DEVELOPMENT OF DECISION MODEL FOR MAINTENANCE ANALYSIS OF

NON-REPAIRABLE COMPONENT BY CONSIDERING THE EXTERNAL

FACTOR

by

ROSMAINI BIN AHMAD

Thesis submitted in fulfilment of the

requirements for the degree
of Master of Science

JULY, 2007
 ACKNOWLEDGEMENTS

BISMILLAHIRRAHMANIRRAHIM

In the name of Allah, the most graceful and merciful. Alhamdulillah, after one and a half

years of effort, I have managed to complete this dissertation. This research has been

carried out in August 2005 till March 2007, at the School of Mechanical Engineering,

Universiti Sains Malaysia (USM). It has been supported by many people. It is a

pleasure to express my gratitude to all these people.

First of all, I would like to express my profound gratitude to my supervisor, Dr. Shahrul

Bin Kamaruddin for his excellent supervision, motivation, ideas, guides and continuous

support during the research.

I am also thankful to Assoc. Prof. Ahmad Shukri Yahya, the lecturer of School of Civil

Engineering for tutoring me on the statistical and mathematical programming. I would

also like to show my appreciation to Mr. T:." Chee Seng, the Assistant General

Manager of Nibong Tebal Paper Mill Sdn. Bhd for providing me the necessary

information and related data for my case study purpose. I wish to express my gratitude

to my beloved parents, Ahmad Bin Ariffin and Siti Rohani Binti Hamad for their

continuous blessing; support, kindness and encouragement. Also thanks to my

postgraduate friends at School of Mechanical Engineering who have assisted me in my

efforts towards completing this research.

Last but not least, I would like to thank to USM for funding part of my study under the

short term grant and to those whom I have not mentioned their names but they have

supported, encouraged and helped me either directly or indirectly in completing my

research.
 TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ii
TABLE OF CONTENTS iii
LIST OF TABLES vi
LIST OF FIGURES vii
LIST OF SYMBOLS viii
LIST OF ABBREVIATION ix
LIST OF APPENDICES x
LIST OF PUBLICATIONS & SEMINARS xi
ABSTRAK xii
ABSTRACT xiv

CHAPTER ONE: INTRODUCTON 1

1.1 Research Background 1
1.2 Problem Statement 3
1.3 Research Objectives 4
1.4 Research Focus and Delimitation 5
1.5 Thesis Overview 5

CHAPTER TWO: LITERATURE REVIEW 8

2.1 Machine Characteristics 8
2.2 The Factor of Machine Component Failure 9
2.2.1 Internal Factor 10
2.2.2 External Factor 11
2.3 Maintenance Strategy 13
2.3.1 Corrective Maintenance Strategy 13
2.3.2 Preventive Maintenance Strategy 14
2.3.3 Preventive Maintenance Application 16
2.3.4 Inspection Model 17
2.3.5 Preventive Replacement Model 18
2.4 Reliability Issue 20
2.4.1 Reliability Models 22
2.4.2 Proportional Hazard Model (PHM) 24

iii
2.5 Literature Findings 27
2.6 Conclusion 28

CHAPTER THREE: DECISION MODEL DEVELOPMENT 29

3.1 Overview of the Decision Model 29
3.1.1 General Structure of the Decision Model 30
3.2 Pareto Analysis (PA) 32
3.3 Physical Failure Analysis (PFA) 32
3.4 Covariate Analysis (CA) 35
3.5 Preventive Maintenance (PM) Modelling 38
3.6 Data Collection Techniques 42
3.7 Conclusion 44

CHAPTER FOUR : RESULT AND DISCUSSION 45

4.1 Case Study Background 45
4.1.1 Production Line 45
4.2 Critical Component Identification 48
External Factors Identification and Censored and Uncensored
4.3
Data Classification 50
4.4 Estimation of External Factors Effect 56
4.5 Determination of Preventive Replacement Time 61
4.5.1 Failure Time Process Test 62
4.5.2 Fitting the Failure Time Distribution 65
4.5.3 Optimum Time of Preventive Replacement 68
4.6 Decision Model Discussion 70
4.7 Conclusion 72

CHAPTER FIVE: CONCLUSION 73

5.1 Concluding Remarks 73
5.2 Future Work 76

iv
REFERENCES 77

PUBLICATIONS & SEMINARS PAPERS 83

APPENDICES
Appendix A: Failure time record for Cutting blade and Transmission A- 1
belt

Appendix B : Data arrangement for Failure Time Process Test (FTPT) B-1
- Cutting blade data

Appendix C : Data arrangement for Failure Time Process Test (FTPT) C- 1

- Transmission belt data

Appendix D : Maintenance cost record (replacement cost) for machine D- 1

types and components failure

Appendix E : Programming of Preventive Replacement Model using E- 1

MATLAB software

v
 LIST OF TABLES

Page

2.1 The failure time function 21

3.1 The sources of data collection and their advantages and 43
disadvantages

4.1 Failure time of transmission belt 51

4.2 Failure time of cutting blade 51

4.3 Physical Failure Analysis (PFA) using FMECA technique 53
4.4 Failure time data of transmission belts and codified values of
influencing covariates 57

4.5 Failure time data of cutting blade and codified values of

influencing covariate 58

4.6 Failure time (Time-Between-Failure) and censored &

uncensored data of cutting blade and transmission belt 65

4.7 e
Comparison of scale, and shape, f3 parameters with
covariates effect and without covariate effect 67

