New Generation Artificial Intelligence-Driven Diagnosis and Maintenance Techniques: Advanced Machine Learning Models, Methods and Applications

Preface
Acknowledgments
Contents
About the Authors
Part I Introduction
1 Overview of Intelligent Fault Diagnosis and Maintenance for Rotating Machinery
	1.1 Introduction
	1.2 Objectives of the Monograph
	1.3 Outline of the Monograph
	References
Part II Deep Learning and Sparse Representation Coupled Intelligent Diagnosis and Maintenance
2 Sparse Model-Driven Deep Learning for Weak Fault Diagnosis of Rolling Bearings
	2.1 Introduction
	2.2 Theoretical Background of Dictionary Learning
	2.3 Methodology
		2.3.1 Sparse Coding via FISTA
		2.3.2 DNN-Based Adaptive Parameter Estimation
		2.3.3 Establishment of Multi-mode Training Data Set
		2.3.4 Deep Network-Based Sparse Denoising Framework
	2.4 Simulation Analysis
		2.4.1 Parameter Description
		2.4.2 Results and Discussion
	2.5 Experiment Analysis
	2.6 Conclusion
	References
3 Memory Residual Regression Autoencoder for Bearing Fault Detection
	3.1 Introduction
	3.2 Technical Preliminaries
		3.2.1 Autoencoder
		3.2.2 Memory-Augmented Autoencoder
		3.2.3 Latent Space Autoregression
	3.3 Proposed Method
		3.3.1 Architecture of Proposed Model
		3.3.2 1D Convolutional Autoencoder
		3.3.3 Memory Module
		3.3.4 Memory Residual Regression
		3.3.5 Loss Function and Anomaly Indicator
	3.4 Experiment Verification
		3.4.1 Datasets Description
		3.4.2 Data Preprocessing
		3.4.3 Performance Evaluation
		3.4.4 Experimental Results
	3.5 Conclusion
	References
Part III Transfer Learning-Based Intelligent Diagnosis and Maintenance
4 Fault Diagnosis of Polytropic Conditions Based on Transfer Learning
	4.1 Introduction
	4.2 Theoretical Foundation
		4.2.1 Variational Mode Decomposition (VMD)
		4.2.2 Feature Extraction
		4.2.3 Extreme Gradient Boosting (XGBoost)
		4.2.4 Manifold Embedded Distribution Alignment (MEDA)
	4.3 The Proposed Architecture
	4.4 Experimental Verification
		4.4.1 Variable Load Datasets from CWRU Lab
		4.4.2 Variable Load Datasets from XJTU Lab
		4.4.3 Variable Speed Datasets from XJTU Lab
	4.5 Conclusion
	References
5 Performance Degradation Assessment Based on Transfer Learning for Bearing
	5.1 Introduction
	5.2 Design of Method and Framework
		5.2.1 Description and Framework of Bearing Degradation Assessment
		5.2.2 Deep Hierarchical Features Extraction
		5.2.3 Joint Geometrical and Statistical Alignment (JGSA)
		5.2.4 Support Vector Machine (SVM)
	5.3 Bearing Experimental Dataset Evaluation
		5.3.1 Bearing Fault Severity Classification
		5.3.2 Bearing Fault Degradation Estimation
	5.4 Conclusions
	References
6 Remaining Useful Life Prediction on Transfer Learning for Bearing
	6.1 Introduction
	6.2 Problem Statement
	6.3 The Framework of the Proposed Methods
		6.3.1 Operational Condition Attention Mechanism
		6.3.2 Global Feature Extraction
		6.3.3 Global Feature Domain Adaptation
	6.4 Case Study
		6.4.1 Dataset Description
		6.4.2 Pseudo-OCs
		6.4.3 Comprehensive Performance Comparison
		6.4.4 Intermediate Feature Analysis
	6.5 Conclusions
	References
Part IV Adversarial Learning-Based Intelligent Diagnosis and Maintenance
7 Deep Sequence Multi-distribution Adversarial Model for Abnormal Condition Detection in Industry
	7.1 Introduction
	7.2 Methodology
		7.2.1 Problem Definition
		7.2.2 Theoretical Basis
	7.3 DSMDA Anomaly Detection
		7.3.1 Model Learning Stage
		7.3.2 Model Testing Stage
	7.4 Experiments
		7.4.1 Experimental Settings
		7.4.2 Experimental Results
	7.5 Discussion and Conclusions
	References
8 Multi-scale Lightweight Fault Diagnosis Model Based on Adversarial Learning
	8.1 Introduction
	8.2 Theoretical Background
		8.2.1 Convolutional Neural Network
		8.2.2 Adversarial Learning Training
		8.2.3 Channel Attention Module
	8.3 The Proposed Multi-scale Lightweight Fault Diagnosis Method
