in with respect to the P-F curve. The most critical point is the state of the insulator at point P on the curve. Point P is an indication of the strength of the pin at which its remaining mechanical strength may not be enough to support the weight of the conductor and other insulators, particularly when there is contribution from wind and/or snow. It is critical that the inspector be able to detect the 'point P' condition of the insulator pin from the guide so

97

that appropriate action can be taken before it reaches a failure state. Some of these actions include to increase inspection intervals or to replace the insulator. It does not matter where on the string the insulator unit with rusty pin is as it only requires one really bad pin to take the whole line down. Therefore, for pin corrosion failure mode, the insulator unit with the worst pin corrosion in the string should be detected as an indicator of the remaining mechanical strength of the string. 5.2.2 Factors Affecting Chipped/Broken Porcelain Insulator Disc The porcelain disc is the part of the insulator that is largely responsible for its electrical performance. It provides the creepage distance that resists flashover that is due to either power frequency, switching surge or lightning surge. Consequently, insulators are designed to withstand high levels of electrical stress. There are two types of cap and pin insulators normally manufactured for transmission line application: the normal type and the anti-fog type. The difference between these two types is determined by the length of their creepage distance. Selection of the types is determined by the environmental condition of the areas through which the transmission line passes. Commonly-used insulators typically have creepage distance of around 300 mm. They are used on transmission lines that operate in clean and moderate environments. Figure 5-7 shows a cutaway drawing of this type of insulator showing the orientation of the ribs underneath the insulator:

Figure 5-7: Cutaway drawing of a normal type cap and pin insulator (Source: [80])

98

The anti-fog type insulators have longer creepage distance, around 440 mm, and are used in areas where the pollution levels are high. Figure 5-8 shows a cutaway drawing of an anti-fog type insulator showing the extended ribs underneath the insulator:

Figure 5-8: Cutaway drawing of an anti-fog type cap and pin insulator (Source: [81]) Electrical failures on porcelain insulators can occur internally or externally. Internal punctures are due to defects in the porcelain material in the form of voids and impurities. These defects result in electrical stress concentrations within the porcelain that could lead to the formation of an electrical path. An internally shorted insulator cannot be identified by visual inspection. External electrical failures occur in the form of flashovers. It is the dielectric breakdown of air as a result of very high electrical stress. On a string, the arc produced by the flashover usually travels away from the porcelain surface. External flashovers are determined by the magnitude and duration of the electrical stress and the condition of the insulator discs [31]. The number of insulators in a string provides the required insulation for the string to withstand flashover. Generally, the higher the line voltage, the longer the insulator string is. If the number of insulators in a string is reduced, this withstand distance will be compromised and therefore flashover is likely to occur. As discussed in Section 5.3, the voltage stress is highest at the line end and reduces in a non-linear fashion as we approach the earth end. Tests conducted by Booker and Jagtiani [82] on insulator strings found out that the performance of the insulators at the line end degrade more rapidly than those at the earth end due to the effect of higher local temperature. They concluded that the primary reason for the higher temperature is due to

99

the magnitude of electrical stresses at the line end. Therefore, it is important that the insulator units at the line end of the insulator string are in good condition to withstand the higher voltage stress. Porcelain is a very dense and tough material. It does not shatter like toughened glass insulators do when hit by an object such as gunshot bullets. Depending on the force of the impact, the porcelain either suffers cracks on the surface, chips, or breaks. Cracked surface, chipped or small-sized fractures do not affect the mechanical and electrical properties of the insulator and the overall integrity of the insulator string. But total breakage of the porcelain disc may pose a problem as it would reduce the creepage distance of the insulator string. If the insulator concerned is located at the line-end of the string where the voltage stress is the highest, then flashover failure may be imminent. Tests on damaged insulator discs have also found that the residual mechanical strength of the insulators is reduced by up to 35% [29]. Several levels of porcelain disc breakage are shown in Figures 5-9, 5-10, 5-11 and 5-12:

Figure 5-9: Chipped porcelain disc (Source: Author’s personal collection)

Figure 5-10: Small breakage of porcelain disc (Source: Author’s personal collection) 100

Figure 5-11: Major radial breakage of porcelain disc (Source: Author’s personal collection)

Figure 5-12: Total porcelain disc breakage (Source: Author’s personal collection) The outer physical condition of the porcelain discs on a string can be determined fairly easily during insulator visual inspection. Figures 5-13 and 5-14 show an insulator string in service which suffers broken discs as perceived during visual inspection of transmission lines.

Figure 5-13: Arrows show partially broken porcelain discs in an insulator string (Source: Author’s personal collection) 101

Figure 5-14: Arrows show 2 totally broken porcelain discs in an insulator string (Source: Author’s personal collection) The levels of defect on porcelain discs can then be categorised as shown in Table 5-1: Category

Condition

1 – No breakage

No observable defects on porcelain disc

2 – Chipped

Porcelain disc has suffered observable chips at the outer edge which do not extend into the second rib underneath the insulator. The insulator still retains its circular shape.

3 – Small breakage Porcelain is observed to be broken up to the second rib underneath the insulator. The insulator is observed to be no more round in shape. 4 – Large breakage The broken porcelain is extended to the head of the insulator. At least half of the insulator disc is gone. 5 – Total breakage

The porcelain disc does not exist.

Table 5-1: Category of in-service porcelain insulator disc defects Defects on porcelain discs generally do not deteriorate with time. The defects are random and mostly are due to public vandalism from rifle fire or slingshots.

102

5.2.3 The Insulator Visual Inspection Process Because pin corrosion and broken porcelain disc can be physically located, utilities normally use visual inspection to detect these defects. To illustrate, the practice of two utilities is given here as examples: 

TNB Practice: In TNB, inspection of insulators on 132 kV and 275 kV transmission lines is done during routine 3-yearly planned maintenance outage of the line. During the inspection, the inspector climbs up the tower and records the condition of the insulator in a form which is part of an inspection report. There is no inspection guide used and the condition of the insulator is determined by the sole judgment of the inspector. The format of the inspection report is in the form of a tick (√) denoting the insulator is in good condition and a cross (X) denoting it is in bad condition. Intermediate insulator conditions are taken note of and written in a separate table in the report form. The lines maintenance engineer then uses the information in the report to decide whether to take no action (for good condition insulators), flag for next maintenance (for partially defective insulators) or replace immediately (for bad insulators). Inspection data is recorded manually in log books and files. Since the condition indicators recorded during inspection are simplistic, it has not been possible for TNB to make a good analysis of the data in order to find a trend for deterioration patterns and to monitor the condition of the insulators. As a result, TNB has not been able to determine the expected serviceable life of its insulators in service. In effect, this situation renders its maintenance planning process ineffective. TNB has been looking for a method or tool to improve its insulator inspection and monitoring including its maintenance planning processes.



Powerlink Practice: Powerlink uses a 4 category visual classification for insulator pin corrosion. These categories are described in Table 5-2 [29]:

103

Category

Rust Type

Description

1

No Rust

Minimal loss of galvanising, only minor flaking and surface discoloration

2

Light Rust

Loss of galvanising, surface rust evident at cement/pin interface, no change in pin diameter

3

Medium Rust

Rust and pitting apparent, expansion of top section of pin, rust bloom increases the pin diameter by 10-20%, cement grout still visible

4

Heavy Rust

Rust bloom increases pin diameter by up to 70%, cement grout no longer visible

Table 5-2: Pin corrosion description used by Powerlink during visual inspection [29] During visual inspection, which is done live either by climbing the tower or from the helicopter, the inspector compares the perceived condition of the insulator with any of the 4 categories and records it in a report form. Preventive action is taken on insulators that have reached categories 3-Medium Rust and 4-Heavy Rust. Powerlink stores its insulator condition based on the categories of defects in a central computer database. Annually, its engineers assess the collected data and make plans for insulator replacements. As highlighted previously, the major problem with visual inspection is the high level of variance in inspection results. This is because the nature of visual inspection heavily relies on the personal judgement and intuition of the inspector. The inspection process as practiced in TNB produces the highest level of variance as the judgment whether an insulator is good or bad is solely based on the knowledge and experience of the inspector. A fairly new staff member would interpret the level of a defect different from one who has several years of experience. Nevertheless, a good insulator which is physically intact and has no defect would always be reported as ‘good’ and a very bad insulator with the entire disc broken would be interpreted as ‘bad’.