4.8 Standard optimal time, Ts of replacement 69

vi
 LIST OF FIGURES

Page

2.1 Flow of the literature review 7

2.2 The types of components in a machine system 8

2.3 Serial configuration and Parallel configuration 9

2.4 The internal and external factors influence on the machine's 10

component failure

2.5 Maintenance strategy classification 13

2.6 Determination the optimum time of PM by minimising the failure 16

cost criteria

2.7 Effects of covariates on hazard rate of the component or system 25

3.1 The general structure of the decision model 31

3.2 The process of PFA in the decision model by using FMECA 33

3.3 The process of CA in the decision model 35

3.4 The process of PM modelling in the decision model 38

3.5 Possible exploratory steps in field failure data analysis before 39

fitting distribution model

4.1 Production line in the company 47

4.2(a) Pareto analysis of the machines in production line 49

4.2(b) Pareto analysis of the component failure in machine Log saw 5 49

4.3 Estimation of covariates on transmission belt failure using PHM 60

4.4 Estimation of covariates on cutting blade failure using PHM 60

4.5 Trend test for transmission belt: No trend 63

4.6 Trend test for cutting blade: No trend 63

4.7 Serial correlation test for transmission belt: No correlation 64

4.8 Serial correlation test for cutting blade: No correlation 64

vii
 LIST OF SYMBOLS

F(t) Cumulative distribution function

a Row vector of the covariates

CrT) Cost function at time, T

C(Ts) Standard cost function at standard time, Ts

Cf Cost of failure replacement

Ck the critically index for failure mode k

Cp Cost of preventive replacement

exp Exponent

f(t) Probability density function

R(t) Reliability function

T Failure time I Optimum time of replacement

t Duration of time used in the FMECA

Ts Standard optimal time of Preventive Replacement

x Column vector of the regression parameters (covariate parameter)

akp The fraction of the component p's failure having failure mode k (that
is, the conditional probability of failure mode k given component p has
failed)

13 Shape parameter of Weibull distribution model

13k The conditional probability that failure mode k will result in the
identified failure effect

e Scale parameter of Weibull distribution model

eo Actual scale parameter of Weibull distribution model by considering

the covariate effect

es Standard scale parameter

A(t) Based line failure rate I hazard rate function

Ao(t) Actual (observed) failure rate I hazard rate function

Ap the failure rate of component p

p Cost ration of Cf / Cp

viii
 LIST OF ABBREVIATION

ARM Age Replacement Model

BRM Block Replacement Model

CA Covariate Analysis

CBM Condition Based Maintenance

CDF Cumulative Distribution Function

CM Corrective Maintenance

FFTD Fitting Failure Time Distribution

FMECA Failure Mode Effect and Criticality Analysis

FTPT Failure Time Process Test

LS Log Saw

PA Pareto Analysis

PDF Probability Density Function

PFA Physical Failure Analysis

PHM Proportional Hazard Model

PM Preventive Maintenance

PR Preventive Replacement

RP Roll Pack

TBF Time Between Failure

TBM Time Based Maintenance

ix
 LIST OF APPENDICES

Appendix - A Failure time record for Cutting blade and Transmission belt

Appendix - B Data arrangement for Failure Time Process Test (FTPT) -

Cutting blade data

Appendix - C Data arrangement for Failure Time Process Test (FTPT) -

Transmission belt data

Appendix - D Maintenance cost record (replacement cost) for machine types

and components failure

Appendix - E Programming of Preventive Replacement Model using MATLAB

software

x
 LIST OF PUBLICATIONS & SEMINARS

1.1 Implementation of dust control system using management and

planning tools (MPT) - A case study

1.2 An investigation on the impact of dust particles on machine's

component failure

1.3 The characteristics of dust particles emitted from the processing

industries and their physical impact on machine components: A
review

1.4 Development of a framework for the relationship of machine's

performance and the impact of dust pollution

1.5 The impact of dust particles on machine's reliability and preventive

maintenance time

1.6 The effect of repair and replacement time for the machine's
component exposed to dust particles

1.7 Identifying the Best Fit Failure Distribution and the Parameters of
Machine's Component: A New Approach

1.8 The Application of Preventive Replacement Strategy on Machine

Component in Deteriorating Condition - A Case Study in the
Processing Industries

1.9 Analysis of failure data from machines production line - Case study in
automotive industry

1.10 Maximum machines availability via inspection policy - Case study in

automotive industry

xi
 PEMBANGUNAN MODEL KEPUTUSAN UNTUK ANALISIS
PENYElENGGARAAN BAGI KOMPONEN YANG TIDAK BOlEH
DIPERBAIKI DENGAN MENGAMBllKIRA FAKTOR lUARAN

ABSTRAK

Penyelenggaraan pencegahan (Preventive Maintenance - PM) merupakan

salah satu strategi yang digunakan untuk mengurangkan masalah kerosakan mesin

yang berlaku secara tiba-tiba. Walaupun begitu, persoalan bilakah masa yang terbaik

untuk menjalankan aktiviti PM merupakan isu terpenting dalam pengaplikasian strategi

ini. Jawapan untuk persoalan tersebut ialah dengan berdasarkan analisis

penyelenggaraan (maintenance analysis). Penyelidikan sedia ada mengenai

penentuan masa untuk menjalankan PM kebanyakannya tidak berdasarkan analisis

penyelenggaraan yang mencukupi yang mana ia tidak mengambilkira faktor luaran

yang mengakibatkan kerosakan mesin dan tidak memahami bag aim ana proses

kerosakan mesin tersebut berlaku (analisis kerosakan fizikal). Oleh sebab itu,

penentuan masa terbaik untuk menjalankan PM tidak cukup tepat dan ia akan

mengurangkan faedah daripada strategi PM.

Dalam penyelidikan ini, satu model pembuat keputusan berdasarkan analisis

penyelenggaraan dengan mengambilkira faktor luaran dan analisis kerosakan fizikal

telah dibangunkan untuk menentukan dengan lebih tepat masa yang terbaik untuk

menjalankan PM. Oleh itu, faedah daripada strategi PM ini dapat dimaksimumkan.

8agi penilaian model ini, satu kajian kes telah dijalankan di industri pemprosesan di

Malaysia. Kajian kes ini memberi tumpuan pada kerosakan komponen mesin yang

tidak boleh diperbaiki (non-repairable component).

