Predictive Maintenance in Dynamic Systems - Advanced Methods, Decision Support Tools and Real-World Applications

Preface

Contents

Contributors

Prologue: Predictive Maintenance in Dynamic Systems

1 From Predictive to Preventive Maintenance in Dynamic Systems: Motivation, Requirements, and Challenges

2 Components and Methodologies for Predictive Maintenance

2.1 Models as Backbone Component

2.2 Methods and Strategies to Realize Predictive Maintenance

3 Beyond State-of-the-Art—Contents of the Book

References

Smart Devices in Production System Maintenance

1 Introduction

2 State of the Art

2.1 Definition of Terms

2.2 Physical Devices/Hardware

2.2.1 Smartphones and Tablets

2.2.2 Smartglasses

2.2.3 Smartwatches

2.3 Market View

2.4 Device Selection and Potentials

3 Application Examples in Maintenance

3.1 Local Data Analysis and Communication for Condition Monitoring

3.2 Remote Expert Solutions

3.3 Process Data Visualization for Process Monitoring

4 Limitations and Challenges

4.1 Hardware Limitations

4.2 User Acceptance

4.3 Information Compression on Smart Devices

4.4 Legal Aspects

5 Summary

References

On the Relevance of Preprocessing in Predictive Maintenance for Dynamic Systems

1 Introduction

2 Preprocessing

2.1 Taxonomy

2.2 Data Cleansing

2.2.1 Outlier Detection Based on Mahalanobis Distance

2.2.2 Outlier Detection Based on ?2 Approximations of Q and T2 Statistics

2.3 Data Normalization

2.4 Data Transformation

2.4.1 Statistical Transformations

2.4.2 Signal Processing

2.5 Missing Values Treatment

2.6 Data Engineering

2.6.1 Feature Selection

2.6.2 Feature Extraction

2.6.3 Feature Discretization

2.7 Imbalanced Data Treatment

2.7.1 Oversampling

2.7.2 Undersampling

2.7.3 Mixed Sampling

2.8 Models

2.8.1 Classification

2.8.2 Regression

3 Experimentation

3.1 Datasets

3.1.1 PHM Challenge 2014

3.1.2 PHM Challenge 2016

3.2 Experimental Schema

3.3 Results

4 Conclusions

References

Part I Anomaly Detection and Localization

A Context-Sensitive Framework for Mining Concept Drifting Data Streams

1 Concept Drifting Data Streams

1.1 Concept Drift

2 A Novel Framework for Online Learning in Adaptive Mode

2.1 Basic Components

2.2 Optimizing for Stream Volatility and Speed

3 Implementation of a Context-Sensitive Staged Learning Framework

3.1 The Use of the Discrete Fourier Transform in Classification and Concept Encoding

3.2 Repository Management

3.3 The Staged Learning Approach

3.3.1 Transition Between Stages

3.4 Space and Time Complexity of Spectral Learning

4 Empirical Study

4.1 Datasets Used for the Empirical Study

4.1.1 Synthetic Data

4.1.2 Synthetic Data Recurring with Noise

4.1.3 Synthetic Data Recurring with a Progressively Increasing Pattern of Drift

4.1.4 Synthetic Data Recurring with an Oscillating Drift Pattern

4.1.5 Real-World Data

4.2 Parameter Values

4.3 Effectiveness of Staged Learning Approach

4.4 Accuracy Evaluation

4.4.1 ARF vs SOL Accuracy of a Concept

4.5 Throughput Evaluation

4.6 Accuracy Versus Throughput Trade-Off

4.7 Memory Consumption Evaluation

5 Sensitivity Analysis

6 Conclusion

7 Future Research

References

Online Time Series Changes Detection Based on Neuro-FuzzyApproach

1 Introduction

2 Fuzzy Online Segmentation-Clustering

2.1 Probabilistic Approach

2.2 Possibilistic Approach

2.3 Online Combined Approach

2.4 Robust Approach

3 Robust Forecasting and Faults Detection in Nonstationary Time Series

4 Conclusions

References

Early Fault Detection in Reciprocating Compressor Valves by Means of Vibration and pV Diagram Analysis

1 Introduction

2 Problem Statement

2.1 Reciprocating Compressor Operation

2.2 Problem Statement

3 Vibration Analysis

3.1 Motivation

3.2 Feature Extraction

3.3 Feature Space

4 Analysis of the pV Diagram

4.1 Motivation

4.2 Feature Extraction

4.3 Feature Space

4.4 Classification

5 Experimental Setup

5.1 Compressor Test Bench

5.2 Test Runs

6 Results

6.1 Vibration Analysis

6.2 pV Diagram Analysis

7 Conclusions

References

A New Hilbert-Huang Transform Technique for Fault Detection in Rolling Element Bearings