104

Powerlink’s practice of using an inspection guide reasonably narrows down the level of subjectivity as the inspector now can compare the actual insulator as he sees it with the defect description from the guide. However, the category numbers used are still fuzzy as during inspection, the condition of the insulator would not exactly be the same as described in the inspection guide. In other words, an insulator which is perceived as category 2-Light Rust could actually be interpreted as ‘about 2’ or ‘some light rust’. To deal with this uncertainty, the principles of fuzzy logic can be used since, as discussed in Chapter 4 of this thesis, it was specifically developed to deal with the fuzziness of human perception and decision making process. It also provides an organized framework for dealing with linguistic quantifiers such as “good”, “bad”, “fair” etc. which comes up during insulator visual defect assessment. The author of this thesis has developed a fuzzy inference system using Matlab’s proprietary Fuzzy Logic Toolbox that can be applied to the insulator visual inspection process. 5.2.4 The Fuzzy Inference System for Visual Inspection of Insulators 5.2.4.1 Assessment of Single Insulators The fuzzy inference system (FIS) which the author developed here models the insulator visual inspection process to assess rusty pins and broken porcelain discs. The respective inspection guides for pin corrosion and broken discs as discussed in the previous section are used to determine the linguistic variables and membership functions. Fuzzy IF-THEN rules are used in the Inference Engine to relate the input and the output membership functions. The Inference Engine is developed using the Mamdani inference technique and using the steps that were discussed in Section 4.3.2. The structure of the system, which follows the generic FIS structure shown in Figure 4-4 earlier, is shown in Figure 5-15:

105

Inputs

Inference Engine

Input 1: Pin Corrosion Category

Input and Output Membership Functions and Fuzzy Rules

Input 2: Porcelain Disc Condition Category

Output

Output: Insulator Condition

Figure 5-15: Structure of insulator inspection FIS The model above adopts the following steps that are taken during insulator inspection: 1. The inspector makes a visual assessment of each insulator unit in the string. 2. To check for rusty pins, the inspector compares the actual condition of the pin and compares it with any of the four categories in Table 5-2. The actual condition of the pin that depicts the closest to any of the four condition ratings is used as one of the inputs to the Inference Engine. 3. To check for the condition of porcelain disc, the inspector compares the actual porcelain condition of the insulator with the description provided in Table 5-1. The actual condition of the porcelain disc that portrays the closest to any of the five categories in the table is used as the other input to the Inference Engine. 4. The output of the Inference Engine indicates the condition of the insulator with respect to the pin corrosion and porcelain insulator defect conditions. It will be shown later that this indicator can be utilised as a measure to plan for insulator maintenance. The essential components of the FIS in Figure 5-15 are as follows: 

Linguistic variables: The input variables used are as follows: o Porcelain disc condition (from Table 5-1): No breakage, Chipped, Small Breakage, Large Breakage, Total Breakage o Rusty pin condition (from Table 5-2): No Rust, Light Rust, Medium Rust, Heavy Rust 106

The output variables used are: o Insulator condition: Very Poor, Poor, Fair, Good, Very Good Membership functions: Each of the fuzzy rules in the FIS is represented by membership functions and IF-THEN rules, which form one half of the knowledge-base used in the FIS. Figures 5-16, 5-17 and 5-18 below show the membership functions for the two inputs and one output FIS used in Fuzzy Rules.

Membership functions for pin rust conditions no-rust 1

lt-rust

med-rust

hvy-rust

0.8 Degree of membership



0.6

0.4

0.2

0 1

1.5

2

2.5 pin-corrosion

3

3.5

4

Figure 5-16: Input membership functions for pin rust conditions

107

Membership functions for porcelain shell conditions good 1

chipped

small-break

large-break

total-break

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5

3 Shell

3.5

4

4.5

5

Figure 5-17: Input membership functions for porcelain shell conditions

Membership functions for insulator condition v.poor 1

poor

fair

good

v.good

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5

3 Condition

3.5

4

4.5

5

Figure 5-18: Output membership functions for insulator condition The design of these membership functions follows the condition rating used during the inspection process. For example, it has been mentioned earlier that there are four condition ratings for pin corrosion used during inspection: ‘1-no rust’, ‘2-light rust’, ‘3-medium rust’ and ‘4-heavy rust’. In Figure 5-16, the same linguistic variables are used to describe the different grades of corrosion. The condition that exactly corresponds to each of the four prescribed condition ratings is assigned the maximum degree of membership (μ = 1.0). It can be then logically 108

inferred that those conditions that do not portray the prescribed condition in its entirety should have lower degrees of membership (μ < 1.0). For instance, if the condition of the pin insulator is between ‘2-light rust’ and ‘3-medium rust’, then it should have partial memberships of either condition ratings. As discussed in Chapter 4, membership functions can take various shapes. In this case, it is assumed that the degree of membership decreases linearly so that the membership functions used are triangular with base spreads not exceeding the immediate adjacent condition rating. The reason for this is that rust on the pin deteriorates progressively from one condition to another with time. For example, there is no possibility that a partial rating of ‘1-no rust’ pin condition will also have a partial rating of ‘4-heavy rust’ condition but there is a possibility that the condition falls between 2 adjacent rating conditions, such as ‘3-medium rust’ and ‘4-heavy rust’ conditions. This makes the crossing point of each membership function at 50% (or μ = 0.5), which has been suggested in Section 4.3.2. The same design rationale is used for creating the other membership functions used in Fuzzy Rules. The input membership functions for porcelain shell conditions are given in Figure 5-17 and the output membership functions for the resultant insulator conditions are shown in Figure 5-18. 

IF-THEN Rules The other half of the knowledge base of the insulator inspection FIS is represented by Fuzzy Rules. Fuzzy Rules specify the relationship between the premises and consequences used in the FIS and is expressed in terms of IFTHEN rules; the IF-part of the rule is the premise and the THEN-part of the rule is the consequence. In Figure 5-15, Fuzzy Rules in the Inference Engine describes the relationship between the different levels of rusty pins (input 1) and porcelain insulator conditions (input 2) with the condition of the insulator unit (output). In this case, inputs 1 and 2 are the antecedents and the output is the consequent.

109

There are four condition ratings (input 1 membership functions) that are used to describe rusty pins and five condition ratings (input 2 membership functions) to describe porcelain insulator conditions which can be used to infer one of the five conditions of the insulator unit. The total number of rules needed in a FIS is determined by the product of the number of membership functions used in the two antecedents. In this case, a total of twenty rules are required to infer the condition of the insulator pin. The rules can be derived using expert knowledge and experience, survey or books/reference manuals [83]. For the FIS used here, the author’s experience and knowledge (based on prior industry experience and research discussed in Chapters 2 and 3 of this thesis) are used as a basis to construct these rules. The rules are shown in Table 5-3. Note that the premises and consequence are all indicated by linguistic variables. For instance: Rule 2 indicates that: IF pin corrosion is NO RUST AND porcelain shell is CHIPPED, THEN insulator condition is GOOD. (‘NO RUST’, ‘CHIPPED’ and ‘GOOD’ are linguistic terms). Rule 18 indicates that: IF pin corrosion is HEAVY RUST AND porcelain shell is SMALL BREAKAGE, THEN insulator condition is VERY POOR. (‘HEAVY RUST’, ‘SMALL BREAKAGE’ and ‘VERY POOR’ are linguistic terms).

110

Rule

Pin corrosion is

1

NO RUST

2 3

NO RUST NO RUST

4

NO RUST

5

NO RUST

6 7 8

LIGHT RUST LIGHT RUST LIGHT RUST

9

LIGHT RUST

10

LIGHT RUST

11

16 17 18

MEDIUM RUST MEDIUM RUST MEDIUM RUST MEDIUM RUST MEDIUM RUST HEAVY RUST HEAVY RUST HEAVY RUST

19

HEAVY RUST

20

HEAVY RUST

12 13 14 15

IF

Porcelain shell condition is GOOD

AND

CHIPPED SMALL BREAKAGE LARGE BREAKAGE TOTAL BREAKAGE GOOD CHIPPED SMALL BREAKAGE LARGE BREAKAGE TOTAL BREAKAGE GOOD

Insulator condition is VERY GOOD GOOD FAIR POOR VERY POOR GOOD FAIR FAIR POOR VERY POOR POOR

THEN

CHIPPED

POOR

SMALL BREAKAGE LARGE BREAKAGE TOTAL BREAKAGE GOOD CHIPPED SMALL BREAKAGE LARGE BREAKAGE TOTAL BREAKAGE

POOR VERY POOR VERY POOR VERY POOR VERY POOR VERY POOR VERY POOR VERY POOR

Table 5-3: IF-THEN rules used in the insulator inspection FIS Combining both the input membership functions and the output membership function with the rules above, a 3-dimensional plot can be obtained to give a snapshot relationship between the inputs and output of the FIS as shown in Figure 5-19 below:

111

condition

4 3 2 1 2 3 4 shell

5

4

3.5

3

2.5

2

1.5

1

pin-corrosion

Figure 5-19: A 3-dimensional plot of insulator inspection FIS showing the relationship between the two inputs and output 

Defuzzification As mentioned in Section 4.3.1, there are four common defuzzification methods that are used in the Mamdani inference technique. The most commonly used defuzzification method is the centroid or Centre-of-Area (COA) and is also the method chosen for the insulator inspection FIS. Recall that the centroid can be calculated using Eq. 4-6. In this insulator inspection FIS, the defuzzified value is used to suggest the appropriate maintenance action to be done on the insulator. Table 5-4 shows a list of appropriate maintenance actions that can be taken on the insulator based on the centroid value of the output. Centroid value

Suggested maintenance decision

1-2

Replace insulator immediately

2-3

Increase monitoring frequency

3-4

Flag for next maintenance cycle

4-5

Do nothing

Table 5-4: Suggested insulator maintenance decision 

Application of the Insulator Inspection FIS

112

To test the FIS, three sample insulator conditions, which were taken from the field, are used as inputs to the FIS. Sample 1 Figure 5-20 shows insulator Sample 1 which was taken from the Prai-Bedong 132 kV line in the Northern part of Malaysia. From physical inspection, it was found that there is no sign of defects on the porcelain shell but the pin seems to suffer some corrosion. Despite a change of colour and rusty surface, the pin still retains its original diameter.

