Atjaunināt sīkdatņu piekrišanu

E-grāmata: Machine Learning and Knowledge Discovery for Engineering Systems Health Management [Taylor & Francis e-book]

Edited by (University of Illinois at Urbana-Champaign, USA), Edited by (Verizon, California, USA)
Citas grāmatas par šo tēmu:
  • Taylor & Francis e-book
  • Cena: 195,66 €*
  • * this price gives unlimited concurrent access for unlimited time
  • Standarta cena: 279,51 €
  • Ietaupiet 30%
Machine Learning and Knowledge Discovery for Engineering Systems Health ManagementPalielināt
Citas grāmatas par šo tēmu:
Taylor & Francis e-booksMore info about Taylor & Francis e-books
Machine Learning and Knowledge Discovery for Engineering Systems Health Management presents state-of-the-art tools and techniques for automatically detecting, diagnosing, and predicting the effects of adverse events in an engineered system. With contributions from many top authorities on the subject, this volume is the first to bring together the two areas of machine learning and systems health management.

Divided into three parts, the book explains how the fundamental algorithms and methods of both physics-based and data-driven approaches effectively address systems health management. The first part of the text describes data-driven methods for anomaly detection, diagnosis, and prognosis of massive data streams and associated performance metrics. It also illustrates the analysis of text reports using novel machine learning approaches that help detect and discriminate between failure modes. The second part focuses on physics-based methods for diagnostics and prognostics, exploring how these methods adapt to observed data. It covers physics-based, data-driven, and hybrid approaches to studying damage propagation and prognostics in composite materials and solid rocket motors. The third part discusses the use of machine learning and physics-based approaches in distributed data centers, aircraft engines, and embedded real-time software systems.

Reflecting the interdisciplinary nature of the field, this book shows how various machine learning and knowledge discovery techniques are used in the analysis of complex engineering systems. It emphasizes the importance of these techniques in managing the intricate interactions within and between the systems to maintain a high degree of reliability.
List of Figures
xi
List of Tables
xxi
Machine Learning and Knowledge Discovery for Engineering Systems Health Management xxiii
Ashok N. Srivastava
Jiawei Han
Editors xxxiii
Contributors xxxv
SECTION I Data-Driven Methods for Systems Health Management
Chapter 1 Mining Data Streams: Systems and Algorithms
3(36)
Charu C. Aggarwal
Deepak S. Turaga
1.1 Introduction
4(3)
1.2 Stream Processing and Mining Challenges
7(2)
1.2.1 Stream Data Management
7(1)
1.2.2 Relational Data Processing on Streams
7(1)
1.2.3 Stream Indexing
8(1)
1.2.4 Stream Mining Algorithms
8(1)
1.3 Stream Processing Systems: Architectural Issues
9(3)
1.4 Stream Data Reduction
12(10)
1.4.1 Broad Applicability
12(1)
1.4.2 One-Pass Constraint
12(1)
1.4.3 Time and Space Efficiency
13(1)
1.4.4 Data Stream Evolution
13(1)
1.4.5 Sampling
13(2)
1.4.6 Sketches
15(2)
1.4.7 Quantization
17(2)
1.4.8 Transform Domain Summarization
19(2)
1.4.9 Summary Statistic Computation
21(1)
1.4.10 Dimensionality Reduction and Forecasting in Data Streams
21(1)
1.5 Stream Mining Algorithms
22(8)
1.5.1 Data Stream Clustering
22(2)
1.5.2 Data Stream Classification
24(1)
1.5.2.1 VFDT Method
25(1)
1.5.2.2 On-Demand Classification
25(1)
1.5.2.3 Ensemble-Based Classification
26(1)
1.5.3 Frequent Pattern Mining
26(1)
1.5.3.1 Entire Data Stream Model
26(1)
1.5.3.2 Sliding Window Model
27(1)
1.5.3.3 Damped Window Model
27(1)
1.5.4 Change Detection in Data Streams
28(1)
1.5.4.1 Velocity Density Estimation
28(1)
1.5.4.2 Stream Cube Analysis of Multidimensional Streams
29(1)
1.5.4.3 Distributed Mining of Data Streams
29(1)
1.6 Optimizing Stream Mining Applications
30(1)
1.7 Conclusions and Research Directions
31(8)
References
32(7)
Chapter 2 A Tutorial on Bayesian Networks for Systems Health Management
39(28)
Arthur Choi
Lu Zheng
Adnan Darwiche
Ole J. Mengshoel
2.1 Introduction
40(1)
2.2 Systems Health Management and Uncertainty
41(3)
2.2.1 An Electrical Circuit Example
42(2)
2.3 Bayesian Networks
44(5)
2.4 Modeling with Bayesian Networks
49(1)
2.5 Reasoning with Bayesian Networks
50(4)
2.5.1 Posterior Marginals
51(1)
2.5.2 Explanations
51(2)
2.5.3 Sensitivity Analysis
53(1)
2.6 Reasoning Algorithms
54(2)
2.7 Learning Bayesian Networks
56(2)
2.7.1 Learning a Bayesian Network's Parameters
57(1)
2.7.2 Learning with Complete Data
58(1)
2.7.3 Learning with Incomplete Data
58(1)
2.8 Applications and Scalability Experiments
58(4)
2.8.1 Electrical Power Systems
59(1)
2.8.2 Bayesian Network Models
60(1)
2.8.3 Experiments with Real-World Data
60(1)
2.8.4 Experiments with Synthetic Data
61(1)
2.8.5 Discussion
62(1)
2.9 Conclusion
62(5)
References
64(3)
Chapter 3 Anomaly Detection in a Fleet of Systems
67(48)