Keputusan daripada anal isis kerosakan fizikal telah mengenalpasti beberapa

maklumat penting iaitu faktor luaran yang mengakibatkan kerosakan komponen mesin

xii
dan pengkelasan data-data "censored" dan "uncensored". Maklumat-maklumat ini

merupakan input yang penting untuk analisis faktor luaran (covariate analysis) dalam

model pembuat keputusan ini. Keputusan daripada analisis faktor luaran menunjukkan

beberapa faktor luaran (covariates) yang dikenalpasti daripada analysis kerosakan

fizikal memberi kesan yang serius kepada kerosakan komponen mesin. Oleh itu, nilai

parameter daripada kesan tersebut digunakan atau diambilkira dalam pengiraan bagi

menentukan masa untuk menjalankan PM dengan lebih tepat. Keputusan daripada

pengiraan ini menunjukkan terdapat perbezaan masa PM yang ketara apabila

pengiraan yang mengambilkira faktor luaran dan pengiraan yang tidak mengambilkira

faktor luaran. Pada kesimpulanya, penentuan masa untuk menjalankan PM akan lebih

tepat sekiranya faktor luaran diambilkira dalam pengiraan terse but.

xiii
 DEVELOPMENT OF DECISION MODEL FOR MAINTENANCE ANALYSIS
OF NON-REPAIRABLE COMPONENT BY CONSIDERING
THE EXTERNAL FACTOR

ABSTRACT

Preventive Maintenance (PM) is one of the strategies that can be applied to

reduce the machine breakdown problems due to unplanned maintenance. However,

the application of PM in term of when is the best time to carry out the PM is an

important issue. The answer to this question should be based on an adequate

maintenance analysis. The existing study in determining the PM time is not based on

an adequate maintenance analysis, where there are not considered the external factor

(covariate) and without understanding the failure mechanism (physical failure).

Therefore, the determination of PM time is not accurate enough and reduced the PM

benefits.

In this research, a decision model of maintenance analysis by considering the

external factor (covariate) and physical failure has been developed in order to

determine the more accurate time of PM. Therefore, the PM strategy that has been

applied can maximise the benefits of PM. In order to validate this model, a case study

was carried out on a processing industry in Malaysia. The case study only focused on

non-repairable components in a machine.

The results from this physical failure analysis have identified the possible

external factors (covariates) that influence the component failure and have classified

the censored and uncensored data. This information is being used for the next analysis

(covariate analysis) in the decision model. The result shows that some identified

covariates have significant effects to the component failure. Thus, the parameter of

covariates effects is used to determine more accurate time of PM. The result of PM
time determination shows that there has a significant difference of PM time with and

without considering the external factor. Therefore, this result implies that more accurate

time of PM should be determined by considering the external factor (covariate).

xv
 CHAPTER ONE
INTRODUCTION

1.0 Introduction

This chapter is divided into five sections. The first section gives an overview of

the research background, the second section discusses the problem statement of the

research, the third section deals with the research objective, the fourth section includes

the research focus and delimitation and the fifth section describes the overview of this

thesis.

1.1 Research Background

After the Second World War, manufacturing industries such as textile, steel and

automotive played an important role in the growth of world economy. Western countries

were the pioneered in these industries. For example, in 1960, 65% of the world's

industrial production was monopolised by United States and Europe (Kister and

Hawkins, 2006). While in Eastern bloc, manufacturing industries were dominated by

Japan and Korea.

This situation increases the competition among the manufacturers to fulfill the

customer's satisfaction in term of time delivery and quality products. Therefore, it forces

the manufacturers to adopt "automation technology". The advantage of the automation

technology is that it increased the productivity with constant quality level. However, the

manufacturers have to invest very high cost in term of equipments, components and

machines. To reach the targeted rates of return on this high investment, the machines

should be fully utilised with minimum cost of stoppage and repair. In reality, machine's

breakdown due to unplanned maintenance (sudden failure) will increase the repair cost

and machine downtime (production lost) (Nakajima, 1986). Therefore, this has brought

the role of maintenance as an important activity in the manufacturing industries.

1
 Maintenance is defined as the combination of activities to restore the

component or machine to a state in which it can perform its designated functions

(Duffuaa et ai, 1999; Dhillon, 2002). Historically, the application of maintenance started

since the First World War for military purposed. At that time, maintenance was applied

with the aim to extend the lifetime of military machines such as truck, tank and aircraft.

In the industries, maintenance is an important element to achieve the organisation

goal, which is for successful competition and increase the product quality (Duffuaa et

ai, 1999). One of the maintenance strategies is Corrective Maintenance (CM). CM is a

traditional strategy that is used to restore the machine or its components to the

required function after the machine failed (Blanchard et ai, 1995). However, this

strategy reflects to high machine downtime (production lost) and maintenance (repair or

replacement) costs (Tsang, 1995). One of the alternatives of this strategy is by applying

the Preventive Maintenance (PM).

PM strategy involves the preventive tasks such as repair, replacement or

inspection at pre-determined interval (Gertsbakh, 1977). This interval is based on

scientific approach such as PM optimisation. The aim of PM's strategy is to reduce the

failure rate of the machine or component, thus the machine downtime and maintenance

(repair or replacement) costs can be reduced. In practise, the application of PM still has

a big gap and it is not based on the adequate analysis (maintenance analysis). This is

because of some reasons, for example, the application of PM strategy required a deep

understanding of the PM concept and the strict rule in maintenance analysis such as in

reliability modelling.

2
1.2 Problem Statement

In applying the PM strategy, the most critical questions are; what is the task and

when is best time to perform PM? These questions can be answered by carrying out a

maintenance analysis (Duffuaa et ai, 1999). Maintenance analysis refers to a process

of understanding the failure mechanism and reliability characteristics of the component

or machine. It includes the root cause analysis (physical failure analysis) for answering

the questions such as why it failed, how failure occurs and what is the failure causes.