		8.3.1 Multi-scale Convolution Operation
		8.3.2 Ghost Transformation Operation
		8.3.3 Micro-adversarial Module
		8.3.4 Depthwise Separable Convolution
		8.3.5 Inverted Residual Block
		8.3.6 The Proposed Method for Condition Monitoring
	8.4 Experimental Verification
		8.4.1 Case 1: Data from CWRU
		8.4.2 Case 2: Data from Laboratory
		8.4.3 Discussions
	8.5 Conclusions
	References
9 Performance Degradation Assessment Based on Adversarial Learning for Bearing
	9.1 Introduction
	9.2 Bearing Real-Time Condition Monitoring Based on Partial Transfer Learning Method
		9.2.1 Problem Formulation
		9.2.2 Partial Domain Adaptation Method
		9.2.3 Proposed Bearing Condition Monitoring Method
	9.3 Experimental Evaluation
	9.4 Conclusion
	References
Part V Graph-Structured Information-Based Intelligent Diagnosis and Maintenance
10 Modelling and Feature Extraction Method Based on Complex Network and Its Application in Machine Fault Diagnosis
	10.1 Introduction
	10.2 Principle
		10.2.1 Recursive Algorithm
		10.2.2 Construction of Sub-network Average Degree
		10.2.3 An Extraction Method Based on FCN Sub-network Average Degree
		10.2.4 Performance Degradation Assessment Method Based on Network Structural Features
	10.3 Data Validation and Analysis
		10.3.1 Extraction of FCN Sub-network Average Degree of Bearing Condition
		10.3.2 Comparative Analysis
	10.4 Assessment of Bearing Performance Degradation
		10.4.1 Degradation Feature Extraction and State Warning
		10.4.2 Degradation State Recognition
	10.5 Conclusion
	References
11 Community Clustering Algorithms and Its Application in Machine Fault Diagnosis
	11.1 Introduction
	11.2 Theoretical Background
		11.2.1 Network Topology Feature Extraction
		11.2.2 Community Clustering
		11.2.3 Construction of Evaluation Index
	11.3 Fault Diagnosis Using Modified Fast Newman Algorithm
	11.4 Experimental Studies
		11.4.1 LFR Standard Test Network Verification
		11.4.2 Laboratory Dataset Validation
	11.5 Conclusion
	References
12 Remaining Life Assessment of Rolling Bearing Based on Graph Neural Network
	12.1 Introduction
	12.2 Graph Sampling Aggregation Network Model
	12.3 Remaining Life Assessment Based on Graph Sampling Aggregation Network
		12.3.1 Lifetime Label Construction
		12.3.2 Multi-domain Attribute Modeling
		12.3.3 Bearing Remaining Life Assessment Modeling
	12.4 Experimental Studies and Analysis
		12.4.1 IMS Full-Life Data
		12.4.2 XJTU-SY Full-Life Data
	12.5 Summary of This Chapter
	References
Part VI Multi-source Information Fusion-Based Intelligent Diagnosis and Maintenance
13 Intelligent Fault Diagnosis Method Based on Multi-source Data and Multi-feature Fusion
	13.1 Introduction
	13.2 Theory Background
		13.2.1 Fusion Convolution
		13.2.2 Structure of the Dense Convolutional Network
	13.3 The Framework of the Proposed Fault Diagnosis Method
		13.3.1 Structure of the Proposed MDAAN
		13.3.2 Domain Adaptation Classification Loss
	13.4 Experiments
		13.4.1 Data Description
		13.4.2 Comparison Methods and Training Setting
		13.4.3 CWRU Dataset Results and Analysis
		13.4.4 XJTU Dataset Results and Analysis
	13.5 Conclusion
	References
14 D-S Evidence Theory and Its Application for Fault Diagnosis of Machinery
	14.1 Introduction
	14.2 D-S Evidence Theory and Its Classical Improvement
		14.2.1 Traditional D-S Theory
		14.2.2 Defects of D-S Evidence Theory
		14.2.3 Classical Improvement of D-S Evidence Theory
	14.3 Application of D-S Evidence Theory in Fault Diagnosis of Gas Turbines
		14.3.1 Gas Turbine System Fault Diagnosis
		14.3.2 Rotor System Fault Diagnosis
		14.3.3 Bearing System Fault Diagnosis
	14.4 Summary and Outlook
	References
Part VII Overall Review and Prospects of Future Research Issues
15 Conclusion, Challenges, and Future Work
	15.1 Conclusion
	15.2 Challenges
	15.3 Future Work
	References