Nikunj Oza
Santanu Das
3.1 What Is a Fleet?
68(1)
3.2 Background
69(10)
3.2.1 Past Work in Fleet Health Monitoring
69(1)
3.2.1.1 Numeric Data
69(3)
3.2.1.2 Text
72(3)
3.2.2 Role of Data Mining in Fleet Health Monitoring
75(1)
3.2.2.1 Anomaly Detection Problem
76(1)
3.2.2.2 Some Preliminaries
76(3)
3.3 Key Issues in Fleet Health Monitoring
79(4)
3.3.1 Heterogeneity
79(3)
3.3.2 Efficiency: Large Data Volume, Distributed Data
82(1)
3.4 Dealing with Heterogeneity
83(21)
3.4.1 Orca
85(1)
3.4.2 GritBot
86(1)
3.4.3 Kernel-Based Anomaly Detection
87(1)
3.4.3.1 Kernel Theory and Operations
88(1)
3.4.3.2 Is It a Mercer Kernel?
89(1)
3.4.3.3 Problem-Specific Kernel Functions
90(2)
3.4.3.4 Information Fusion
92(1)
3.4.3.5 One-Class SVMs: An Overview
93(2)
3.4.3.6 An Example of Multiple-Kernel Anomaly Detection
95(3)
3.4.4 Text Mining
98(1)
3.4.4.1 Data and Algorithms
98(3)
3.4.4.2 Results
101(3)
3.5 Dealing with Efficiency Issues: Distributed Anomaly Detection
104(6)
3.5.1 Gaussian Mixture Model
105(3)
3.5.2 Distance Based
108(2)
3.6 Conclusions
110(5)
References
111(4)
Chapter 4 Discriminative Topic Models
115(32)
Hanhuai Shan
Amrudin Agovic
Arindam Banerjee
4.1 Introduction
116(2)
4.2 Latent Dirichlet Allocation and Supervised Latent Dirichlet Allocation
118(3)
4.2.1 Latent Dirichlet Allocation
118(2)
4.2.2 Supervised Latent Dirichlet Allocation
120(1)
4.3 Discriminative Latent Dirichlet Allocation
121(3)
4.4 Inference and Parameter Estimation
124(7)
4.4.1 Variational Approximation
124(3)
4.4.1.1 Inference
127(1)
4.4.1.2 Parameter Estimation
128(2)
4.4.2 Variational EM Algorithm
130(1)
4.4.3 Prediction
131(1)
4.5 Experimental Results on ASRS Data Sets
131(13)
4.5.1 DLDA Properties
132(1)
4.5.1.1 Discriminative Latent Dirichlet Allocation versus Latent Dirichlet Allocation
132(3)
4.5.1.2 Fast DLDA versus Other Classification Algorithms
135(2)
4.5.2 Fast DLDA on ASRS
137(1)
4.5.2.1 Topics from Fast DLDA
137(1)
4.5.2.2 Relationship between Classes and Topics
138(4)
4.5.2.3 Classification Results
142(2)
4.6 Conclusion
144(3)
References
145(2)
Chapter 5 Prognostic Performance Metrics
147(34)
Kai Goebel
Abhinav Saxena
Sankalita Saha
Bhaskar Saha
Jose Celaya
5.1 Introduction
148(2)
5.2 Background
150(5)
5.2.1 Prediction Categorization
150(1)
5.2.1.1 Forecasting
150(1)
5.2.1.2 Prognostics
151(1)
5.2.2 Prediction Methods
152(1)
5.2.3 Performance Evaluation Methods
152(3)
5.3 Metrics for Prognostic Applications
155(19)
5.3.1 Certification Metrics
155(1)
5.3.2 Cost-Benefit Metrics
155(1)
5.3.2.1 MTBF-to-MTBUR Ratio Method
155(1)
5.3.2.2 Life Cycle Cost
155(1)
5.3.2.3 Return on Investment
155(1)
5.3.2.4 Technical Value
155(1)
5.3.2.5 Total Value
156(1)
5.3.3 Metrics for Computational Performance
156(1)
5.3.4 Metrics for Reliability Analysis
156(1)
5.3.4.1 Constant Rate Reliability Metrics
157(1)
5.3.4.2 Probability of Success Metrics
157(1)
5.3.5 Metrics for Prognostics Algorithm Performance
157(1)
5.3.5.1 Challenges
158(2)
5.3.6 Error-Based Metrics
160(1)
5.3.6.1 FP, FN, and ROC
160(2)
5.3.6.2 Spread-Based Metrics
162(1)
5.3.6.3 Anomaly Correlation Coefficient
162(1)
5.3.6.4 Prognostic Horizon
163(1)
5.3.6.5 α-λ Performance
164(1)
5.3.6.6 Relative Accuracy
165(1)
5.3.6.7 Cumulative Relative Accuracy
166(1)
5.3.6.8 Convergence
166(1)