Then the logical task such as replacement, repair or inspection is decided. While, the

best time to carry out this task is determined by PM models (also known as PM

optimisation model) such as the inspection model and the replacement model.

The research presented in this thesis is focused on the process of maintenance

analysis in order to determine the accurate time of PM, thus the benefits of the PM

strategy can be maximised. According to Dekker (1996), most of the maintenance

analysis is performed without understanding the root causes of the machine failure.

Therefore, it may result in the wrong information and data for reliability analysis and

affected to the determination of PM time (using PM optimisation model). Furthermore,

many of the researches and articles in the maintenance analysis assumed that the

failure (lifetime) of the component or machine depends only on aging (time usage)

factor (in this research it refers as internal factor) (see e.g. Clarotti et ai, 2004: Percy

and Kobbacy, 2000). However, in reality most of the component or machine failures are

influenced not only by the internal factor (age-time usage) but also by the external

factor (Lam and Zhang, 2003). The external factor would be the effects of

environmental (dust, humidity and heat), human skills, product types and maintenance

activities. In fact, when the external factor is not considered in the maintenance

analysis, it may give rise to errors in the identification of machine characteristics (failure

rate and failure distribution). Consequently, determination of the time of PM is not

3
accurate enough and it reduces the PM benefits (cost saving, high failure rate and

production lost).

The emphasis of this research is to develop a decision model for maintenance

analysis by considering not only the internal factor but also the external factor. The

advantages of this model are that it includes the root failure causes analysis (physical

failure analysis) and covariate (refers to external factor) analysis before the time of PM

is determined using the PM optimisation model. The main objective of the model is to

determine a more accurate time of PM by minimising the failure cost.

1.3 Research Objectives

The general objective of this research is to apply the Preventive Maintenance

(PM) strategy for solving machine breakdown problems. The specific objectives of this

research are as below;

1. Develop a decision model for the maintenance analysis to analyse the failure

data of non-repairable component by considering the external factors.

2. Verify and validate the decision model developed in real case study at a

processing industry.

3. Identify the critical component (non-repairable component) that causes major

machine breakdown using Pareto Analysis (PA).

4. Perform the Physical Failure Analysis (PFA) on critical component (non-

repairable component) using Failure Mode Effect and Criticality Analysis

(FMECA) techniques.

5. Codify and measure the effects of the external factor (covariate) on component

failure time by carrying out the Covariate Analysis (CA) based on Proportional

Hazard Model (PHM).

6. Determine the accurate time of Preventive Replacement (PR) using

maintenance optimisation model based on results from previous objective.

1.4 Research Focus and Delimitation

This research is carried out based on some limitations, there are;

• The application of Preventive Maintenance (PM) strategy is based on Time-

Based Maintenance (TBM) and it focuses on the determination of the

Preventive Replacement (PR) time.

• Only the cost criterion is considered in the PR analysis.

• Only non-repairable components in repairable system (machine) are studied. In

other words, only the critical components that contribute to the repairable

system failure are studied.

• Only a processing industry has been considered in the case study.

1.5 Thesis Overview

The overview of this thesis is as follows; Chapter two provides a literature

review of the related subjects. Chapter three, deals with the development of the

decision model. Chapter four presents the decision model validation, which a case

study is carried out at processing industry in Malaysia. Chapter five presents the

discussion of the decision model validation. Chapter six addresses with the conclusions

and recommendations of the future research.

5
 CHAPTER TWO
LITERATURE REVIEW

2.0 Introduction

This chapter is focuses on the literature review on the topic of machine

characteristic, factors of machine component failure, maintenance strategy and

reliability issue. The flow of the literature review is illustrated in Figure 2.1. In the topic

of machine characteristics, the components of a machine and their configuration are

discussed. In the next topic, two failure factors are discussed; there are internal factors

and external factors. In the topic of the maintenance strategy, the concept and

application of Preventive Maintenance (PM) based on PM optimisation models are

reviewed. In the topic of reliability issue, the emphasis is given on the reliability analysis

of failure time based on time-dependent and time-independent analysis. The final topic

in this chapter discusses the findings of the literature review.

,,,-------------------------------------------------------------------------.,,
, ,,,
,,
,,
Component
,,
Types of component ,,
in a machine configuration
,,
,,
____________________________________
,,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ______ 1

,,------------------------------------
,
-----------------------------------,

Internal factors The Factor of Internal factors

(Aging-time Machine (Environmental and
usage) Component Human error)
Failure

------------------------------------,
-----------------------------------.
/-F==:!==~
Maintenance
Strategy

Time-Based Inspection Model

Maintenance
Condition-Based Replacement
Maintenance Model
,
-----------------------------------~
-----------------------------------.,

Time-dependent Time-Independent
Analvsis Analvsis

Reliability Proportional Hazard

Issue Model

Exponential Model

Normal Model

Lognormal Model

Weibull Model

Figure 2.1: Flow of the Literature Review

7
2.1 Machine Characteristics

Machine can be described as any mechanical or electrical device that transmits

or modifies energy to perform or assist in the performance of human tasks (Shigley and

Mischke, 2001). Machine generally consists of three types of components i.e.

mechanical, electrical and electronic. Figure 2.2 shows the combination of these three

components in a machine, which refers to "a machine system" (Duffuaa et ai, 1999;

Bohoris, 1996). The function of the mechanical component is to transfer the energy into

motion, or move materials in a controlled way. Electrical component is used to change

electrical energy into another form of energy to perform a task. This energy can be

mechanical (motion), magnetic, thermal (heat) or chemical. Electronic components are

devices that are used for controlling the voltages and/or current in different signal.