Figure 5-20: Insulator Sample 1 Comparing the pin rust condition in Figure 5-20 above with the four pin rust categories in Table 5-2, it is decided that the pin rust condition above falls within pin rust categories 2-light rust and 3-medium rust; the membership value of the 3medium rust condition (μmedium rust(pin corrosion) = 0.8) is higher than the 2-light rust condition (μlight rust(pin corrosion) = 0.2). This is shown in Figure 5-21:

113

Membership functions for pin rust conditions no-rust 1

lt-rust

med-rust

hvy-rust

Degree of membership

0.8

0.6

0.4

0.2

0 1

2

1.5

3

2.5 pin-corrosion

4

3.5

Figure 5-21: Pin rust condition for insulator Sample 1 It can also be seen from Figure 6-20 that the porcelain shell condition of the sample insulator has no defects. This means that it has a high membership of category 1-very good porcelain shell condition (μvery

(shell) = 1.0). This is

good

shown in Figure 5-22:

Membership functions for porcelain shell conditions good 1

chipped

small-break

large-break

total-break

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5

3 Shell

3.5

4

4.5

5

Figure 5-22: Porcelain shell condition for insulator Sample 1 Referring to the rules in Table 5-3, the conditions shown in Figures 5-21 and 5-22 invoke Rule 6 and Rule 11 of the FIS.

114

Rule 6 indicates that: IF pin corrosion is LIGHT RUST AND porcelain shell is GOOD, THEN insulator condition is GOOD. Rule 11 indicates that: IF pin corrosion is MEDIUM RUST AND porcelain shell is GOOD, THEN insulator condition is POOR. The AND operator used in the rules infers that the minimum criterion is used in the resultant. The results of invoking rule Rules 6 and 11 are shown in Figures 623 and 6-24 respectively:

lt-rust

1 0.8 0.6 0.4 0.2 0 1

good

1 1.5

2

2.5 3 pin-corrosion

3.5

4

min

μlt-rust (pin corrosion) =0.2

0.8 0.6 0.4 0.2 0

good 1

1

1.5

2

0.8 0.6

2.5

3.5 3 Condition

4

μgood(condition)=0.2

0.4 0.2 0 1

1.5

2

2.5

3 3.5 Shell

4

4.5

5

μgood (shell)=1.0

Figure 5-23: Resultant of invoking Rule 6

115

4.5

5

med-rust

1 0.8 0.6 0.4 0.2

poor

1 0 1

1.5

2 2.5 pin-corrosion

3

3.5

min

4

0.8

μmed-rust(pin corrosion)=0.8

0.6 0.4 0.2

good 1

0

0.8

1

1.5

2

0.6 0.4

2.5

3 3.5 Condition

4

4.5

5

μpoor(condition)=0.8

0.2 0 1

1.5

2

2.5

3 3.5 Shell

4

4.5

5

μgood(shell)=1.0

Figure 5-24: Resultant of invoking Rule 11 The resultants of Rules 6 and 11 are then aggregated to indicate the overall condition of the insulator. This is shown in Figure 5-25 below:

poor

1

good

0.8 0.6 0.4 0.2 0 1

1.5

2

2.5

3 3.5 Condition

4

4.5

5

Figure 5-25: Aggregation of Rules 6 and 11 resultants The above aggregation shown in Figure 5-25 is still represented by fuzzy membership functions. In order to determine the overall condition of the insulator, it is necessary to resolve the above aggregation into a crisp value. This step is called defuzzification.

116

Applying Eq. 4-6 to the shaded areas in Figure 5-25, the centroid zCOA is found to be 2.55 (i.e. zCOA = 2.55). Cross-referencing this value to Table 5-4, it is suggested that the monitoring frequency of this sample insulator be increased. Sample 2 Figure 5-26 below shows insulator Sample 2. The insulator is taken from a 132kV line in the Southern part of Malaysia.

Figure 5-26: Insulator Sample 2 It can be seen that the insulator above is chipped-off at the outer edge of the porcelain shell. The chip is considered to be small as the insulator is still generally round in shape. Comparing with the category descriptions in Table 5-1, it is decided that the shell condition corresponds to both categories 2-chipped and 3small breakage, with a higher membership of category 2-chipped than category 3small breakage. This is represented in Figure 5-27.

117

Membership functions for porcelain shell conditions good 1

chipped

small-break

large-break

total-break

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5

3 Shell

3.5

4

4.5

5

Figure 5-27: Porcelain shell condition for insulator Sample 2 There seems to be no signs of rust at the insulator pin which, from Table 5-2, corresponds to category 1-no rust. This is represented in Figure 5-28.

Membership functions for pin rust conditions no-rust 1

lt-rust

med-rust

hvy-rust

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5 pin-corrosion

3

3.5

4

Figure 5-28: Pin rust condition for insulator Sample 2 Referring the above conditions as shown in Figure 5-27 and 5-28 to Table 5-3, Rule 2 and Rule 3 are invoked. Rule 2 indicates that: IF pin corrosion is NO RUST AND porcelain shell is CHIPPED, THEN insulator condition is GOOD.

118

Rule 3 indicates that: IF pin corrosion is NO RUST AND porcelain shell is SMALL BREAK, THEN insulator condition is FAIR. Again, the AND operator in the rules implies that the minimum criterion is used in the inference. The results of firing rules 2 and 3 are as shown in Figures 5-29 and 5-30 respectively.

no-rust

1 0.8

min

0.6 0.4 0.2 0 1.5

1

2

2.5 3 pin-corrosion

3.5

4

good

1

μno rust(pin corrosion)=1.0

0.8 0.6 0.4 0.2

chipped

1

0 1

1.5

2

2.5

3

0.8 Condition 0.6 0.4 0.2 0 1

1.5

2

2.5

3 Shell

3.5

4

4.5

5

μchipped(shell)=0.7 Figure 5-29: Resultant of invoking Rule 2

119

3.5

4

4.5

5

1

no-rust

0.8

min

0.6 0.4 0.2 0 1

1.5

2

2.5

3

3.5

fair

1

4

pin-corrosion

0.8

μno rust(pin corrosion)=1.0

0.6 04 0.2

small-break

1

0

0.8

1

1.5

2.5

2

0.6

3.5 3 Condition

4

4.5

5

0.4 0.2 0 1

1.5

2

2.5

3.5

3

4

4.5

5

Shell

μsmall break(shell)=0.3

Figure 5-30: Resultant of invoking Rule 3 The results of Rules 2 and 3 are then aggregated to indicate the overall condition of insulator Sample 2. This is shown in Figure 5-31.

fair

1

good

0.8 0.6 0.4 0.2 0 1

1.5

2

2.5

3 3.5 Condition

4

4.5

5

Figure 5-31: Aggregation of Rules 2 and 3 resultants Applying Eq. 4-6 to the shaded area in Figure 5-31, the centroid is found to be 3.71 (zCOA = 3.71). Cross-referencing this value to Table 5-4, it is suggested that insulator Sample 2 be thoroughly inspected in the next maintenance cycle.

120

Sample 3 Figure 5-32 shows insulator sample 3 which had been taken down from a Powerlink 132 kV transmission line. The insulator was removed from the line because it appeared to have suffered severe corrosion at the pin and most of the edge of the porcelain shell has been chipped off, but it still retains its round shape.

Figure 5-32: Insulator Sample 3 Comparing the physical condition of the porcelain shell with the category descriptions in Table 5-1, it is suggested that the shell condition is corresponding to full membership of category 2-chipped. This is shown in Figure 5-33.

Membership functions for porcelain shell conditions good 1

chipped

small-break

large-break

total-break

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5

3 Shell

3.5

4

4.5

5

Figure 5-33: Porcelain shell condition for insulator sample 3

121

As for the pin rust condition, it decided that the pin condition is corresponding to both category 3-medium rust and category 4-heavy rust from Table 5-2, with a higher membership of category 3-medium rust. This is shown in Figure 5-34.

Membership functions for pin rust conditions no-rust 1

lt-rust

med-rust

hvy-rust

Degree of membership

0.8

0.6

0.4

0.2

0 1

1.5

2

2.5 pin-corrosion

3

3.5

4

Figure 5-34: Pin rust condition for insulator sample 3 Referring to Table 5-3, the above conditions shown in Figures 5-34 and 5-35 will invoke Rules 12 and 17. Rule 12 indicates that: IF pin corrosion is MEDIUM RUST AND porcelain shell is CHIPPED, THEN insulator condition is POOR. Rule 17 indicates that: IF pin corrosion is HEAVY RUST AND porcelain shell is CHIPPED, THEN insulator condition is VERY POOR. The results of firing these 2 rules are shown in Figures 5-35 and 5-36 respectively. Again, the AND operator in the rules implies that the minimum criterion is used in the inference. The results of firing rules 12 and 17 are as shown in Figures 5-35 and 5-36 respectively.

122

med-rust

1 0.8

min

0.6 0.4 0.2 0 1

2

1.5

3

2.5

poor

1

4

3.5

pin-corrosion

0.8

μmedium rust(pin corrosion)=0.9

0.6 0.4 0.2

1

0

chipped

1

0.8

2

1.5

2.5

3

3.5

4

4.5

5

Condition

0.6 0.4 0.2 0 1

1.5

2

2.5

3

Shell

3.5

4

4.5

5

μchipped(shell)=1.0

Figure 5-35: Resultant of invoking Rule 12

1

hvy-rust

0.8

min

0.6

0.4

0.2

1

v.poor

0 1

1.5

2

2.5

3

3.5

0.8

4

pin-corrosion 0.6

μheavy rust(pin corrosion)=0.1

0.4 0.2

1

chipped

0 1

0.8

1.5

2

2.5

3

3.5

Condition

4

4.5

5

0.6

0.4

0.2

0 1

1.5

2

2.5

3

3.5

4

4.5

5

Shell

μchipped(shell)=1.0

Figure 5-36: Resultant of invoking Rule 17 The results of Rules 12 and 17 are then aggregated to indicate the overall condition of insulator Sample 2. This is shown in Figure 5-37.