5.3.6.9 Robustness
167(1)
5.3.6.10 RUL Online Precision Index
168(1)
5.3.7 Incorporating Uncertainty Estimates
169(2)
5.3.8 Guidelines for Applying Prognostics Metrics
171(1)
5.3.8.1 Guidelines on Choosing Performance Parameters
172(1)
5.3.8.2 Guidelines for Dealing with Uncertainties
172(1)
5.3.8.3 Guidelines to Resolve Ambiguities
173(1)
5.4 Summary
174(7)
Acknowledgments
174(1)
References
174(7)
SECTION II Physics-Based Methods for Systems Health Management
Chapter 6 Gaussian Process Damage Prognosis under Random and Flight Profile Fatigue Loading
181(22)
Aditi Chattopadhyay
Subhasish Mohanty
6.1 Introduction
182(1)
6.2 Theoretical Approach
183(8)
6.2.1 Physics-Based Fatigue Damage Prognosis Model
183(2)
6.2.2 Bayesian Framework for Damage Prediction
185(1)
6.2.3 Damage Prediction Using GP Regression
186(1)
6.2.3.1 Covariance Matrix to Kernel Matrix
187(1)
6.2.3.2 Mean and Variance of the Predicted Damage at (n + 1)th Damage Level
188(1)
6.2.3.3 Kernel Function Selection
189(1)
6.2.3.4 Hyperparameter Determination
189(2)
6.3 Numerical Results and Discussion
191(8)
6.3.1 Fatigue Experiment and Data Collection
191(3)
6.3.2 GP Input-Output Data
194(1)
6.3.3 Future Damage State Prediction
195(1)
6.3.3.1 Case I
195(2)
6.3.3.2 Case II
197(1)
6.3.3.3 Case III
198(1)
6.4 Conclusion
199(4)
References
200(3)
Chapter 7 Bayesian Analysis for Fatigue Damage Prognostics and Remaining Useful Life Prediction
203(30)
Xuefei Guan
Yongming Liu
7.1 Introduction
204(2)
7.2 Probabilistic Modeling for Hierarchical Uncertainties
206(4)
7.2.1 Model Choice Uncertainty
206(2)
7.2.2 Parameter, Mechanism, and Measurement Uncertainties
208(2)
7.3 An Efficient Algorithm for Continuous Bayesian Updating
210(5)
7.3.1 Bayesian Updating with MDI Reparameterization
211(3)
7.3.2 The Updating Algorithm
214(1)
7.4 MCMC Methodology in the General State Space
215(5)
7.4.1 The M-H Algorithms in the General State Space
216(3)
7.4.2 A Factorized M-H Algorithm
219(1)
7.5 Fatigue Damage Prognostics and RUL Prediction
220(11)
7.5.1 Component Experimental Data
220(1)
7.5.2 Fatigue Crack Growth Models
221(3)
7.5.3 Bayesian Updating for Crack Growth Prognostics and RUL Predictions
224(2)
7.5.4 Model Probabilities, Bayes Factors, and Parameter Statistics
226(2)
7.5.5 Bayesian Model Averaging
228(1)
7.5.6 Comparisons of the Overall Performance and Efficiency
229(1)
7.5.7 Discussion
230(1)
7.6 Conclusion
231(2)
Appendices
233(106)
Appendix 7.1 Proof of Detailed Balance Equation of the Factorized M-H Algorithm in Section 7.4.2 and Demonstration Examples
233(4)
Appendix 7.2 Demonstration of the Performance and Efficiency of the Proposed Bayesian Updating Algorithm with MDI Reparameterization in Section 7.3
237(8)
References
240(5)
Chapter 8 Physics-Based Methods of Failure Analysis and Diagnostics in Human Space Flight
245(40)
Vadim N. Smelyanskiy
Dmitry G. Luchinsky
Vasyl V. Hafiychuk
Viatcheslav V. Osipov
Igor Kulikov
Ann Patterson-Hine
8.1 Introduction
246(1)
8.2 Dynamical Inference of Stochastic Nonlinear Models
247(5)
8.3 The Lorenz System
252(3)
8.3.1 Parameter Estimation with Strong Dynamical Noise
252(2)
8.3.2 Model Reconstruction with Strong Dynamical Noise
254(1)
8.4 The Three Tank Problem
255(6)
8.5 In-Flight Decision Support for SRMS
261(7)
8.5.1 Internal Ballistics of SRMs