Electrical Electronic

Figure 2.2: The Types of Components in a Machine System

These three components within a machine system may be related to one

another in two primary ways, either in a serial or in a parallel configuration as shown in

Figure 2.3 (Ebeling, 1997). Figure 2.3a and 2.3b show the diagram of serial and

parallel configurations, respectively. The characteristic of serial configuration is that all

the components must function for the machine to function, while in a parallel

configuration, at least one component must function for the machine to function.

8
 -II---HL--.....JH~c~
a
b

o =Components o =Components
a. Serial b. Parallel

Figure 2.3: a. Serial Configuration and b. Parallel Configuration

In practice, most machines have been designed in serial relationship

components. Consequently, failure in at least one of the component in serial

configuration will cause the whole system (machine) to breakdown.

2.2 The Factor of Machine Component Failure

As mentioned before, most of the machines are designed in serial configuration

and in many cases of the machine breakdown (failure) are due to unplanned

maintenance. Most of the unplanned maintenance is due to component failures.

According to Lam and Zhang (2003), failure of the machine component is influenced by

internal and external factors and the classification of these factors is illustrated in

Figure 2.4.

9
 The Fadors of
Component
Failure
I Internal Factor I External Factor

t
r-------------------, '-lI
.----------------1 \II
r--------------------
:, Aging (time usage) ,: : Environmental : I Human Errors ,:
1_____ _____________ __ I
causes , ~-------------------~
L ________________ J

Figure 2.4: The Internal and External Factors Influence on the Machine's
Component Failure

Referring to Figure 2.4, the internal factor refers to the cause of aging (time

usage) (Lam and Zhang, 2003; Lam and Yeh, 1994), while the external factor consists

of environmental causes (Lam and Zhang, 2003) and human errors (Clarotti et ai,

2004). The details of internal and external factors are discussed in the following

section.

2.2.1 Internal Factors

Internal factor refers to the age (time usage) of a component and it is normal

causes of component failure (Lam and Zhang, 2003). In many cases, increasing in the

component aging will be followed by reducing in the component performance due to

failure and it occurs for all types of components (mechanical, electrical and electronic)

(Valdez-Flores and Feldman, 1989: Moustafa et ai, 2004). In other words, this situation

is nature process in a component lifetime.

If we refer a component that operates in a machine (as a system), this factor

also results on reliability reduction, means that the failure rate of the component is

increased by machine age (time usage). Failure rate of a component (that operating in

a machine) can exist in three states; there are decreasing failure rate (early state),

constant failure rate (useful state) and increasing failure rate (deteriorating state)

10
(Ebeling, 1997). In reliability study, internal factor (age-time usage) is the basic

measurement of failure rate. It can be measured in different unit of time including

hours, days, working shifts, product cycle or some other units (Tong et ai, 2002). For

example, Freeman (1996) and Brown et al (2004) used the time unit of years in

analysing the failure rate of transformer. Detail of reliability analysis based on internal

factor (age-time) is discussed in Section 2.4.1.

As mentioned before, the failure of a component is not only because of internal

factor but it also influences by external factor. External factor such as environmental

causes and human errors may reduce the component lifetime and increase the

component failure rate.

2.2.2 External Factors

The environmental causes such as over heated (temperature), high humidity

level, extreme dust condition and over dosed radiation may influence to the component

failure. For instance, Oyebisi (2000) discussed the effects of high temperature and

humidity on electronic component. He states that when the working temperature and

humidity drastically change (increase or decrease), it affects to the performance of the

component due to failure. Other study carried out by Saka (1987), where he reports

that the fungi which are the products of high humidity environment have a serious

effect on the capacitance value of some types of capacitor. He found that a ceramic

capacitor has a high percentage changes in capacitance value up to 167.75 when

enclosed in a high fungous environment. Martorell et al (1999) studied the effects of

environment in Nuclear Power Plant (NPPs) on the lifetime of the mechanical

component and they found that over dosed of radiation from nuclear process has a

significant affects on the material properties of mechanical component and will reduce

the component lifetime.

11
 Human error is another cause of external factor that may influence on the

component failure. Human error is defined as the failure to perform a specified task that

result in damage to property and equipment (Dhillon and Liu, 2006). According to

Wang and Hwang (2004), human errors can be divided by two types: first is the critical

human error which will cause system breakdown (unplanned breakdown) and second

is latent human error which does not lead to immediate system breakdown. Unskilled

technicians or maintenance crews may lead to the component failure as well as

machine breakdown through maintenance activities. For example, Dhillon and Liu

(2006) stated that component failure may occur due to maintenance error by carrying

out incorrect repair or preventive action. For examples, incorrect calibration of

component and application of the wrong grease at appropriate points of the

component. In addition, Latorella and Prabhu (2000) presented the most common

maintenance errors related to industry; there are incorrect installation of components,

fitting of the wrong parts, electrical wiring discrepancies (including cross connection),

loose objects (tools) and inadequate lubricant. Even the human error is unavoidable,

but the percentage of error of unskilled operators and technicians can be reduced by

intensive training program or improving the maintenance management system (Wang

and Hwang, 2004).

In fact, both of internal and external factors are related to each other. Internal

factor (age-time usage) presents the actual component lifetime (with certain failure rate

state) and the effects of external factor may reduce or extent the component lifetime as

well as the failure rate. For example, the effect of proper and adequate maintenance

may extent the component lifetime (as well as reduce the failure rate), while

maintenance error may reduce the component lifetime (as well as increase the failure

rate). Therefore, determination of an appropriate maintenance strategy at the right time

may extent the component lifetime and reduces the component failure rate (Martorell et

ai, 1999; Oyebisi 2000; Wang and Hwang, 2004).

2.3 Maintenance Strategy

As mentioned in Section 1.1, maintenance is a set of activities or tasks to

restore a machine or a component to a state in which it can perform its designated

functions (Duffuaa et ai, 1999; Dhillon, 2002). In practice, maintenance is classified into

two strategies; there are Corrective Maintenance (CM) and Preventive Maintenance

(PM) (Hall, 1997) as shown in Figure 2.5. Detail of each strategy is discussed in the

following section.

