INTRODUCTION OF INTELLIGENT MACHINE FAULT DIAGNOSISAND PROGNOSIS

CONTENTS

PREFACE

ACKNOWLEDGMENTS

ABOUT THE AUTHORS

INTRODUCTION

1.1. BACKGROUND

1.2. MAINTENANCE STRATEGY

1.3. DEFINITION OF FAULT DIAGNOSIS AND PROGNOSIS

1.4. CONDITION BASED MAINTENANCE SYSTEM ARCHITECTURE

1.5. FAULT DIAGNOSIS FRAMEWORK

1.6. ROLE OF CONDITION MONITORING, FAULT DIAGNOSIS AND PROGNOSIS

1.7. PROBLEMS DURING IMPLEMENTATION

1.8. RELATED TECHNIQUES

1.9. REFERENCES

DATA ACQUISITION, PROCESSING AND ANALYSIS

2.1. INTRODUCTION

2.2. DATA ACQUISITION

2.2.1. Selecting A Proper Measure

2.2.2. Vibration Transducers

1) Proximity Probes

2) Velocity Transducers

3) Accelerometers

4) Key-Phasor

2.2.3. Transducer Selection

2.2.4. Transducer Mounting

1) Stud Mounting

2) Cement Mounting

3) Wax Mounting

4) Adhesive Mounting

5) Magnetic Mounting

6) Hand-Held

2.2.5. Transducer Location (Measuring Point)

2.2.6. Frequency Spans

2.2.7. Data Display

1) Time Waveform

2) Spectrum

3) Orbit

2.3. DATA PROCESSING

2.3.1. Data preprocessing

1) Wavelet Transform

Continuous Wavelet Transform (CWT)

2) Averaging

3) Enveloping

4) Cesptrum

2.4. DATA ANALYSIS

2.4.1. Features in Time Domain

1) Cumulants

2) Upper and Lower Bound of Histogram

4) Auto-Regression Coefficients

2.4.2. Features in Frequency Domain

1) Fourier Transform

2) Spectral Analysis

3) Frequency parameter indices

4) Higher order spectra (HOS)

2.4.3. Features in Time-Frequency Domain

1) Short Time Fourier Transform (STFT)