1 Introduction

2 Minimum Entropy Deconvolution Filter

3 The Proposed eHT Technique for Bearing Fault Detection

3.1 Brief Discussion of Mathematical Morphology Analysis

3.1.1 Structural Element (SE)

3.1.2 Dilation and Erosion

3.1.3 Closing and Opening

3.2 The Proposed Morphological Filter

3.3 The Proposed eHT Technique

4 Application of the Proposed eHT Technique for Bearing Fault Detection

4.1 Experimental Setup and Instrumentations

4.2 Performance Evaluation

4.2.1 Validation of Morphological-Based Filtering Technique

4.2.2 Validation of the Normality Measure

4.3 Evaluation of the Proposed eHT Technique

4.3.1 Condition Monitoring of a Healthy Bearing

4.3.2 Outer Race Fault Detection

4.3.3 Inner Race Fault Detection

4.3.4 Rolling Element Fault Detection

5 Conclusion

References

Comparison of Genetic and Incremental Learning Methods for Neural Network-Based Electrical Machine Fault Detection

1 Introduction

2 Electrical Machine Fault Detection

3 Genetic Algorithm for Neural Network Learning

3.1 Initialization and Parameterization

3.2 Phenotype Representation

3.3 Recombination Operator

3.3.1 Arithmetic Crossover

3.3.2 Multipoint Crossover

3.3.3 Local Intermediate Crossover

3.4 Mutation Operator

3.4.1 Gaussian Mutation

3.4.2 Random Mutation

3.4.3 Post-Processing Based on Local Random Mutation

3.5 Fitness Function

3.6 Selection Operator

3.6.1 Tournament Selection

3.6.2 Elitism

3.7 Stopping Criteria

4 Incremental Algorithm for Neurofuzzy Network Learning

4.1 Numerical and Fuzzy Data

4.2 Network Architecture

4.3 Fuzzy Neuron

4.3.1 Triangular Norm and Conorm

4.3.2 Neuron Model

4.4 Granular Region

4.5 Granularity Adaptation

4.6 Developing Granules

4.7 Adapting Connection Weights

4.8 Learning Algorithm

5 Results and Discussion

5.1 Preliminaries

5.2 Genetic EANN for Fault Detection

5.3 Incremental EGNN for Fault Detection

5.4 Comparative Analyses and Discussion

6 Conclusion

References

Evolving Fuzzy Model for Fault Detection and Fault Identification of Dynamic Processes

1 Introduction

2 Evolving Fuzzy Model

2.1 Fuzzy Cloud-Based Model Structure

2.2 Evolving Mechanism

3 Fault Detection and Identification

3.1 Learning/Training Phase

3.2 Fault Detection Phase

3.3 Fault Identification Phase

4 Description of the HVAC Process Model

4.1 Possible Faults on HVAC System

5 Experimental Results

6 Conclusion

References

An Online RFID Localization in the Manufacturing Shopfloor

1 Introduction

2 RFID Localization System

3 eT2QFNN Architecture

3.1 Input Layer

3.2 Quantum Layer

3.3 Rule Layer

3.4 Output Processing Layer

3.5 Output Layer

4 eT2QFNN Learning Policy

4.1 Rule Growing Mechanism

4.2 Parameter Adjustment

4.2.1 Fuzzy Rule Initialization

4.2.2 Winning Rule Update

5 Experiments and Data Analysis

5.1 Experiment Setup

5.2 Comparison with Existing Results

6 Conclusions

References

Part II Prognostics and Forecasting

Physical Model-Based Prognostics and Health Monitoring to Enable Predictive Maintenance

1 Introduction

2 Challenges in the Field of Predictive Maintenance

2.1 Combining Diagnosis and Prognosis

2.2 System Versus Component Level

2.3 Monitoring of Usage, Loads, Condition or Health

2.4 Interpretation of Monitoring Data

2.5 Data-Driven or Model-Based Prognostics

2.6 Selection of Most Suitable Approach and Technique

2.7 Data Quality

3 Structural Health and Condition Monitoring

3.1 Sensors

3.2 Vibration and Vibration-Based Monitoring

4 Physical Model-Based Prognostics

4.1 Relation Between Usage, Loads and Degradation Rate

4.2 Developing a Prognostic Method

4.3 Comparison to Data-Driven Approaches

5 Decision Support Tools

5.1 Guidelines for Selecting Suitable Approach

5.2 Critical Part Selection

6 Case Studies

6.1 Maritime Systems

6.2 Railway Infrastructure

6.3 Wind Turbines

7 Conclusions

References

On Prognostic Algorithm Design and Fundamental Precision Limits in Long-Term Prediction

1 Introduction

2 Cramér–Rao Lower Bounds

2.1 Bayesian Cramér–Rao Lower Bounds

2.2 BCRLBs for Discrete-Time Dynamical Systems

3 Methodology for Prognostic Algorithm Design

3.1 Conditional Predictive Bayesian Cramér–Rao Lower Bounds

3.2 Analytic Computation of MCP-BCRLBs

4 Case Study: End-of-Discharge Time Prognosis of Lithium-Ion Batteries

4.1 State-Space Model

4.2 Prognostic Algorithm

4.3 Avoiding Monte Carlo Simulations in EoD Prognostic Algorithms

4.4 Prognostic Algorithm Design: Known Future Operating Profiles

4.5 Prognostic Algorithm Design: Statistical Characterizations of Future Operating Profiles

5 Conclusions

Acronyms

References

Performance Degradation Monitoring and Quantification: A Wastewater Treatment Plant Case Study