Maintenance Strategy

Corrective Maintenance Preventive Maintenance

Cond ition-Based Time-Based

Maintenance Maintenance

Figure 2.5: Maintenance Strategy Classification

2.3.1 Corrective Maintenance Strategy

Corrective Maintenance (CM), also known as Run-To-Failure or reactive

strategy is a traditional strategy that used to restore (repaired or replaced) the machine

or component to it required function after it has failed (Blanchard et ai, 1995). This

strategy reflects to high machine downtime (production lost) and maintenance (repair or

replacement) costs due to sudden failure (Tsang, 1995). Moreover, continuous

application of CM will effect to the machine performance. For example, Swanson

(2001) shows that there are strong relationships between the application of CM

strategy and PM strategy to the machine performance. Based on survey and analysis,

the author found that the application of CM strategy significantly reduced the machine

performance in term of reliability, availability, production lost and product quality.

1'1
Therefore, an alternative of CM strategy is by applying the Preventive Maintenance

(PM).

2.3.2 Preventive Maintenance Strategy

Preventive Maintenance (PM) is widely applied in industry for asset

management such as production machines. The concept of PM is to perform the

maintenance activities before the machine fails (Gertsbakh, 1977; Lofsten, 1999). The

main objective of PM is to reduce the failure rate or failure frequency of the machine. It

contributes to reducing cost, minimising machine downtime (production lost), increasing

productivity and improving quality (Usher et ai, 1998). In other word, PM is a series of

pre-planned tasks such as repair, replace, overhaul and inspection with the aimed to

maximise reliability and availability of the machine. The application of PM's strategy

can be carried out based on two approaches; there are Time-Based Maintenance

(TBM) and Condition-Based Maintenance (CBM) (Duffuaa et ai, 1999).

In TBM approach (also known as statistical-reliability based), PM is performed

based on machine aging (time usage) (Mann et ai, 1995). The aging of the machine is

determined based on the use of statistical and reliability analysis. As mentioned before,

aging can be measured in different unit of time such as operating time, cycle time,

working shifts, working days or some others unit of time. The objective of TBM is to

determine the optimum intervals (time) of PM in order to minimise the total cost of

failure (reducing failure rate) and machine downtime (production lost). TBM is usually

applied on single or non-repairable component such as machine tool (Jianqiang and

Koew, 1997; Bahrami et ai, 1998). In addition, TBM is feasible when the machine or

component is in deteriorating state (or other word in increasing failure rate) and the

cost of PM is less then the cost of CM (Mann et ai, 1995). In relation of this, Mobley

(1990) stated that the cost of CM can be in excess of three times of PM. Reasons for

this include;

14
 • CM will extend the downtime due to unavailability of components, or labour.

• CM can result in overtime.

• CM is not executed as efficiently as PM.

In CBM approach, PM is carried out will depend on the condition of the

machine, which involves the use of sensor or other indicator (Andersen and

Rasmussen, 1999). The condition of the machine is measured directly to its operation

based on the different variables such as vibration, temperature, lubricating oil,

contaminants and noise level. Under CBM, PM is performed only "when needed" or just

before failure. One of the advantages of CBM is its feasibility to apply at any state of

the machine lifetime, whether in early state (decreasing failure rate), useful state

(constant failure rate) or uncertain state (random failure rate) (Tsang et ai, 2006). In

fact, CBM is suitable when applying on expensive equipment or complex system such

as transformer, turbine system, electronic board, power station, electric motor, engine

and generator. For instance, Han and Song (2003) reviewed the success of CBM

application on electrical equipment such as power transformer, generator and induction

motor. In other cases, Jardine et al (2001) applied CBM on mine haul truck wheel

motor based on oil analysis and they determine the optimum time of repair. While

Beebe (2003) applied CBM to study the steam turbine performance and determine the

optimum time of PM based on vibration analysis and oil analysis.

However, the application of CBM requires an expensive equipment or system

(such as special indicator) for continuous monitoring. Moreover, as discussed in

Section 2.1 and 2.2, most of the machine failure (breakdown) in production line is due

to the component failure. This component may fail without given any signs (based on

failure mode only) and failure may influence by internal (age-time usage) and external

1 t::.
factors (environmental causes and human error) (Dekker, 1996). Therefore, based on

this reason, the application of TBM approach is more suitable for PM application.

2.3.3 Preventive Maintenance Application

The main issue in PM application is in determining the optimal time to carry out

the PM's task (replacement or inspection) (Jardine, 1973). If the task is made too early,

the components may not have been utilised to full capacity. If the interval is too long, it

reflects too high machine down time due to unplanned maintenance (restoration due to

sudden failure). Moreover, most manufacturers recommend that PM intervals must be

followed to preserve warranty rights. The determination of these intervals by the

manufacturer may not reach to maximum benefits of PM strategy (Bahrami et ai, 2000).

One of the solutions of this matter is using the maintenance optimisation model or PM

optimisation model.

According to Dekker (1995), PM optimisation models refer to the mathematical

models whose aim is to determine the optimum time/interval of PM which maximises or

minimises some criteria such as failure cost, profit, downtime, availability and reliability.

The concept of PM optimisation model is illustrated in Figure 2.6.

Maintenance (replacement)
cost! cycle

Failure cost /
cycle

C(tp). PM cost /
Optimum
cost
cycle

t p• t, time
Optimum time

Figure 2.6: Determination the Optimum Time of PM by Minimising the Failure Cost
Criteria

16
 Referring to Figure 2.6, it shows that in reality the failure cost / cycle will

increase, while the PM cost / cycle will decrease by aging (time usage). The cross point

between the failure cost line and PM cost line presents the optimum time to perform

PM in order to minimise the failure cost criterion. Inspection and Preventive

Replacement (PR) models are examples of PM optimisation models that widely

discussed in literature (Scarf, 1997). Detail of Inspection and Preventive Replacement

(PR) models in solving various maintenance problems is discussed in following section.