2) Wavelet Transform

2.5. REFERENCES

FEATURE EXTRACTION AND CLUSTERING

3.1. INTRODUCTION

3.2. DEFINITION OF SOME BASIC CONCEPTS

3.2.1. Pattern and Feature Vector

3.2.2. Class

3.3. PARAMETER EVALUATION TECHNIQUE

3.4. PRINCIPAL COMPONENT ANALYSIS (PCA)

3.5. INDEPENDENT COMPONENT ANALYSIS (ICA)

3.6. KERNEL PCA

3.7. KERNEL ICA

3.8. FISHER DISCRIMINANT ANALYSIS (FDA)

3.9. LINEAR DISCRIMINANT ANALYSIS (LDA)

3.10. GENERALIZED DISCRIMINANT ANALYSIS (GDA)

3.11. CLUSTERING

3.11.2 K-Means Clustering

3.11.3. Hierarchical Clustering

3.12. OTHER TECHNIQUES

3.13. REFERENCES

FEATURE SELECTION

4.1. INTRODUCTION

4.2. INDIVIDUAL FEATURE EVALUATION (IFE) BASED ON SPACE DISTRIBUTION

4.3. CONDITIONAL ENTROPY

4.4. BACKWARD FEATURE SELECTION

4.5. FORWARD FEATURE SELECTION

4.6. BRANCH AND BOUND FEATURE SELECTION

4.7. PLUS L-TAKE AWAY R FEATURE SELECTION

4.8. FLOATING FORWARD FEATURE SELECTION

4.9. DISTANCE EVALUATION TECHNIQUE

4.10. TAGUCHI METHOD-BASED FEATURE SELECTION

4.11. GENETIC ALGORITHM

4.11.1. General concept

1) Selection

2) Crossover

3) Mutation

4.11.2. Differences from other Traditional Methods

4.11.3. Simple Genetic Algorithm (SGA)

4.11.4 Feature Selection Using GA

4.12. SUMMARY

4.13. REFERENCES

FAULT CLASSIFICATION ALGORITHMS

5.1. INTRODUCTION

5.2. LINEAR CLASSIFIER

5.2.1. Linear Separation of Finite Set of Vectors

5.2.2. Perceptron Algorithm

5.2.3 Kozinec’s Algorithm

5.2.4. Multi-Class Linear Classifier

5.3. QUADRATIC CLASSIFIER

5.4. BAYESIAN CLASSIFIER

5.5. K-NEAREST NEIGHBORS (K-NN)

5.6. SELF-ORGANIZING FEATURE MAP (SOFM) NEURAL NETWORK

5.7. LEARNING VECTOR QUANTIZATION (LVQ) NEURAL NETWORK

5.8. RADIAL BASIS FUNCTION (RBF) NEURAL NETWORK

5.9. ART-KOHONEN NEURAL NETWORK (ART-KNN)

5.10. SUPPORT VECTOR MACHINES (SVMS)

5.10.1. Wavelet SVM

5.10.2. Multi-Class Classification

5.10.3. Sequential Minimal Optimization (SMO)

5.11. DECISION TREE

5.11.1. Building Decision Tree

5.11.2 Pruning Decision Tree

5.12. RANDOM FOREST

5.12.1. Random Forest

1) CART Methodology

2) Tree Building

3) Stopping Tree Building

5.12.2. Random Forests Algorithm (RF)

1) Two Randomized Procedures in RF Tree Building

2) Convergence of RF

3) Accuracy of RF Depending on Strength and Correlation

5.12.3. Genetic Algorithm

5.13. ADAPTIVE NEURO-FUZZY INTEGRATED SYSTEM (ANFIS)

5.13.1 Classification and Regression Trees (CART)

1) Tree Building

2) Tree Pruning

5.13.2. Adaptive Neuro-Fuzzy Inference System (ANFIS)

1) Architecture of ANFIS Based on CART

2) Learning Algorithm of ANFIS

5.14. CASE STUDIES: FAULT DIAGNOSIS OF INDUCTION MOTORS

5.14.1. Wavelet SVM

1) Experiment And Data Acquisition

2) Feature Calculation

3) Feature Extraction and Reduction

4) Training and Classification

5) Result And Discussion

5.14.2. Decision Tree

5.14.3 Random Forest

1) Data Acquisition and Feature Calculation

2) Fault Diagnosis Result and Discussion

5.14.4. Cart-Anfis

1) Data Acquisition

2) Feature Calculation

3)Feature Selection and Classification

5.15. REFERENCES

DECISION FUSION ALGORITHMS

6.1. INTRODUCTION

6.2. FUSION APPLICATION AREAS

6.3. FUSION ARCHITECTURES

6.3.1. Data-Level Fusion

6.3.2. Feature-Level Fusion

6.3.3. Decision-Level Fusion

6.4. FUSION TECHNIQUES AT DECISION-LEVEL

6.4.1. Voting Method

6.4.2. Bayesian Belief Fusion

6.4.3. Behavior Knowledge Space (BKS)

6.4.4. Dempster-Shafer Theory

6.4.5. Multi-Agent Fusion

6.4.6. Decision Templates (DTS)

6.5. CLASSIFIER SELECTION

6.5.1 2-Classifier Correlation Analysis

6.5.2. Multi-Classifier Correlation Analysis

6.6. DECISION FUSION SYSTEM

6.6.1. Level 1 To Level 3: Preparation for Decision Fusion

6.6.2. Level 4: Data Fusion at Decision Level

6.6.3. Level 5: Classifier Selection

6.6.4. Level 6: Decision Fusion

6.7 FAULTS DIAGNOSIS OF TEST-RIG MOTORS USING FUSION TECHNIQUES

6.7.1. Data Acquisition

6.7.2. Feature Calculation and Classification

6.7.3. Classifiers Selection and Fusion

6.7.4.Classifiers Fusion Comparison

6.8. FAULTS DIAGNOSIS OF ELEVATOR MOTOR USING FUSION TECHNIQUES

6.8.1. Data Acquisition

6.8.2. Feature Calculation and Classification

6.8.3. Classifiers Fusion

6.9. DECISION-LEVEL FUSION DIAGNOSIS USING TRANSIENT CURRENT SIGNAL

6.9.1. Experiments and Data Acquisition

6.9.2. Signal Preprocessing and Wavelet Transform

6.9.3. Features Calculation and Classification;

6.9.4. Fusion Performance Evaluation

6.10. REFERENCES

FAULT PROGNOSIS ALGORITHMS

7.1. INTRODUCTION

7.2. PROGNOSIS APPROACHES

7.2.1. Rule-Based Approaches

7.2.2. Fuzzy Logic Approaches

7.2.3. Model-Based Approaches

1) Physics-Based Model

2) System Dynamic Model

3) Probabilistic Model

7.2.4. Trend-Based Evolutionary Approach

7.2.5. Data-Driven Model Based Approach

1) Neural Networks

2) Support Vector Regression (SVR)

(1) Linear Support Vector Regression

2) Nonlinear Support Vector Regression

7.2.6. State Estimator-Based Prognosis

7.2.7. Statistical Reliability and Usage-Based Approaches

7.2.8. Adaptive Prognosis

7.2.9. Data Miningand Automated Rule Extraction

7.2.10. Distributed Prognosis System Architecture

7.3. APPLICATIONS

7.3.1. Bearing Prognosis

1) Spall Initiation Model

2) Spall Progression Model

7.3.2. Gear Prognosis

7.3.3. Low-Methane Compressor Prognosis

7.3.4. Machine-Tool Prognosis

7.4. REFERENCES

APPENDIX

INDEX