123

v.poor 1

poor

0.8 0.6 0.4 0.2 0 1

1.5

2

2.5

3 3.5 Condition

4

4.5

5

Figure 5-37: Aggregation of Rule 12 and Rule 17 resultants Applying Eq. 4-6 to the shaded area in Figure 5-37, the centroid is found to be 1.96 (zCOA = 1.96). Cross-referencing this value to Table 5-4, it implies that insulator Sample 3 needs to be replaced immediately. 

Discussion on the results of insulator inspection FIS From the sample applications above, it has been shown that, together with an insulator inspection guide for rusty pins and broken porcelain discs, the insulator inspection FIS can assist visual evaluation of defective insulator units on the string by inferring the defect level of each insulator

with respect to its physical

appearance and condition. With the insulator inspection FIS, the inspector is only required to indicate the appearance of rust at the insulator pin and chipping/breakage of the porcelain shell by comparing these conditions with the condition ratings used in Tables 5-1 and 5-2 respectively. With the insulator inspection FIS, inference is undertaken by the Inference Engine of the FIS and not by the inspector thereby relieving him/her from making any judgement with regards to the overall condition of the insulator. Because the same knowledgebase is used in the Inference Engine, consistency in the output is achieved. The heart of the insulator inspection FIS is the knowledge-base that forms the Inference Engine. In this case, the knowledge of the human expert is translated to IF-THEN rules used in the Inference Engine. It can be argued that the level of expertise from one human expert may differ from another such that one expert’s design of the rules may not be the same with another but as long as they are fixed in the Inference Engine, the outputs of the FIS will always be consistent. In the

124

insulator inspection FIS model shown in this thesis, the author’s knowledge and experience was directly incorporated in the Inference Engine. In the absence of an expert, the knowledge-base can be created indirectly by knowledge engineers using such knowledge elicitation techniques as knowledge representation and knowledge modelling [84]. It has been shown that the insulator inspection FIS works by converting qualitative indicators of the insulator conditions to numerical values that make it possible to quantify the insulators’ level of defects. Apart from making it possible to store these numerical values in a computer database for future trending and deterioration monitoring, it has been shown that the numerical values can also be used to suggest the maintenance crew taking appropriate maintenance actions. Because the maintenance actions are also standardized as shown in Table 5-4, consistency in decision-making for maintenance actions is therefore possible. There are, however, some noticeable limitations of the insulator inspection FIS. These include: o Despite the use of the inspection guide, subjectivity in determining the insulator defect level still exists as the results of the inspection now depends on how well the inspector compares the perceived actual insulator condition with those conditions indicated in the inspection guide. However, in the case of TNB inspection practice, the level of subjectivity can be significantly smaller than the level before since the inspectors now use an inspection guide that provides a basis for them to determine the defect level. o The use of descriptive defect conditions in the inspection guide may not be as effective as a pictorial representation of defect levels. o The output of the FIS is restricted by the design and number of membership functions and rules of the knowledge–base in the Inference Engine. In this particular design, triangular membership functions were used by the author to represent the linguistic terms to indicate the different defect levels. Different membership functions such as trapezoidal, Gaussian or sigmoid may yield different outputs. Defect

125

determination could also be improved by adding more defect levels in the inspection guide. For example, in Table 5-2, four categories of rust levels are used in the FIS, namely “No Rust”, “Light Rust”, “Medium Rust” and “Heavy Rust”. By adding intermediate categories such as “Very Light Rust”, “Light-Medium Rust” and “Medium-Heavy Rust”, the variability between one defect level to another will become smaller. However, more rules will need to be defined in the knowledge-base. 

Recommendations In order to improve insulator defect assessment, it is suggested that: o TNB to initiate use of insulator inspection guide and defect assessment criteria together with the insulator inspection FIS. TNB should also consider doing away with manual recording of inspection data in files and establishing a computerised database system that can be used to store and analyse the insulator condition data. It is foreseeable that this method will assist TNB to achieve more effective maintenance planning of its insulator assets especially after several inspection cycles when trends in insulator data will become apparent. o Powerlink to continue using existing insulator inspection guide together with the insulator inspection FIS. The FIS should be able to assist Powerlink inspectors in improving the insulator evaluation process because insulator conditions that fall between two defect categories can now be quantified.

5.2.4.2 Assessment of Multiple Insulators in a String As shown in the example assessments of insulator units in Section 5.6.1, with the insulator inspection FIS, each insulator in a string is given a numerical rating that indicates its pin rust and porcelain shell conditions. Depending on the defuzzified output of the FIS, the unit is recommended to be replaced immediately, flagged for next inspection cycle or put into closer monitoring frequency.

126

For insulator strings with rusty pins, it is very likely that the same rust conditions would appear on each of the insulator units in the string. This is because each insulator unit in the string is exposed to the same environmental conditions. The overall condition rating of the string is then simply the arithmetic mean of all the individual condition ratings of insulator units in the string. By doing this, each string is given a numerical rating that indicates its overall pin rust conditions. This numerical rating of each string can then be cross-referenced back to Table 5-4 for suggested maintenance action. The given numerical rating makes it possible to store the condition indicator of defective porcelains in a database which would be difficult to do if qualitative indicators were used. With each inspection cycle, the database can be updated to indicate the latest condition rating of the insulator string and this makes it possible to track the deterioration rate of the insulator string due to pin rust. Broken porcelain shells are easily detectible on a string as their defective appearance would stand out in a string of perfect insulator units. Broken porcelain shells generally occur in areas that are accessible to the public but secluded from populated places such as residential or urban developments. This is because vandals, who cause many of the broken porcelain shells by making them targets for rifle or slingshot practice, normally operate in such isolated but accessible areas. Broken porcelain shells are therefore random in nature and, unlike pin corrosion, they do not deteriorate with time and are not affected by environmental conditions. Chipped porcelain shells do not affect the electrical and mechanical properties of the insulators but, as explained in Section 5.4, totally broken porcelain shells result in reduction of the insulator’s insulating property and reduction of the insulator’s residual mechanical strength by up to 35%. Therefore it is imperative that insulators with broken porcelain shells be replaced immediately. The numerical rating obtained from the FIS which infers the condition of the insulator based on the condition of the porcelain shells can be used as a guide to replace the insulator. For insulators that have broken porcelain shells, TNB practice is to only replace those that are broken in the string and leave those units that are still intact on the string. On the other hand, Powerlink practice is to replace the whole string with new insulators. Gillespie [85] explains that the reason for this is that Powerlink’s operational policy is not to mix

127

different batches of insulator units in a string. In the author’s opinion, Powerlink’s practice is better than that of TNB as, from the asset management perspective, having insulator units of the same batch in the string makes it easier to track its field performance. 5.2.4.3 Assessment of Multiple Insulator Strings in a Transmission Line The serviceable life of a transmission line is predominantly determined by the environment that it passes through. For the line which is installed with porcelain cap and pin insulators, corrosion of the steel pin largely determines the remaining mechanical strength, thus the serviceable life, of the insulators. Steel corrosion deteriorates with time and the rate of corrosion deterioration is ascertained by how aggressive the environment is. The aggressiveness of the site can be obtained from a variety of sources such as: 

Back tracking from actual service experience



From national or state corrosion surveys by various agencies



From national or state weather data



From actual on-site corrosion monitoring programme using tokens



From pollution severity classification recommended by the International Electrotechnical Commission (IEC)



From best guess approximations of the relative severity of the site compared to others (for example, a line section close to the sea would experience a higher corrosion rate compared to one which is located in a sheltered valley).

From the above sources, it is possible to differentiate the aggressiveness of one site from another. It is therefore possible to assign specific codes to quantify the aggressiveness of the environment. For example, from classification of pollution severity recommended by the IEC [86], a coding scale to identify the aggressiveness of each environment can be introduced. This is shown in Table 5-5:

128

Environmental Code

Environmental Condition General Environment

Topographical Environment

Industrial Environment

Water Environment

10 9 8

Clean

All mountain areas and most agricultural areas with good rainfall

Remote from industrial and housing areas

Areas located more than 10 km from sea coast

7 6 5

Mild

Agricultural areas subject to occasional periods of below average rainfall

Industrial areas with low gaseous discharge or within 2 km of sources of conductive dust

Areas located 3 to 10 km from ocean water

4 3

Corrosive

Areas with moderate dust deposits and low rainfall but high relative humidity

High density industrial areas or within 2 km of chemical plants, fertiliser factories etc.

Areas between 1 and 3 km from ocean water or sources of soluble salts

2 1

Highly Corrosive

Arid areas with low rainfall patterns for long periods

Industrial areas which produce thick conducive deposits

Areas within 1 km of the sea coast

Table 5-5: Introduction of environmental coding for areas based on IEC’s pollution severity classification [86] A more accurate representation of how severe or corrosive the environment is in a specific location is by monitoring the rate of corrosion using sample tokens. This is done by measuring the time it takes for the sample tokens, which are put within the area where the tower is located, to change its state due to corrosion. This method, however, takes time. Marshall [87] made the measurements and suggested the following codes as shown in Table 5-6 for the rate of corrosion deterioration of zinc tokens, which corresponds to the severity of the environment.

129

Zinc Loss Rates per Year

Environmental Code

Environmental

(G/M2/YR)

Condition

6

10

Clean

7.5

8

Mild

10

6

15

4

30

2

60

1

Corrosive Highly Corrosive

Table 5-6: Environmental coding for zinc loss [87] The purpose of quantifying the different types of environment is so that it can be used in a predictive model which provides an estimate of the year recommended to replace the insulators. The model proposed is based on the assumption that the rate of corrosion deterioration increases proportionately with the severity of the environment. With the model, it is not necessary to establish the age of the insulators to estimate the recommended replacement year; it only requires the condition of the insulator strings on the line at the time of inspection and the environment in which the line section is located. The model proposed is given in Eq. 5-1:

  SR  RC   EC  RY  IY   ML    100    where RY = Recommended Replacement Year IY = Year of inspection ML = Maximum life of insulators in a clean environment = 50 years (refer Section 5.3) SR = Insulator string condition rating at the time of inspection = Arithmetic mean of the rating conditions of all insulators in the string RC = Minimum insulator string condition rating that requires replacement = 2 (refer Table 5-5) EC = Environmental code

130

Note that in the insulator replacement model, the replacement criteria for the insulator string is when its rating condition is poor (SR = 2). This is the same rating condition used in the insulator inspection FIS that suggests immediate insulator replacement. Such a condition warrants that the insulator be replaced immediately. For insulator string condition rating between 2 and 5, the insulator replacement model can be used to estimate the year it is recommended to be replaced. With regular monitoring, the accumulated numerical data from the proposed inspection method will make it possible to improve the estimate. Some numerical examples of the insulator replacement model, assuming last insulator inspection year is in 2003, are shown below to illustrate the application of the insulator replacement model: 1. Very good condition insulators (SR = 5) in a clean environment (EC = 10)

  5  2  10  RY  2003  50     2018  100   From the above calculation, it is recommended that the insulators be replaced in year 2018. 2. Fair condition insulators (SR = 3) in a mild environment (EC = 7)

  3  2  7  RY  2003  50     2007  100   It is recommended that the insulators be replaced in year 2007. 3. Good condition insulators (SR = 4) in an aggressive environment (EC = 3)

  4  2  3  RY  2003  50     2006  100   It is recommended that the insulators be replaced in year 2006 For long transmission lines that traverse different types of environment, it is possible to assign an environmental code to different sections of the line that are exposed to similar environments. For example, a transmission line may have a total of 250 towers from end to end. The first section of 50 towers may be located near the sea coast where the environment is very corrosive (EC = 2). The next 50 towers may run through a

131

mountainous and/or forest areas where the environment is relatively cleaner (EC = 9). The rest of the towers may be located near or within a residential area where the environment is harsher than the previous one (EC = 4). It is then possible to use the environmental codes and the insulator string condition ratings for the towers in the respective environments to estimate the year required to replace the insulators in each section. The insulator replacement model makes it possible for utilities to make budgetary plans for bulk replacement of insulators based on their condition ratings and the environment they are operating in. For long lines, it is possible to use the model to plan for staggered replacement of different sections of the line rather than replacing all the insulators of the line at one time. In this way, utilities are able to achieve significant savings in capital expenditure. 5.3 Application of the FIS on Other Transmission Line Components To effectively manage maintenance costs, normal utility practice during an inspection exercise is to check not just one component of the transmission line but all the components. For example, during a tower top inspection in TNB, the inspector is required to visually inspect all of the transmission line items in the inspection checklist which includes the foundations, the tower body, the cross arms, the conductors (including vibration dampers, spacers, and jumpers), the hardware fittings (such as clamps, yoke plates, and extension links), the earth wire and, of course, the insulators. Since the inspection is done visually, there exists the same kind of uncertainty that the inspector encounters when assessing these components. From the experience gathered from designing and applying the FIS to the visual inspection of porcelain insulators, the same design principles can be used to develop and apply a knowledge-based FIS on the other components of the transmission line. The system could take the structure outlined in Figure 5-38.

132

Components Insulators

Inputs

Inference Engine

Output

Porcelain Shell Breakage Pin Corrosion

Knowledge Base 1

Conductors Joint resistance Joint temperature

Knowledge Base 2

Knowledge Base 5

Structure Deformation

Tower Condition

Knowledge Base 3

Corrosion Foundations Corrosion Displacement

Knowledge Base 4

Figure 5-38: Proposed structure of tower inspection FIS The steps that need to be taken, which also exemplify the steps taken to develop the insulator inspection as described in this thesis, are as follows: 1. Identify the function(s) of the components and their failure modes. 2. Identify the indicators of failure for each component. For example, failure of the tower body can be indicated by how much deflected or bent a certain tower bracings appear and failure of a wooden cross arm can be determined by how severe the cracks on the wood surface appears. For failures that deteriorate with time, the P-F curve of the component should be determined so that the critical point P can be identified. 3. For each component condition, define numerical categories of defect based on how the different levels of defect can be visually detected. Either pictorial samples or linguistic descriptions of defect levels can be used. This also makes up the inspection guide that is used by the inspectors during inspection. 4. Based on the respective defect levels, develop membership functions for the inputs and outputs of the FIS.

133

5. Develop fuzzy IF-THEN rules and choose the appropriate defuzzification technique for the FIS. 6. Design a standard table of suggested maintenance actions. Steps 1 and 2 may require tests and experiments done on the component in order to determine the right deterioration mechanisms. The information gathered in these steps is then transferred to the knowledge-base of the FIS. Together with the inspection guide, the FIS can prove to be useful in the field during tower inspection exercise as not only it can significantly reduce subjectivity associated with the inspection process but also can make available expert knowledge at site. With the standard maintenance actions which are inferred from the output of the FIS, the inspectors can readily suggest the appropriate maintenance actions based on the defect condition of the components. 5.4 Potential Savings due to Introduction of FIS in Tenaga Nasional Berhad’s (TNB) Transmission Line Inspection and Maintenance Practice This section looks at the possible savings that can be achieved by Tenaga Nasional Berhad (TNB) due to the introduction of improved inspection process. The currency used is in Malaysian Ringgit (A$ 1 ~ RM 2.80) as data is from TNB. TNB’s transmission system spans the whole of Peninsular Malaysia, connecting power stations to load centers. The system operates at 132 kV, 275 kV and 500 kV voltage levels and forms an integrated network known as the National Grid. Overhead transmission lines currently make up 95.7% (16, 137 km) of the network [88]. TNB uses double-circuit steel lattice towers as the standard design for its transmission line. Double circuit lines are used to ensure good supply reliability (N-1 operation contingency) and to maximize land utilization for transmission line right-of-way. To protect the transmission lines from lightning strikes, TNB uses two earth wires strung at the top of the phase conductors. TNB also uses timber cross arms on a majority of its suspension towers to increase line insulation. To assess its transmission line operational performance, TNB uses a set of key performance indicators. One of the indicators used by TNB is the transmission line

134

outage rate. Figure A-2 provides a snapshot of TNB’s transmission line forced outage rate for the periods between 1998/1999 and 2002/2003 [88].

Figure 5-39: TNB transmission line forced outages 1997-2003 [88] It can be seen from Figure A-2 that the annualized 275 kV and 132 kV transmission line performance in 2002/2003 is 1.09 per 100 km-circuit and 1.03 per 100 km-circuit respectively. Most forced outages experienced by TNB have been due to bad weather or lighting strikes. This is not surprising because Malaysia is located in the tropics where the isokeraunic level is about 180 thunder days per year [89]. Each year TNB spends between RM 32 million and RM 37 million on inspection and maintenance of its transmission line [90]. This expenditure is due to current inspection and maintenance activities which include: 

Monthly routine ground patrol inspection for transmission lines located in accessible regions



3-yearly routine climbing inspection for all transmission lines



Yearly routine climbing inspection for lines that cross roads, highways, railways and rivers



Yearly helicopter inspection for transmission lines located in inaccessible regions



Bulk replacement of wooden cross arms and cap and pin insulators that have been installed for 25 years 135

As with many utilities worldwide, TNB has been working towards reducing its maintenance and inspection expenditure whilst maintaining or further reducing its outage rate. One way of achieving this objective is to reduce the frequency of inspection and maintenance based on the condition of the component. The improved inspection process using the FIS as proposed in this thesis provides a means for TNB to use a numerical rating scheme to quantify the condition of the component. Specific maintenance actions can therefore be made on those components that are bad and/or achieving its end of service life. Expenditures can therefore be channelled to components with such conditions rather than on the whole component population. The net effect is savings in inspection and maintenance expenditures. At the same time, information regarding the condition of the components that are collected using the FIS can be stored in a computer. As more data are collected, an analysis of the component deterioration trend can be made. The trends would enable TNB to make projections of future component replacement or repair activities thus taking better control of its maintenance budget. It is rather difficult to estimate the exact savings that can be achieved by introducing the new inspection process, but based on the current annual transmission line inspection and maintenance expenditure, a savings of 10% would translate to between RM 3 million to RM 4 million a year. With time, the savings could potentially be more significant as more data about component condition is available for future maintenance projections. 5.5 Chapter Summary In this chapter, we have primarily presented the development of a novel methodology that utilizes principles of fuzzy logic to handle uncertainties associated with visual inspection of porcelain cap and pin insulators. Current practices of two utilities namely TNB (Malaysia) and Powerlink Queensland (Australia) were discussed which indicated the need for such a methodology to overcome the problem of uncertainty. Together with an inspection guide, it has been shown that using the insulator inspection fuzzy inference system significantly reduces the subjectivity of assessing insulator pin rust

136

and porcelain shell conditions. The development of the fuzzy inference system, whereby the design of input/output membership functions and linguistic variables were utilized in the rule base, has also been explained. It has also been elucidated how the fuzzy inference system was used to determine the condition of three sample insulators taken from the field. It has been further shown that qualitative analysis based on linguistic indicators used during assessment of the insulators is converted to crisp numbers by the fuzzy inference system. The results of the output of the fuzzy inference system and how they can be used as a decision support for maintenance have also been discussed. The numerical rating of each insulator unit has made it possible to define the rating of an insulator string. It has been shown that the numerical rating of multiple insulator strings in a transmission line can be used in a mathematical model to estimate the appropriate replacement date for bulk replacements due to pin corrosion deterioration. Further in this chapter, we have also discussed the application of the FIS on visual assessment exercise of other components of the transmission line. Finally, we have discussed the possible savings that might be achievable if the new inspection methodology is implemented in TNB.