261(2)
8.5.2 Estimation of the Parameters of Nozzle Blocking
263(1)
8.5.3 Predicting "Misses" in the Fault Detection
264(4)
8.6 Modal Dynamics Based Damage Detection
268(6)
8.6.1 Mathematical Model of Pristine Plate
268(3)
8.6.2 Mathematical Model of Damaged Plate
271(3)
8.7 Dynamical Inference of a Set of Coupled Oscillators
274(5)
8.7.1 General Inferential Framework for a Set of Coupled Oscillators
274(3)
8.7.2 Numerical Example
277(2)
8.8 Conclusion
279(6)
References
282(3)
Chapter 9 Model-Based Tools and Techniques for Real-Time System and Software Health Management
285(54)
Sherif Abdelwahed
Abhishek Dubey
Gabor Karsai
Nagabhushan Mahadevan
9.1 Introduction
286(2)
9.2 Related Work
288(3)
9.2.1 Failure Propagation Models
288(1)
9.2.2 Fault Recovery
289(1)
9.2.3 Fault Detection and Health Management of Software
290(1)
9.3 Fault Diagnostics using Timed Failure Propagation Graphs
291(20)
9.3.1 The Timed Failure Propagation Graph Model
291(3)
9.3.2 Reasoning Algorithm
294(2)
9.3.2.1 Hypothesis Ranking
296(1)
9.3.2.2 Reasoner Performance
297(1)
9.3.3 TFPG Examples
298(3)
9.3.4 Distributed Reasoning
301(1)
9.3.4.1 Overview
302(1)
9.3.4.2 Extensions to the TFPG Model
302(2)
9.3.4.3 Extensions to the Reasoner
304(1)
9.3.4.4 Synchronous Event Processing
305(1)
9.3.4.5 Asynchronous Event Processing
306(1)
9.3.5 Distributed TFPG Examples
307(3)
9.3.5.1 Discussion
310(1)
9.4 Application of TFPG for Diagnosing Software Failures
311(15)
9.4.1 ARINC Component Framework
312(2)
9.4.2 Health Management in ACM
314(1)
9.4.2.1 Component-Level Health Management
314(1)
9.4.2.2 System-Level Health Manager
314(1)
9.4.2.3 Component-Level Detection
314(2)
9.4.2.4 Component-Level Mitigation
316(1)
9.4.3 Software Fault Propagation Model
317(5)
9.4.3.1 Complexity of the Generated Model
322(1)
9.4.4 The Diagnosis Process
323(3)
9.5 Application of TFPG for Prognostics of Impending Faults
326(5)
9.5.1 Failure Criticality
326(2)
9.5.2 State Estimation Plausibility
328(1)
9.5.3 Time Proximity
329(1)
9.5.4 Time to Criticality
330(1)
9.6 Relation to Machine Learning and Data Mining
331(1)
9.7 Summary
332(7)
Acknowledgments
332(1)
References
332(7)
SECTION III Applications
Chapter 10 Real-Time Identification of Performance Problems in Large Distributed Systems
339(24)
Moises Goldszmidt
Dawn Woodard
Peter Bodik
10.1 Introduction
340(2)
10.2 Problem Definition
342(1)
10.3 From Collected Signals to Fingerprints
343(4)
10.3.1 Summarizing the State of the Datacenter
344(1)
10.3.2 Crisis Modeling
345(1)
10.3.3 Selecting the Relevant Signals
346(1)
10.4 Identifying a Crisis
347(5)
10.4.1 Cluster Modeling
348(1)
10.4.2 Computing the Probability of the Crisis Label
349(3)
10.4.3 Prior Specification
352(1)
10.5 Experiments and Results
352(5)
10.5.1 System Under Study and Data
353(2)
10.5.2 Experiments
355(1)
10.5.3 Offline Clustering
355(2)
10.5.4 Operational Setting
357(1)
10.6 Discussion
357(2)
10.7 Conclusions
359(4)
Acknowledgments
359(1)
References
359(4)
Chapter 11 A Combined Model-Based and Data-Driven Prognostic Approach for Aircraft System Life Management
363(32)
Marcos Orchard
George Vachtsevanos
Kai Goebel
11.1 Introduction
364(2)
11.2 An Integrated Fault Diagnosis and Failure Prognosis Architecture
366(10)
11.2.1 Sensing and Data Processing