1 Introduction

1.1 Energy Consumption on WWTPs

1.2 Energy Savings Through Maintenance

2 Methodology

3 Results

3.1 External Recirculation Pumping System

3.1.1 Experimental Setup

3.1.2 Application of the Methodology

3.1.3 Results and Discussion

3.2 Plant Input Pumping System

3.2.1 Experimental Setup

3.2.2 Application of the Methodology

3.2.3 Results and Discussion

3.3 Aeration System Blowers

3.3.1 Experimental Setup

3.3.2 Application of the Methodology

3.3.3 Results and Discussion

4 Conclusions and Future Works

References

Fuzzy Rule-Based Modeling for Interval-Valued Data: An Application to High and Low Stock Prices Forecasting

1 Introduction

2 Interval Arithmetic

3 Interval Fuzzy Rule-Based Model

3.1 Interval Participatory Learning Fuzzy Clustering with Adaptive Distances

3.2 Rules Consequent Parameters Identification

3.3 iFRB Identification Procedure

4 Computational Experiments

4.1 Performance Assignment

4.2 Empirical Results

5 Conclusion

References

Part III Diagnosis, Optimization and Control

Reasoning from First Principles for Self-adaptive and Autonomous Systems

1 Introduction

2 Example

3 Model-Based Reasoning

3.1 Model-Based Diagnosis

3.2 Abductive Diagnosis

3.3 Summary on Model-Based Reasoning for Diagnosis

4 Modeling for Diagnosis and Repair

5 Self-adaptation Using Models

6 Related Research

7 Conclusions

References

Decentralized Modular Approach for Fault Diagnosis of a Class of Hybrid Dynamic Systems: Application to a Multicellular Converter

1 Learning from Data Streams

1.1 Basic Definitions and Motivation

1.2 State of the Art

1.3 Contribution of the Proposed Approach

2 Proposed Approach

2.1 System Decomposition

2.2 Discrete Component Modeling

2.3 Residual Generation Based on System Continuous Dynamics

2.4 Enriched Local Models Building

2.5 Local Hybrid Diagnoser Construction

2.6 Equivalence Between Centralized and Decentralized Diagnosis Structures

2.7 Computation Complexity Analysis

3 Experimental Results

4 Conclusion

References

Automated Process Optimization in Manufacturing Systems Based on Static and Dynamic Prediction Models

1 Introduction

1.1 Our Approach

2 Problem Statement

2.1 Process Optimization Based on Parameters

2.2 Process Optimization Based on Process Values Trends

3 Establishment of Predictive Models

3.1 Iterative Construction of Static Predictive Mappings (Parameters Quality)

3.1.1 Expert Knowledge Initialization

3.1.2 Hybrid Design of Experiments (HDoE)

3.1.3 Predictive Mapping Models Construction

3.2 Time-Series-Based Forecast Models (Process Values Quality) Learning and Adaptation

3.2.1 Learning by a Nonlinear (Fuzzy) Version of PLS (PLS-Fuzzy)

3.2.2 On-Line Model Adaptation with Increased Flexibility

4 Process Optimization with Predictive Models

4.1 Static Case (Mappings as Surrogates)

4.1.1 Evolutionary Algorithms for Solving Many-Objective Optimization Problems

4.1.2 A New Efficient Method for Multi-Objective EA (DECMO2)

4.2 Dynamic Case (Time-Series-Based Forecast Models as Surrogates)

4.2.1 Optimization Strategies

4.2.2 Reducing Dimensionality of the Optimization Space

5 Some Results from a Chip Production Process

5.1 Application Scenario

5.2 Experimental Setup and Data Collection

5.3 Results

5.3.1 Static Phase (Based on Process Parameter Settings)

5.3.2 Dynamic Case (Based on Time-Series of Process Values)

6 Conclusion and Outlook

References

Distributed Chance-Constrained Model Predictive Control for Condition-Based Maintenance Planning for Railway Infrastructures

1 Introduction

2 Preliminaries

2.1 Hybrid and Distributed MPC

2.2 Chance-Constrained MPC

3 Problem Formulation

3.1 Deterioration Model

3.2 Local Chance-Constrained MPC Problem

3.3 Two-Stage Robust Scenario-Based Approach

3.4 MLD-MPC Problem

4 Distributed Optimization

4.1 Dantzig-Wolfe Decomposition

4.2 Constraint Tightening

5 Case Studies

5.1 Settings

5.2 Representative Run

5.3 Computational Comparisons

5.4 Comparison with Alternative Approaches

6 Conclusions and Future Work

Appendix

Parameters for Case Study

Cyclic Approach

References

Index