2.3.4 Inspection Model

Inspection is a process of identifying the current state of the machine by

detecting the hidden failure or sign that may lead to major damage and it is usually

applies on complex system (Jardine, 1973). Hence, preventive actions such as minor

modification, minor repair, minor replacement and/or cleaning can be taken before the

major failure occurs (Hauge et ai, 2002). Examples of inspection activities are physical

records (checking) and condition monitoring techniques such as vibration, noise,

radiation, oil and temperature tests. The benefit of inspection is to prevent the machine

from major failure (Bahrami et ai, 1998). Many inspection models have been proposed

to determine the optimum time to carry out the inspection and the earliest basic model

was introduced by Barlow et al (1963). Following the introduction by Barlow, various

inspection models have been developed and modified based on various case study

problems.

For example, inspection model is widely applied on emergency and storage

equipment or system. For instance, Jardine (1973) proposed an optimal inspection

model on the equipment that is used in emergency condition such as fire extinguisher

and alarm system. The benefit of this model is to determine the optimum time of

inspection in order to maximise the availability of these equipments that used in

emergency condition. While the system which in storage for a long time, its reliability

17
may go down with time. Thus, a system in storage should be inspected and maintained

at periodic time to reach the higher reliability and availability. Ito and Nakagawa (1995)

suggested an optimal inspection model for storage system such as weapon

equipments. The objective of this model is to determine the optimum time of inspection

to maximise the reliability of this storage system. On the other hand, Ito et al (1995)

extended the optimal inspection model for storage system such as missiles and spare

parts of aircraft, which the objective of this model is to maximise the reliability of the

system.

In other cases, inspection model also has also been applied for the case of

randomly failing or randomly shocks system. For instance, Mathew and Kennedy

(2002) developed an optimal inspection model based on failure due to random shock

load. The objective of this model is to reduce failure frequency as well as to minimise

the failure cost. Chelbi and Ait-Kadi (1999) and (1995); Hariga (1996) developed and

proposed an inspection model for determining the optimal inspection time for a system

subjected to random failure. The objective of this model (inspection model) is to

minimise the cost of preventive and corrective action. Chelbi and Ait-Kadi (2000)

proposed the inspection models for the case of randomly failing system, which the

objective is to maximise the system availability.

2.3.5 Preventive Replacement Model

According to Wang (2002), replacement is a perfect maintenance because

component will function as good as new after it was replaced. Like inspection model,

the main objective of PR model is to determine the optimum time (interval) of PR in

order to maximise or minimise some optimisation criteria such as failure cost, profit,

downtime, availability and reliability. However, the application of PR model is only

feasible when the cost of PR is less than the corrective replacement (under CM

strategy) cost and the failure rate of the component should be increasing mode, which

18
can be determined based on reliability analysis (Jardine, 1973; Ebeling, 1997; Barlow

and Proschan, 1965). There are two basic PR models which are widely discussed in

the literature; first is Block Replacement Model (BRM) and second is Age Replacement

Model (ARM) (Handlarski, 1980; Aven and Jensen, 1999).

In BRM (also known as constant-interval replacement model), a component is

replace at constant intervals or time irrespective of the age of the component and

carried out the failure replacement (sudden failure) whenever necessary (Barlow and

Proschan, 1965). However, BRM is not a popular model for real application. The

reason is BRM exist in complicated form and practically it is difficult to apply. Moreover,

according to Bahrami et al (2000), BRM has the undesirable characteristic that

relatively good components are replaced more frequently than required. Consequently,

it reflects to high maintenance cost (component replacement), production lost (machine

downtime due to PR) and the component is not fully utilised. The alternative of BRM is

ARM. In ARM, a component is always replaced according to its age, T or failure,

whichever occurs first, where T is constant (Barlow and Hunter, 1960). In fact, ARM is

more practical and benefits to apply compared to BRM. For example Jiang et al (2006)

shows that the cost saving by using PR based on ARM is higher compared with BRM.

This is because PR by using BRM performs more PR actions (more then needed),

which reflects to high maintenance cost and production lost.

In practice, the application of PR model has been modified and extended based

on BRM or ARM in order to solve many maintenance problems and it is usually applies

to non-repairable component. For instance, Huang et at (1995) proposed a standard

solution of ARM in order to minimise the replacement cost of failure and it applied for

the replacement problem of drilling tool. Jianqiang and Keow (1997) modified the ARM

that applied in computer-integrated manufacturing environment. From the model they

determined the optimum time of tool replacement in order to minimise the failure and

19
production cost. Even the application of BRM is not practical and less benefit in term of

cost saving, Bahrami et al (2000) proposed a new perspective of BRM in order to

reduce this deficiencies and it is applied on machine tool in the crankshaft line at a car

engine manufacturing company. The main objective of this model (new perspective

BRM) is to determine the optimum time of PR with minimise the total machine

downtime. The detail reviews of various PR models for solving the single-unit problem

has been discussed by W~ng (2002), which various existing replacement models were

summarised, classified and compared.

According to Dekker (1996), the application of PM based on optimisation

models can lead to totally wrong result if the information of failure characteristics

(lifetime) of the component is not fully defined and understand. This information can be

determined by carrying out the reliability analysis. Reliability analysis concerns about

the failure data (aging-time usage) analysis in order to identify the state of component

lifetime, whether decreasing failure rate, constant failure rate or increasing failure rate.

The reliability analysis is discussed in detail in the following section.

2.4 Reliability Issue

Reliability is a word with many different connotations and applications. For

example, when applied to human being, it refers to the ability of the person to perform

certain tasks according to a specified standard. For a component application, reliability

means the ability of the component to fulfil what is required of it (Crowder et ai, 1991).

Reliability is measured in different unit of time such as operating time, cycle time,

working shifts, working days or some others unit of time. The basic measurement of

reliability is the failure rate, where high component failure rate means lower reliability

(Ghodrati, 2005).

20
 In maintenance analysis (refers to PM application based on TBM approach),

reliability is an essential part (sub-analysis) to present the characteristics of the

component failure (lifetime) through it failure time distribution. In reliability analysis, the

characteristics of the component failure can be represented by the random variable T

(time to failure) and it can be formed into different failure time functions as shows in

Table 2.1.

Table 2.1: The Failure Time Function (Ebeling, 1997)

Failure function Purpose

Cumulative distribution function (CDF)

F(t) = 1-R(t) =Pr {T < t} Presents the failure distribution in cumulative

form
(Probability that a failure occurs before
time, t)
Probability density function (PDF)
Describe the shape of the failure distribution
f(t) =dF(t)/dt =- dR(t)/dt
Failure rate or Hazard rate function Identify the characteristics of failure distribution,
whether in Increasing failure rate (IFR), Constant
Nt)= f(t)/R(t) failure rate (CFR) or Decreasing failure rate
(DFR)

Referring to Table 2.1, the first failure function is Cumulative Distribution

Function (CDF), F(t) which presents the failure distribution of the component in

cumulative form. The second failure function is Probability Density Function (PDF), f(t)

which describes the shape of failure distribution such as Weibull, Exponential, Normal

and Lognormal distributions. The final failure function is failure rate or hazard rate, Nt)
which determine the characteristics of failure distribution whether in increasing,

constant or decreasing failure rates. In addition, each of these failure functions can be

presented by different form based on different reliability models.

71
2.4.1 Reliability Models

The failure time (failure distribution) of a component can be modelled using

different reliability models (also known as time-dependent failure model) such as

Exponential, Normal, Lognormal and Weibull. Exponential distribution model is used to

model the failure time with constant failure rate. This model is one of the important

models for modelling the failure time of electric and electronic component or system

(Ebeling, 1997). For instance, Tong et al (2002) applied the exponential distribution for

assessing performance of lifetime (failure rate) index of electronic component. They

developed more simple procedure for testing the lifetime performance index under the

exponential distribution.

Normal and Lognormal distributions models are widely used to model the failure

time with increasing failure rate for fatigue and wear-out phenomena (Ebeling, 1997). In

many cases, it usually uses in modelling the failure time of the cutting tool. For

example, Chelbi and Kadi (1999) used the Normal distribution to model the failure time

of cutting tool and it applied for determining the optimum time of inspection. In other

cases, Jianqiang and Keow (1997) used the Lognormal distribution to fit the failure time

of the cutting tool for the purposed of determination the best replacement interval.

One of the most popular reliability models in maintenance application is Weibull

distribution model. Weibull distribution model is widely used for characterising the

mechanical component (Ghodrati, 2005; J6zwiak, 1997). For example, Tam et al

(2006) applied the Weibull distribution to model the lifetime of multi-component system

in order to determine the optimal maintenance interval. While Farrero et al (2002) used

the Weibull model for failure data analysis (failure time modelling) on production system

in order to determine the optimum time of replacement stocks.

22
 In addition, Weibull distribution model usually can be presented with two

parameters. There are scale parameter, e and shape parameter, /3. Scale parameter, e
shows the lifetime of the component. While shape parameter /3, presents the

characteristic of component lifetime whether in decreasing, constant or increasing

failure rate (Ebeling, 1997). The characteristics of component lifetime can be identified

based on the different value of parameter, /3, where;

/3 < 1, represents the decreasing failure rate

/3 =1, represents the constant failure rate

/3 > 1, represents the increasing failure rate

However, the application of this conventional reliability models (Exponential,

Normal, Lognormal and Weibull) that have been presented in this section assume that

the failure data (failure time) used is time-dependent (internal factor). In the scope of

this research, it means that the reliability analysis that carried out using these reliability

models only considered the internal factor (age-time usage) influencing to the

component failure (failure time). In practise, this assumption is not always true

because, as discussed in Section 2.2, failure of a component may influence by both of

internal and external factors (Lam and Zhang, 2003). Consequently, if only the internal

factor is considered in reliability analysis, the result may not be accurate enough and

effects in the calculation of PM time. Therefore, the reliability analysis should be

performed by considering the external factor in order to reduce this effect. The

reliability analysis which is considered the external factor can be performed by

integrating the conventional reliability models (Exponential, Normal, Lognormal and

Weibull models) with Proportional Hazard Model (PHM).

2.4.2 Proportional Hazard Model (PHM)

As mentioned before, in reliability analysis based on time-dependent (refers to

internal factor), the conventional reliability models are used for modelled the failure

distribution of the component. However, when the external factor is considered, which

is referred to time-independent, the reliability analysis can be performed based on

regression type analysis by integrating with Proportional Hazard Model (PHM) (Cox,

1972). Generally, PHM describes the relationship between the reliability analysis based

on internal factor (time-dependent) and changing the external factors or covariates

(Coit and English, 1999).

PHM is widely used to analyse the failure time data, where the failure time

depends on external factor or covariates (time-independent). The basic assumption of

PHM is that the actual hazard rate (failure rate), Ao(t) of a component is the product of

a based line (time-dependent) hazard rate A(t) and a positive (time-independent)

functional term exp(a, x) , basically independent of time, incorporating the effects of a

number of covariates. Mathematically, PHM can be expressed in failure rate function as

below;

Ao (t) =A (t) exp(a, x) ---------------------- (2. 1)

Where a, is a row vector consisting of the covariates and x is a column vector

consisting of the regression parameters. On the other hand, the hazard rate (observed

hazard rate) of a system or component will change when the covariates change.

The concept of covariates that have an effects on based line (time-dependent)

hazard rate A(t) are shown in Figure 2.7. Baseline hazard rate is the failure rate without

covariate effects (only internal factor effects - aging and accumulative wear), while the

')4