137

CHAPTER 6: CONCLUSIONS AND FUTURE WORK

6.1 Summary of the Research The main goal of the research presented in this thesis was to develop a novel methodology that would improve utility practice of visual defect assessment of transmission line components. In reaching this objective, this thesis seeks to answer the research question, “How can we reduce or remove the uncertainty associated with inspector visual assessment during field inspection so that the information gathered from the field inspection can be used effectively to manage maintenance actions?” The answer to this question is, as has been proposed in this thesis, a knowledge-based fuzzy inference system. This chapter presents an overall conclusion of the research and reflects on the objectives listed in Section 1.2. The first stage of the study was aimed at identifying the functions and understanding the failure modes of the major transmission line components. This was an essential part of the study because the reason utilities carry out visual inspection was to identify defects that could lead to imminent failure. In Chapter 2, it was highlighted that environmental factors such as wind, weather and atmospheric pollution have a major influence on failures of transmission line components. Except for very severe cases such as major storms with high incidental winds or earthquakes which could result in immediate transmission line failure, failures of transmission line components generally start with small defects followed by gradual deterioration. Failure, as defined in this thesis, is loss of function. In this regard, Chapter 2 presented the main functions of the major components and their associated failure modes. These are summarized in Table 6-1.

138

Component

MAIN FunctionS

Failure Modes

Steel

Withstand static and

Buckling or collapse due to structural

Tower/Structure

dynamic forces

failure made worse by effects of corrosion

Foundations

Transfer mechanical loads

Foundation failure due to corrosion

to ground for stability

of steel reinforcement bar Excessive soil movement

Conductors Insulators

Carry rated current while

Conductor snap due to corrosion,

maintaining even sag

vibration fatigue or annealing

Provide mechanical

Broken pin due to corrosion or metal

support

fatigue

Withstand electrical

Damaged insulator shells

stresses Table 6-1: Summary of transmission line components, their functions and failure modes As shown in Table 6-1, a majority of the defects is due to deterioration of steel components as a result of corrosion. Having understood the functions of transmission line components and their failure modes, the second phase of the study was to review the available methods for detecting these defects from the various sources of literature and the third chapter includes information about various types of component monitoring systems. It was found that, depending on the component and its functions, a majority of these methods utilized equipments that were able to detect either electrical or physical defects on transmission lines. It was also found that, among the major components, the insulator has the greatest number of different equipments available determination of its condition. However, these equipments were either found to be expensive (i.e. the Daylight Corona camera), still in the development stages (i.e. analysis of sample polymeric insulator sheds using Fourier transform infrared spectroscope) or highly affected by the environment (i.e. the thermovision camera or corona phone). Chapter three also provided a review of inspection and maintenance practices of utilities worldwide, which was the third objective of the research. For this purpose, the results of

139

two industry surveys were analysed: one was conducted by CIGRE [1] on 90 utilities worldwide and the other by McMahon [2] on utilities in Australia and New Zealand. The CIGRE survey indicated that utilities regard the most typical form of defect is corrosion attack on steel components. However, the significance of both the survey results to this research is that, despite the advent of new monitoring equipments, both the survey results indicated that the most widely used utility practice for location of defects on transmission lines is by visual inspection. The method requires inspectors to visually inspect the component either from the air (in helicopters), on the ground (using visual aids such as binoculars or zoom cameras) or at the top of the tower (by climbing). The main problem with visual inspection, as discussed towards the end of the chapter three, is that there exists a high level of subjectivity and uncertainty when the inspectors are making the defect evaluation. Another problem is that because the defects are recognized by their perceived physical conditions, defect indicators are normally reported qualitatively using vague and linguistic terms such as “medium crack”, “heavy rust”, “small deflection” and “large breakage”. Unfortunately, inspectors are human beings and not machines or robots that can be programmed to produce accurate results all the time. As a consequence, there is a large variance in defect reporting (which, in time, makes it difficult for utilities to monitor the condition of the component) that can lead to wrong or ineffective maintenance decisions. This thesis recommends solving this problem by using the artificial intelligence technique of fuzzy logic, a technique which was first introduced in the 1960’s to deal with the fuzziness of human judgment. The aim of chapter four in this thesis was to provide an understanding of the principles of fuzzy logic for it to be applied in this research. In this chapter, the main building blocks of fuzzy logic, namely the construction of membership functions and IF-THEN rules, were discussed. It was shown in the chapter how the combination of membership functions that represent linguistic variables and IF-THEN rules were implemented in a knowledgebased Fuzzy Inference System (FIS), one of the ways by which fuzzy logic is used in realworld applications. To assist in understanding further the subject matter, several applications of fuzzy logic in the utility environment were also discussed in this chapter. The fifth and main objective of this research was to design, apply and test a knowledgebased FIS on the utility visual inspection of one of the most important components of the

140

transmission line – the porcelain cap and pin insulators. For this purpose, chapter five examined two of the most common failure mechanisms of porcelain cap and pin insulators: corrosion of the steel pin and breakage of the porcelain shells. Pin rust increases gradually with time and the rate of deterioration is affected by the environment wherein the insulator is located. Breakage of porcelain shells is mostly due to vandalism, does not increase with time and normally occurs on lines that are located in areas accessible to the public. Both types of defects are normally detected visually by utilities during routine inspections. For the purpose of this research, current insulator inspection practice of two utilities, Powerlink Queensland (Australia) and Tenaga Nasional Berhad Malaysia (TNB), was studied. Table 6-2 provides a summary comparison of the inspection methods of both utilities. TNB

powerlink

 Visual inspection is done by

 Visual inspection is done by

linesmen during routine line

linesmen during routine live line

outage for inspection

inspection

 No inspection guides are used

 For rust conditions, a 4-category

 Inspection results are recorded

rust in inspection guide is used (scale: 1-no rust to 4-heavy rust)

in inspection form: (√) means good – no action, (x) means bad

 Maintenance decision is based

– replace immediately; anything

on the reported condition of the

in between is recorded in

insulator referred to one of four

separate forms

defect categories

 Maintenance is done based on

 Bulk replacement is taken if

the reported condition of the

necessary

insulator  Bulk replacement is sometimes taken if the same defects are detected on many insulators Table 6-2: Summary of TNB and Powerlink porcelain cap and pin insulator inspection practice It was deduced that the TNB practice results in high level of variance in defect reporting due to the linesmen not using any inspection guides when conducting the inspection. 141

Powerlink practice narrows down the variance level but there is still uncertainty in the defect evaluation because the actual perceived condition of the insulator may not be exactly the same as described in the inspection guide. Chapter five then proceeded with the design of a knowledge-based FIS for use during insulator visual inspection. The first step shown was to develop defect rating tables based on the linguistic description of the severity of porcelain shell breakage (see Table 5-1) and pin rust (see Table 5-2). Using Matlab’s proprietary Fuzzy Logic Toolbox program, these defect ratings were then converted to triangular membership functions. The overall condition ratings of the insulator were also represented by membership functions. In the FIS, the relationship between the inputs (pin rust condition and porcelain shell breakage condition membership functions) and the output (insulator overall condition membership function) was represented by IF-THEN rules. For this purpose, a table of twenty IFTHEN rules (see Table 5-3) was then created to infer the overall condition of the insulator based on its pin rust and porcelain shell conditions respectively. Mamdani’s Center-of–Area (COA) defuzzification method was used in the FIS, where the overall condition of the insulator was inferred by calculating the point which is central to the area under the aggregated output membership function. Finally, the output of the FIS was cross-referenced to a table of suggested maintenance action (see Table 5-4) to guide the user to either “do nothing”, “flag for next maintenance cycle”, “increase monitoring frequency” or “replace immediately”. The insulator inspection FIS was then tested on three sample insulators taken from the field. Experience from the three sample applications showed that: 

together with an insulator inspection guide for rusty pins and broken porcelain shells, the insulator inspection FIS assisted visual evaluation of defective insulator units on the string by inferring the defect level of each insulator with respect to its physical appearance and condition.



the inspector was only required to indicate the appearance of rust at the insulator pin and chipping/breakage of the porcelain shell by comparing these conditions with the condition ratings used in Tables 5-1 and 5-2 respectively.



inference was undertaken by the Inference Engine of the FIS and not by the inspector thereby relieving the inspector from making any judgment with regards

142

to the overall condition of the insulator. This greatly reduced subjectivity and uncertainty when evaluating insulator defects. 

because the output of the FIS was referred to standardized maintenance actions (see Table 5-4), consistency in decision-making for maintenance actions was therefore possible.

Chapter five also showed how the output of the insulator inspection FIS was used in a mathematical model that was developed to assist maintenance managers in making bulk replacement of rusty porcelain cap and pin insulators. The assumptions used in the model were that the rate of pin rust deterioration over time follows a linear relationship and that the rate of pin rust deterioration is higher in areas of aggressive environments. For this intention, an environmental rating condition based on the International Electrotechnical Commission’s (IEC) definition of pollution severity was created. Application of the model on several combinations of insulator and environmental conditions showed the model’s usability as a tool for managing insulator assets, thus achieving the sixth and final objective of this thesis. As finally discussed in chapter five, the ideas used during the design of the insulator inspection FIS can be applied towards the design and implementation of a transmission line inspection FIS which takes into account the visual assessment of other transmission line components namely the tower structure, the conductors and the foundations. True to this thesis’s title and research objectives, thus it has been shown “The Application of Knowledge-based Fuzzy Inference System on High Voltage Transmission Line Maintenance”. 6.2 Major Research Contributions Several benefits of the FIS include: 

Together with visual inspection guide, FIS can effectively reduce the level of uncertainty when assessing transmission line defects. This results in a more objective and consistent evaluation of defects as well as provides support to making maintenance decisions.

143



FIS works by transforming fuzzy qualitative indicators that are prevalent when assessing defects visually to crisp numerical values that can be used as defect rating.



Numerical values make it possible to represent transmission line component defect condition in computer database for trending and future maintenance strategies.



Numerical values can be used in a replacement model to plan for bulk replacement of transmission line components.



Programmed into mobile computing devices, FIS makes available expert knowledge at site



FIS can be fine-tuned by adjusting membership functions and IF-THEN rules based on available expert knowledge.

6.3 Future Work The following are recommendations for further work on this research topic, which in the author’s opinion, will prove to be particularly fruitful: 

Examining the effect of using different membership function shapes such as trapezoidal, Gaussian or sigmoid on the output of the FIS.



Further reducing the level of uncertainty associated with visual inspection of defects. The use of a pictorial representation of defect levels in the inspection guide (rather than descriptions of defect as used in this research) may help achieve this objective.



Examining the sensitivity of the knowledge-based FIS by using more defect levels in the inspection guide. However, this would entail more rules being defined in the FIS inference engine.



Incorporating the Matlab codes in mobile computing device platform for practical implementation by field inspectors.

144

Bibliographical References [1] [2] [3]

[4] [5] [6] [7] [8] [9] [10] [11]

[12] [13] [14] [15]

[16]

[17]

[18] [19]

R. Paschen, J. Rogier, D. Hughes, B. Rassineux, and J. B. G. F. d. Silva, "Assessment of existing overhead line supports," in ELECTRA: CIGRE, April 2003. B. McMahon, "Reliability and maintenance practices for Australian and New Zealand HV transmission lines," presented at Second International Conference on Reliability of Transmission and Distribution Equipment, 1995. G. Jaensch, H. Hoffmann, and A. Markees, "Locating Defects in High Voltage Transmission Lines," presented at IEEE 8th International Conference on Transmission & Distribution Construction, Operation & Live-Line Maintenance, 1998. L. A. Zadeh, "Fuzzy Sets," Information & Control, vol. 8, pp. 338-353, 1965. H. M. Ryan, High Voltage Engineering and Testing, 2nd ed. London, United Kingdom: The Institution of Electrical Engineers, UK, 2001. M. F. Ishac and H. B. White, "Effect of tornado loads on transmission lines," Power Delivery, IEEE Transactions on, vol. 10, pp. 445-451, 1995. T. Gillespie, "Overhead Line Design - Electrical," in Postgraduate Course in Electrical Supply Engineering. Brisbane: Queensland University of Technology, 2002. G. Karady, "Transmission Lines," http://www.eas.asu.edu/~karady/360_stuff/Lectures/5, 2003. BS EN ISO 1461:1999 Hot dip galvanized coatings on fabricated iron and steel articles. Specifications and test methods: British Standards International, 1999. D. Douglass, "New Overhead Conductors for Higher Rating & Reliability," 2003. G. J. Oberst, Jr., "Lattice tower ground line corrosion and mitigation; a case study," presented at IEEE 8th International Conference on Transmission & Distribution Construction, Operation & Live-Line Maintenance Proceedings, 1998. C. Bayliss, Transmission and Distribution Electrical Engineering, 2nd ed. Manchester: Newnes, 2001. J. P. Broomfield, Corrosion of Steel in Concrete. London: E & FN Spon, 1997. P. Dimitrakakis, "Corrosion of Steel-Reinforced Concrete Poles," presented at Solving Corrosion Problems in Reinforced Concrete Symposium, Parkville, Victoria, 1986. R. J. T. Schweiner, K.E.; Lindsey, K.E., "Transmission line emergency restoration philosophy at los angeles department of water and power," presented at Transmission and Distribution Construction, Operation and Live-Line Maintenance, 2003. 2003 IEEE ESMO. 2003 IEEE 10th International Conference on, 2003. J. C. Pohlman, K. E. Lindsey, and R. F. Corpuz, "Controlling the economic risk from catastrophic failure of overhead transmission lines," presented at IEEE 8th International Conference on Transmission & Distribution Construction, Operation & Live-Line Maintenance, 1996. L. N. E. Agrawal, "Planning and training reduce restoration time for damaged transmission lines in India," presented at IEEE 9th International Conference on Transmission and Distribution Construction, Operation and Live-line Maintenance, 2000. "ACSR Aluminum Conductor, Steel Reinforced, Bare," 2003. C. Lee and G. Mulherrin, "Overhead Line Design - Mechanical," in Postgraduate Course in Electricity Supply Engineering. Brisbane: Queensland University of Technology, 2002. 145

[20]

[21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35]

[36] [37] [38]

[39]

I. M. Zamora, A.J.; Criado, R.; Alonso, C.; Saenz, J.R., "Uprating using hightemperature electrical conductors [overhead power lines]," in Electricity Distribution, 2001. Part 1: Contributions. CIRED. 16th International Conference and Exhibition on (IEE Conf. Publ No. 482), vol. 1, 2001, pp. 5 pp. vol.1. G. Orawski, "Overhead lines-the state of the art," Power Engineering Journal, vol. 7, pp. 221-231, 1993. "ELF Electromagnetic Fields and the Risk of Cancer," Advisory Group on NonIonizing Radiation, National Radiological Protection Board,UK, Oxon 2001. D. Hafmeister, "Background Paper on Power Line Fields and Public Health," www.calpoly.edu/~dhafemei/background2.html. D. R. Shannon, "Life Expectancy of ACSR Conductors under Live-Line and OffLine Conditions," http://www.shannontechnology.com/corrosion.html, 2002. W. D. Marshall and T. E. Jacobs, "Transmission Line Conductor Joint Resistance's Measured Live Line (Live-Line Measurement Project at Transpower New Zealand)," www.pls.co.nz/resources/joints.pdf July 16, 1997. "Grid Services GmbH: Defects [3]," Grid Services GmbH, 2002. "Safe Design Tension with repsect to Aeolian Vibrations. Part 1: Single Unprotected Conductors," in Electra, October 1999, pp. 186. R. R. Gibbon, P. H. Juul, H. B. White, and W. J. Wijker, "Damage to Overhead Lines due to Conductor Gallopping," in ELECTRA, 1984. C. Lee, "Reliability Assessment of Transmission Line Components," in School of Electrical & Electronics Systems Engineering. Brisbane: Queensland University of Technology, 1996, pp. 179. G. d. Troia, "Effects of High Temperature Operation on Overhead Transmission Full Tension Joints and Conductors," SC22 WG12 CIGRE, 2000. R. S. Gorur, E. A. Cherney, and J. T. Burnham, Outdoor Insulators: Ravi S. Gorur, Inc., 1999. B. Jubb, "Current overseas practice with transmission at 132 kV and above," presented at IEE Colloquium on Review of Outdoor Insulation Materials, 1996. IEEE Standard 100, The New IEEE Standard Dictionary of Electrical and Electronic Terms: IEEE, 1996. H. M. Schneider, J. F. Hall, G. Karady, and J. Renowden, "Nonceramic insulators for transmission lines," IEEE Transactions on Power Delivery, vol. 4, pp. 2214-2221, 1989. T. Kikuchi, S. Nishimura, M. Nagao, K. Izumi, Y. Kubota, and M. Sakata, "Survey on the use of non-ceramic composite insulators," Dielectrics and Electrical Insulation, IEEE Transactions on [see also Electrical Insulation, IEEE Transactions on], vol. 6, pp. 548-556, 1999. A. Phillips, "Ceramic vs. Polymer (Non-Ceramic Insulators)," presented at 1st Annual Overhead Transmission Line Equipment, Inspection & Maintenance Practices Conference, Monterey, California, USA, 2002. K. Morita, T. Imakoma, and Y. Suzuki, "Hardware Corrosion of Suspension Insulators on Overhead Transmission Lines," presented at 6th Conference on Electricity Power Supply Industry (CEPSI), Jakarta, 1986. H. Dietz, H. Karner, K. H. Muller, H. Patrunky, H. J. Voss, P. Verma, and G. Schenk, "Latest developments and experience with composite longrod insulators," presented at International Conference on Large High Voltage Electric Systems, Paris, France, 1986. M. J. Geramian, Z. Hamnabard, M. Mohseni, and S. Dolatshahi, "An Assessment on Brittle Fracture of Composite Insulators' Rod and Factors Improving their

146

[40]

[41]

[42] [43] [44] [45] [46] [47] [48]

[49] [50] [51] [52]

[53]

[54]

[55]

Properties," presented at 16th International Conference and Exhibition on Electricity Distribution, 2001. J. Montesinos, R. S. Gorur, B. Mobasher, and D. Kingsbury, "Mechanism of brittle fracture in nonceramic insulators," in Dielectrics and Electrical Insulation, IEEE Transactions on [see also Electrical Insulation, IEEE Transactions on], vol. 9: Experimental, 2002, pp. 236-243. J. T. Burnham, T. Baker, A. Bernstorf, C. d. Tourreil, J. George, R. Gorur, R. Hartings, B. Hill, A. Jagtiani, and D. Mitchell, "IEEE Task Force Report: Brittle Fracture in Nonceramic Insulators," IEEE Transactions On Power Delivery, vol. 17, pp. 848-856, 2002. R. Gorur, "Condition Assessment of Polymer Insulators," 2003. J. T. Burnham and R. J. Waidelich, "Gunshot Damage to Ceramic and Nonceramic Insulators," IEEE Transactions On Power Delivery, vol. 12, pp. 16511656, 1997. M. Vitelli, V. Tucci, and C. Petrarca, "Temperature Distribution along an Outdoor Insulator Subjected to Different Pollution Levels," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 7, 2000. R. S. Matsuoka, H.; Kondo, K.; Gorur, R.S., "Investigation of field energized RTV coated porcelain insulators," presented at Electrical Insulating Materials, 1995. International Symposium on, 1995. J. S. T. Looms, Insulators for High Voltages, 1st ed. London: Peter Peregrinus Ltd., 1988. T. L. Landers, R. J. Richeda, E. Krizanskas, J. R. Stewart, and R. A. Brown, "High phase order economics: constructing a new transmission line," Power Delivery, IEEE Transactions on, vol. 13, pp. 1521-1526, 1998. L. Bauer, P. Ulardic, and J. Muller, "Reinforcement strategies for extending the service life of 110 kV overhead transmission lines," presented at Electricity Distribution. Part 1. Contributions. 14th International Conference and Exhibition on (IEE Conf. Publ. No. 438), 1997. "Havard Engineering Inc Webpage: Towertest," 2003. J. Duxbury, "Foundation Steel Corrosion Inspection & Repair at BCHydro," presented at 1st Annual Overhead Transmission Line Equipment, Inspection & Maintenance Practices Conference, Monterey, California, USA, 2002. "ASTM C876: Standard Test Method for Half-Cell Potentials of uncoated Reinforcing Steel in Concrete," American Society of Testing and Materials 1999. D. G. Havard, G. Bellamy, P. G. Buchan, H. A. Ewing, D. J. Horrocks, S. G. Krishnasamy, J. Motlis, and K. S. Yoshiki-Gravelsins, "Aged ACSR conductors. I. Testing procedures for conductors and line items," presented at Transmission and Distribution Conference, 1991., Proceedings of the 1991 IEEE Power Engineering Society, 1991. J. Snell and J. Renowden, "Improving results of thermographic inspections of electrical transmission and distribution lines," in Transmission and Distribution Construction, Operation and Live-Line Maintenance Proceedings. 2000 IEEE ESMO 2000 IEEE 9th International Conference on, 2000, pp. 135-144. D. G. Havard, M. K. Bissada, C. G. Fajardo, D. J. Horrocks, J. R. Meale, J. Y. Motlis, M. Tabatabai, and K. S. Yoshiki-Gravelsins, "Aged ACSR conductors. II. Prediction of remaining life," Power Delivery, IEEE Transactions on, vol. 7, pp. 588595, 1992. M. Ostendorp, "Inspection, Assessment, and End-of-Service-Life Guidelines for Porcelain Cap and Pin Insulators," presented at 1st Annual Overhead

147

[56] [57]

[58]

[59] [60]

[61] [62] [63] [64] [65]

[66] [67] [68]

[69] [70] [71] [72] [73]

Transmission Line Equipment, Inspection & Maintenance Practices Conference, Monterey, California, USA, 2002. G. H. Vaillancourt, J. P. Bellerive, M. St-Jean, and C. Jean, "New Live Line Tester for Porcelain Suspension Insulators on High-Voltage Power Lines," IEEE Transactions On Power Delivery, vol. 9, pp. 208-219, 1994. G. H. Vaillancourt, S. Carignan, and C. Jean, "Experience with the detection of faulty composite insulators on high-voltage power lines by the electric field measurement method," IEEE Transactions on Power Delivery, vol. 13, pp. 661-666, 1998. D. H. Shaffner, D. L. Ruff, and G. H. Vaillancourt, "Experience with a composite insulator testing instrument based on the electric field method," presented at Transmission and Distribution Construction, Operation and Live-Line Maintenance Proceedings. 2000 IEEE ESMO - 2000 IEEE 9th International Conference on, 2000. M. Lindner, S. Elstein, P. Lindner, J. M. Topaz, and A. J. Phillips, "Daylight Corona Discharge Imager," presented at High Voltage Engineering Symposium, 1999. W. L. Broskey and A. Phillips, "Implementation of Daytime Corona Inspection at Allegheney Power," presented at 1st Annual Overhead Transmission Line Equipment, Inspection & Maintenance Practices Conference, Monterey, California, USA, 2002. S. Shihab and K. L. Wong, "Detection of faulty components on power lines using radio frequency signatures and signal processing techniques," presented at Power Engineering Society Winter Meeting, 2000. IEEE, 2000. A. Krivda, G. Cash, D. Birtwhistle, and G. George, "Condition monitoring of EPDM polymer insulators," presented at Eleventh International Symposium on High Voltage Engineering, 1999. Y. C. Cheng, L. Ding, C. R. Li, and X. H. Ma, "Online detecting insulator corona for diagnosing faulty insulators on transmission line," IEEE 2001. C. R. Li and Y. C. Cheng, "A technique of detecting faulty insulator strings on ground," presented at High Voltage Engineering Symposium, 1999. I. Ramirez-Vazquez and J. L. Fierro-Chavez, "Criteria for the Diagnostic of Polluted Ceramic Insulators Based on the Leakage Current Monitoring Technique," presented at IEEE Conference on Electrical Insulation and Dielectric Phenomena, 1999. J. Moubray, Reliability-centered Maintenance: Industrial Press Inc., 1992. A. Stewart, "Airborne Inspection Technology: Market Survey," presented at 1st Annual Overhead Transmission Line Equipment, Inspection & Maintenance Practices Conference, Monterey, California, USA, 2002. M. Ostendorp, "Innovative Airborne Inventory and Inspection Technology for Electric Power Line Condition Assessments and Defect Reporting," presented at IEEE 9th International Conference on Transmission and Distribution Construction, Operation and Live-Line Maintenance Proceedings, 2000. "Services: Haverfield Air Service Team," 2000. "Services: Haverfield Air Service Team," 2000. K. S. L. Leung, W., "Fuzzy concepts in expert systems," Computer, vol. 21, pp. 4356, 1988. G. Klir and B. Yuan, Fuzzy Sets and Fuzzy Logic: Theory and Applications. Upper Saddle River, New Jersey: Prentice Hall Inc., 1995. Fuzzy Logic Toolbox for use with Matlab. Natick, MA: The MathWorks Inc., 1998.

148

[74] [75] [76]

[77] [78] [79] [80] [81] [82]

[83] [84] [85] [86] [87] [88] [89] [90]

D. D. Bouchaffra, "CSE 513 Soft Computing (Ch. 4 Fuzzy Inference System)," 2002. P. Kumar, M. Jamil, M. S. Thomas, and Moinuddin, "Fuzzy approach to fault classification for transmission line protection," presented at TENCON 99. Proceedings of the IEEE Region 10 Conference, 1999. I. Hathout, "Soft reliability assessment of existing transmission lines," presented at Proceedings of ISUMA - NAFIPS '95 The Third International Symposium on Uncertainty Modeling and Analysis and Annual Conference of the North American Fuzzy Information Processing Society, 1995. P. Marannino, A. Berizzi, M. Merlo, and G. Demartini, "A rule-based fuzzy logic approach for the voltage collapse risk classification," presented at Power Engineering Society Winter Meeting, 2002. IEEE, 2002. S. W. Mofizul Islam, T.; Ledwich, G., "A novel fuzzy logic approach to transformer fault diagnosis," Dielectrics and Electrical Insulation, IEEE Transactions on [see also Electrical Insulation, IEEE Transactions on], vol. 7, pp. 177-186, 2000. "Line Life, Component Condition and Utility Actions," report by CIGRE SC22 WG13, August 1994. "Iran Insulator - Standard suspension insulator Ball & Socket," in Standard Suspension Insulator Ball & Socket: http://www.iraninsulator.com/pages/201.html, 2002. "Iran Insulator - Anti-fog suspension insulator Ball & Socket," in Anti-fog Suspension Insulator Ball & Socket: http://www.iraninsulator.com/pages/206.html, 2002. A. S. Jagtiani and J. R. Booker, "Aging of porcelain suspension insulators under mechanical and electrical stress on EHV AC lines," presented at Transmission and Distribution Construction and Live Line Maintenance, 1995. ESMO-95 Proceedings., Seventh International Conference on, 1995. J. M. Mendel, Uncertain Rule-Based Fuzzy Logic Systems: Introduction and New Directions. New Jersey: Prentice Hall Inc., 2001. V. D. Hunt, Artificial Intelligence & Expert Systems Sourcebook. New York: Chapman & Hall, 1986. T. Gillespie, "Personal communication." Powerlink, Brisbane, Queensland, 2003. IEC 60518: Guide for the Selection of Insulation in Respect of Polluted Conditions: InternationaI Electrotechnical Committee, 1986. W. Marshall, "Condition Assessment Linked to Predictive Maintenance Modeling: A Valuable Management Tool," presented at ESMO, Orlando, 1998. M. Z. Meah, "TNB Transmission Division Business Plan (FY2004-FY2008)," Tenaga Nasional Berhad, Kuala Lumpur 2003. A. Halim, A. Bakar, and S. Imai, "Auto-reclose performance on 275 kV and 132 kV transmission line in Malaysia," presented at Transmission and Distribution Conference and Exhibition 2002: Asia Pacific. IEEE/PES, 2002. R. Othman, "E-mail Communication." Tenaga Nasional Berhad, Kuala Lumpur, Malaysia, 31/7/2003.

149