368(3)
11.2.2 Selection and Extraction of CIs
371(3)
11.2.3 Diagnostics and Prognostics Modules
374(2)
11.3 PF Algorithms in a Combined Model-Based/Data-Driven Framework for Failure Prognosis
376(9)
11.3.1 PF Algorithms and Failure Prognosis
377(4)
11.3.2 Uncertainty Measure-Based Feedback Loops for the Extension of Remaining Useful Life
381(2)
11.3.2.1 DS-Based Approach to RUL Extension
383(1)
11.3.2.2 CIS-Based Approach to Rule Extension
384(1)
11.4 Case Study: Load Reduction and Effects on Fatigue Crack Growth in Aircraft Components
385(7)
11.5 Conclusions
392(3)
References
392(3)
Chapter 12 Hybrid Models for Engine Health Management
395(28)
Allan J. Volponi
Ravi Rajamani
12.1 Introduction
395(2)
12.2 Background
397(5)
12.3 Hybrid Model Process
402(5)
12.4 Example
407(4)
12.5 Verification and Validation
411(3)
12.6 Transient Diagnostics and Anomaly Detection
414(1)
12.7 V&V of Diagnostic Algorithms
415(2)
12.8 Validation
417(2)
12.9 Software Specifications and Design Descriptions
419(1)
12.10 Verification
420(1)
12.11 Summary
421(2)
Acknowledgment
421(1)
References
422(1)
Chapter 13 Extracting Critical Information from Free Text Data for Systems Health Management
423(28)
Anne Kao
Stephen Poteet
David Augustine
13.1 Introduction
424(1)
13.2 Problem Description
425(2)
13.3 Related Work
427(2)
13.4 Partname Matching by Analysis of Text Characteristics
429(10)
13.4.1 The Part Name Reference Problem
429(2)
13.4.2 Our Solution
431(4)
13.4.3 Experiments
435(1)
13.4.4 Results
436(3)
13.5 QUBIT
439(6)
13.5.1 The Ad Hoc Query Problem
439(1)
13.5.2 Suggester
439(1)
13.5.2.1 Finding Terms with a High Co-occurrence with the Search Term
440(1)
13.5.2.2 Using Fuzzy String Matching Algorithms to Generate a Potential Match
441(1)
13.5.2.3 Using Regular Expressions to Generate a Potential Match
441(1)
13.5.2.4 Using Latent Semantic Analysis or Other Methods that Generate Topic Relateness
442(1)
13.5.2.5 Using a Knowledge Base
443(1)
13.5.2.6 Using Suggestions Entered by Peers
443(1)
13.5.3 Query Builder
444(1)
13.6 Future Research Directions
445(2)
13.7 Conclusion
447(4)
References
448(3)
Index 451
Ashok N. Srivastava is the Principal Scientist for Data Mining and Systems Health Management at NASA. Dr. Srivastava has received many awards, including the IEEE Computer Society Technical Achievement Award, the NASA Exceptional Achievement Medal, NASA Group Achievement Awards, the IBM Golden Circle Award, and a U.S. Department of Education Merit Fellowship. His current research focuses on the development of data mining algorithms for anomaly detection in massive data streams, kernel methods in machine learning, and text mining algorithms.

Jiawei Han is an Abel Bliss Professor of Computer Science at the University of Illinois. He is also the Director of the Information Network Academic Research Center, which is supported by the U.S. Army Research Lab. A fellow of ACM and IEEE, Dr. Han has received numerous honors, including IEEE W. Wallace McDowell Award, IEEE Computer Society Technical Achievement Award, ACM SIGKDD Innovation Award, IBM Faculty awards, and HP Innovation awards. His research interests include data mining, information network analysis, and database systems.
Permanent link: https://www.kriso.lv/db/9780429151309_pe.html
Keywords: