Rolf lsermann

Fault-Diagnosis Systems
Rolf Isermann

Fault-Diagnosis Systems
An Introduction from Fault Detection
to Fault Tolerance

With 227 Figures

~Springer
Professor Dr. Rolf Isermann
TU Darmstadt
Institut fi.ir Automatisierungstechnik
Fachgebiet Regelungstechnik und Prozessautomatisierung
Landgraf-Georg-Str. 4
64283 Darmstadt
Germany
risermann@iat.tu-darmstadt.de

Library of Congress Control Number: 2005932861

ISBN 978-3-540-24112-6 Springer Berlin Heidelberg New York

This work is subject to copyright. All rights are reserved, whether the whole or part of the material
is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broad-
casting, reproduction on microfilm or in other ways, and storage in data banks. Duplication of this
publication or parts thereof is permitted only under the provisions of the German Copyright Law of
September 9, 1965, in its current version, and permission for use must always be obtained from Sprin-
ger. Violations are liable to prosecution under German Copyright Law.

Springer is a part of Springer Science+ Business Media

springeronline.com

© Springer-Verlag Berlin Heidelberg 2006

The use of general descriptive names, registered names, trademarks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant pro-
tective laws and regulations and therefore free for general use.

Typesetting: Digital data supplied by author

Final processing by PTP-Berlin Protago-TEX-Production GmbH, Germany
Cover-Design:
Printed on acid-free a er 62/3141/Yu- 54 3 2 1 0
Preface

With increasing demands for efficiency and product quality and progressing integra-
tion of automatic control systems in high-cost and safety-critical processes, the field
of supervision (or monitoring), fault detection and fault diagnosis plays an important
role. The classical way of supervision is to check the limits of single variables and
alarming of operators. However, this can be improved significantly by taking into
account the information hidden in all measurements and automatic actions to keep
the systems in operation.
During the last decades theoretical and experimental research has shown new
ways to detect and diagnose faults. One distinguishes fault detection to recognize
that a fault happened, and fault diagnosis to find the cause and location of the fault.
Advanced methods of fault detection are based on mathematical signal and process
models and on methods of system theory and process modelling to generate fault
symptoms. Fault diagnosis methods use causal fault-symptom-relationships by ap-
plying methods from statistical decision, artificial intelligence and soft computing.
Therefore, efficient supervision, fault detection and diagnosis is a challenging field
by encompassing physical oriented system theory, experiments and computations.
The considered subjects are also known as fault detection and isolation (FDI) or
fault detection and diagnosis (FDD).
A further important field is fault management. This means to avoid shut-downs
by early fault detection and actions like process condition-based maintenance or re-
pair. If sudden faults, failures or malfunctions cannot be avoided, fault-tolerant sys-
tems are required. Through methods of fault detection and reconfiguration of redun-
dant components, break-down and in the case of safety-critical processes accidents
may be avoided.
The book is intended to give an introduction to advanced supervision, fault detec-
tion and diagnosis and fault-tolerant systems for processes with mainly continuous,
sampled signals. Of special interest is an application-oriented approach with methods
which have proven their performance in practical applications.
The material is the result of many own research projects during the last 25
years on fault detection and diagnosis, but also of publications by many other re-
search groups. The development of the field can especially be followed by the IFAC-
VI

Symposia series "SAFEPROCESS", which was initiated 1991 in Baden-Baden and

then repeated all three years in Helsinki, Hull, Budapest, Washington, Beijing, and
the IFAC-Workshop "On-line fault detection and supervision in the chemical indus-
tries" started in Kyoto (1986), and then held in Newark, Newcastle, Folaize and
Cheju, but also by other conferences.
The book is dedicated as an introduction in teaching the field of fault detec-
tion and diagnosis, and fault-tolerant systems for graduate students or students of
higher semesters of electrical and electronic engineering, mechanical and chemical
engineering and computer science. As the treated field is in a phase of increasing
importance for technical and also non-technical systems, it has been tried to present
the material in an easy to understand and transparent way and with realistic perspec-
tives for the application of the treated and discussed methods. Therefore the book
is also oriented towards practising engineers in research and development, design
and manufacturing. Preconditions are basic undergraduate courses of system theory,
automatic control, mechanical and/or electrical engineering.
The author is greatful to his research associates, who performed many theoretical
and practical research projects on the subject of this book since 1975, among them
H. Siebert, L. Billmann, G. Geiger, W. Goedecke, S. Nold, U. Raab, B. Freyermuth,
S. Leonhardt, R. Deibert, T. Hofling, T. Pfeufer, M. Ayoubi, P. Balle, D. Fiissel,
0. Moseler, A. Wolfram, F. Kimmich, A. Schwarte, M. Vogt, M. Miinchhof, D. Fi-
scher, F. Haus and I. Unger. Following chapters or sections were worked out by:
8.4.4: F. Kimmich, 9.2.3: M. Vogt, 10.4.2: P. Balle, 11.4.2: I. Unger, 13: F. Haus,
15.2, 16, 17.3 and 23.2: D. Fiissel. I appreciate these contributions highly as valu-
able inputs to this book.
Finally, I especially would like to thank Brigitte Hoppe for the laborious and pre-
cise text setting, including the figures and tables in camera-ready form.

Darmstadt, February 2005

Rolf Isermann
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Process automation and process supervision . . . . . . . . . . . . . . . . . . . . 1
1.2 Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Historical notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Part I Fundamentals

2 Supervision and fault management of processes - tasks and

terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Basic tasks of supervision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Faults, failures, malfunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Reliability, availability, safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Fault tolerance and redundancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5 Knowledge-based fault detection and diagnosis . . . . . . . . . . . . . . . . . 25
2.5.1 Analytic symptom generation.......................... 26
2.5.2 Heuristic symptom generation . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5.3 Fault diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Implementation issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 Reliability, Availability and Maintainability (RAM) . . . . . . . . . . . . . . . . 31

3.1 Reliability................................................. 31
3.1.1 Type of faults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.2 Reliability estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.3 Connected elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Maintainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Availability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4 Fault management for total life cycles . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Some failure statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.1 Statistics of components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
VIII Contents

3.5.2 Statistics of systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4 Safety, Dependability and System Integrity . . . . . . . . . . . . . . . . . . . . . . . 49

4.1 Reliability analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2 Event tree analysis (ETA) and fault tree analysis (FTA) . . . . . . . . . . . 50
4.3 Failure mode and effects analysis (FMEA) . . . . . . . . . . . . . . . . . . . . . 51
4.4 Hazard-analysis (HA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.5 Risk classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.6 Integrated reliability and safety design . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Part 11 Fault-Detection Methods

5 Process Models and Fault Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5.1 Fault models............................................... 63
5 .1.1 Basic fault models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1.2 Examples for fault models............................. 64
5.2 Process models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2.1 Theoretical and experimental modelling . . . . . . . . . . . . . . . . . 71
5.2.2 Static process models................................. 73
5.2.3 Linear dynamic process models . . . . . . . . . . . . . . . . . . . . . . . . 74
5.2.4 Nonlinear process models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6 Signal models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.1 Harmonic oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.1.1 Single oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.1.2 Superposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.1.3 Amplitude modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.1.4 Frequency and phase modulation . . . . . . . . . . . . . . . . . . . . . . . 85
6.1.5 Beating (Libration) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.1.6 Superposition and nonlinear characteristics. . . . . . . . . . . . . . . 88
6.2 Stochastic signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.2.1 Continuous-time stochastic signals. . . . . . . . . . . . . . . . . . . . . . 89
6.2.2 Discrete-time stochastic signals . . . . . . . . . . . . . . . . . . . . . . . . 91
6.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

7 Fault detection with limit checking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

7.1 Limit checking of absolute values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2 Trend checking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.3 Change detection with binary thresholds . . . . . . . . . . . . . . . . . . . . . . . 97
7.3.1 Estimation of mean and variance . . . . . . . . . . . . . . . . . . . . . . . 97
7 .3.2 Statistical tests for change detection. . . . . . . . . . . . . . . . . . . . . 100
7.4 Change detection with fuzzy thresholds . . . . . . . . . . . . . . . . . . . . . . . . 107
 Contents IX

7.5 Adaptive thresholds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

7.6 Plausibility checks .......................................... 108
7.7 Problems ................................................. 109

8 Fault detection with signal models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

8.1 Analysis of periodic signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
8.1.1 Bandpass filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
8.1.2 Fourier analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
8.1.3 Correlation Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
8.1.4 Fourier transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.1.5 Fast Fourier transformation (FFT) . . . . . . . . . . . . . . . . . . . . . . 117
8.1.6 Maximum entropy spectral estimation . . . . . . . . . . . . . . . . . . . 122
8.1.7 Cepstrum analysis ................................... 127
8.2 Analysis of non-stationary periodic signals . . . . . . . . . . . . . . . . . . . . . 129
8.2.1 Short-time Fourier transform . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.2.2 Wavelet transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.3 Analysis of stochastic signals ................................. 131
8.3.1 Correlation analysis .................................. 131
8.3.2 Spectrum analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.3.3 Signal parameter estimation with ARMA-models . . . . . . . . . 133
8.4 Vibration analysis of machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
8.4.1 Vibrations of rotating machines . . . . . . . . . . . . . . . . . . . . . . . . 134
8.4.2 Vibration signal models ............................... 135
8.4.3 Vibration analysis methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.4.4 Speed signal analysis of combustion engines ............. 141
8.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

9 Fault detection with process-identification methods ................ 147

9.1 Identification with correlation functions ........................ 149
9 .1.1 Estimation of correlation functions . . . . . . . . . . . . . . . . . . . . . 149
9.1.2 Convolution ......................................... 150
9.2 Parameter estimation for linear processes ....................... 151
9.2.1 Method of least squares (LS) .......................... 152
9.2.2 Extended least squares (ELS) method ................... 158
9.2.3 Modifications of basic recursive estimators. . . . . . . . . . . . . . . 159
9.2.4 Parameter estimation of time-varying processes ........... 164
9.2.5 Parameter estimation for continuous-time signals ......... 167
9.2.6 Parameter estimation in closed loop ..................... 169
9.3 Identification of nonlinear processes . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
9.3.1 Parameter estimation for nonlinear static processes ........ 171
9.3.2 Parameter Estimation with Classical Nonlinear Models .... 172
9.3.3 Artificial Neural Networks for Identification. . . . . . . . . . . . . . 173
9.3.4 Identification with Grid-based Look-up Tables for Static
Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
X Contents

9.3.5 Parameter Estimation for Non-continuously Differentiable

Nonlinear Processes (Friction and Backlash) ............. 190
9.4 Symptom generation with identification models ................. 193
9.5 Problems ................................................. 195

10 Fault detection with parity equations ............................. 197

10.1 Parity equations with transfer functions ........................ 197
10.2 Parity equations with state-space models . . . . . . . . . . . . . . . . . . . . . . . 203
10.2.1 Continuous-time parity approach . . . . . . . . . . . . . . . . . . . . . . . 203
10.2.2 Discrete-time parity approach .......................... 207
10.3 Properties of residuals ....................................... 210
10.3.1 Generation of enhanced residuals ....................... 210
10.3.2 Generation of structured residuals ...................... 213
10.3.3 Sensitivity of parity equations. . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.4 Parity equations for nonlinear processes . . . . . . . . . . . . . . . . . . . . . . . . 220
10.4.1 Parity equations for special nonlinear processes ........... 220
10.4.2 Parity equation for nonlinear, local linear models .......... 221
10.5 Parameter estimation with parity equations . . . . . . . . . . . . . . . . . . . . . 226
10.6 Problems ................................................. 228

11 Fault detection with state observers and state estimation . . . . . . . . . . . . 231

11.1 State observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
11.1.1 Additive faults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
11.1.2 Multiplicative faults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
11.1.3 Fault isolation with state observers . . . . . . . . . . . . . . . . . . . . . . 234
11.2 State estimation (Kalman filter) ............................... 237
11.3 Output observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
11.4 Comparison of the parity- and observer-based approaches . . . . . . . . . 246
11.4.1 Comparison of residual equations. . . . . . . . . . . . . . . . . . . . . . . 246
11.4.2 Comparison by simulations ............................ 247
11.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252

12 Fault detection of control loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

12.1 Effects of faults on the closed loop performance ................. 253
12.2 Signal-based methods for closed-loop supervision ............... 254
12.3 Methods for the detection of oscillations in closed loops . . . . . . . . . . 260
12.4 Model-based methods for closed-loop supervision ............... 261
12.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

13 Fault detection with Principal Component Analysis (PCA) . . . . . . . . . . 267

13.1 Principal components ....................................... 267
13.2 Fault detection with PCA .................................... 273
 Contents XI

14 Comparison and combination of fault-detection methods . . . . . . . . . . . 279

14.1 Assumptions of model-based fault detection . . . . . . . . . . . . . . . . . . . . 279
14.2 Suitability of model-based fault-detection methods .............. 282
14.3 Combination of different fault-detection methods ................ 283

Part Ill Fault-Diagnosis Methods

15 Diagnosis procedures and problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

15.1 Introduction to fault diagnosis ................................ 287
15.2 Problems of fault diagnosis .................................. 289
15.2.1 Diagnosis knowledge representation .................... 290
15.2.2 Prior knowledge ..................................... 290
15.2.3 Typical statistical symptom distributions ................. 291
15.2.4 Data size ........................................... 291
15.2.5 Symptom representation .............................. 292
15.3 Problems ................................................. 292

16 Fault diagnosis with classification methods . . . . . . . . . . . . . . . . . . . . . . . . 295

16.1 Simple pattern classification methods .......................... 295
16.2 Bayes Classification ........................................ 296
16.3 Geometric classifiers ........................................ 299
16.4 Polynomial classification .................................... 300
16.5 Decision trees ............................................. 302
16.6 Neural Networks for fault diagnosis ........................... 305
16.6.1 Multi-layer perceptron networks ........................ 306
16.6.2 Radial-basis function networks ......................... 307
16.6.3 Clustering and self-organizing networks ................. 309
16.7 Problems ................................................. 310

17 Fault diagnosis with inference methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 311

17.1 Fault trees ................................................. 312
17.2 Approximate reasoning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
17 .2.1 Forward chaining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
17 .2.2 Backward chaining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
17 .2.3 Summary and comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
17.3 Hybrid neuro-fuzzy systems .................................. 324
17.3.1 Structures .......................................... 324
17 .3.2 Identification of membership functions . . . . . . . . . . . . . . . . . . 329
17 .3.3 Identification of rules with predefined membership functions 330
17.3.4 Optimization methods ................................ 332
17.3.5 Self-learning classification tree (SELECT) ............... 333
17.4 Problems ................................................. 343
XII Contents

Part IV Fault-Tolerant Systems

18 Fault-tolerant design ........................................... 347

18.1 Basic redundant structures ................................... 347
18.2 Degradation steps .......................................... 351
18.3 Problems ................................................. 353

19 Fault-tolerant components and control . . . . . . . . . . . . . . . . . . . . . . . . . . . 355

19.1 Fault-tolerant sensors ....................................... 355
19.1.1 Hardware sensor redundancy .......................... 355
19.1.2 Analytical sensor redundancy .......................... 356
19.2 Fault-tolerant actuators ...................................... 358
19.3 Fault-tolerant communication ................................ 359
19.4 Fault-tolerant control systems ................................ 360
19.5 Automatic fault-management system .......................... 362
19.6 Problems ................................................. 365

Part V Application Examples

20 Fault detection and diagnosis of DC motor drives . . . . . . . . . . . . . . . . . . 369

20.1 DC motor ................................................. 369
20.1.1 DC motor test bench ................................. 369
20.1.2 Parity equations ..................................... 371
20.1.3 Parameter estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
20.1.4 Experimental results for fault detection . . . . . . . . . . . . . . . . . . 373
20.1.5 Experimental results for fault diagnosis with SELECT ..... 376
20.2 Electrical throttle valve actuator .............................. 382
20.2.1 Actuator setup ....................................... 382
20.2.2 Parameter estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
20.2.3 Parity equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
20.2.4 Diagnostic equipment for quality control . . . . . . . . . . . . . . . . 388

21 Fault detection and diagnosis of a centrifugal pump-pipe-system .... 391

21.1 The pump-pipe-tank system .................................. 391
21.2 Mathematical models of the centrifugal pump ................... 392
21.3 Parity equations and parameter estimation ...................... 393

22 Fault detection and diagnosis of an automotive suspension and the

tire pressures .................................................. 401
22.1 Mathematical model of a suspension and the test rig ............. 401
22.2 Parameter estimation (test rig) ................................ 402
22.3 Parity equations (test rig) .................................... 404
22.4 Experimental results with a driving vehicle . . . . . . . . . . . . . . . . . . . . . 405
 Contents XIII

22.5 Shock absorber fault detection during driving . . . . . . . . . . . . . . . . . . . 405

22.6 Tire pressure supervision with spectral analysis . . . . . . . . . . . . . . . . . 409

23 Appendix ..................................................... 413

23.1 Terminology in fault detection and diagnosis .................... 413
23.2 State variable filtering of noisy signals to obtain signal derivations . 415
23.3 Fuzzy logic- a short introduction ............................. 416
23.3.1 Basics .............................................. 416
23.3.2 Simplification for Fault Diagnosis . . . . . . . . . . . . . . . . . . . . . . 420
23.4 Estimation of physical parameters for dynamic processes ......... 420
23.4.1 Introduction ......................................... 420
23.4.2 On the model structure for processes with lumped parameters 422
23.4.3 Calculation of the physical process coefficients . . . . . . . . . . . 424
23.4.4 Example: Second order electrical circuit ................. 425
23.5 From Parallel to Hierarchical Rule Structures ................... 426
23.5.1 Parallel Rule Bases ................................... 427
23.5.2 Hierarchical Rule Bases ............................... 428

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435

Index ............................................................. 469

List of symbols

Only frequently used symbols and abbreviations are given.

Letter symbols

a parameters of differential of difference equations

b parameters of differential or difference equations
c spring constant, constant, concentration, stiffness
d damping coefficient
e equation error, control deviation e = w - y, number e
2.71828 ...
f fault, frequency (f = 1/Tp, Tp period), function f( .. .)
g gravitational acceleration, function g( ... ), impulse response
integer, gear ratio, index, vC1 (imaginary unit)
j integer, index
k discrete number, discrete time k = tj To = 0, 1, 2, ... (T0 : sam-
pling time)
l index, length
m mass, order number
n rotational speed, order number, disturbance signal
p pressure, index, controller parameter, probability density func-
tion, process parameter
q controller parameter, failure density
r index, radius, reference variable, residual
s Laplace variables= 8 + iw, symptom
t continuous time, principal component coordinate
u input signal change !::!.. U
V speed, specific volume, disturbance signal
w reference value, setpoint
X space coordinate, state variable
y output signal change !::J..Y, space coordinate, control variable
change I::!..Y, signal
XVI Contents

z space coordinate, disturbance variable change !::J..Z, z-transform

variable z = exp T0 s
x estimated or observed variable
.X estimation error
.X average, steady-state value
xo amplitude
Xoo value in steady state

A area
B magnetic flux density
c capacitance
D damping ratio, diameter
E module of elasticity, energy, potential, bulk modulus
F filter transfer function, force
G weight, transfer function
H magnetic field strength, height
I electrical current, mechanical momentum, torsion, second mo-
ment of area
J moment of inertia
K constant, gain
L inductance
N discrete number, windings number
p power, probability
Q generalized force, heat, unreliability function
R electrical resistance, covariance or correlation function, reliabil-
ity function, risk number
s spectral density, sum, performance criterion
T absolute temperature, torque, time constant
To sampling time
u input variable, manipulated variable (control input), voltage
V volume
X space coordinate
y output variable, space coordinate, control variable
z space coordinate, disturbance variable
a vector
A matrix
AT transposed matrix
I identity matrix
() parameter vector
p covariance matrix
t data vector

a coefficient, angle
f3 coefficient, angle
y specific weight, correcting factor
 Contents XVII

decay factor, impulse function

correlation function
efficiency
temperature
thermal conductivity, forgetting factor, failure rate
friction coefficient, permeability, membership function
kinematic viscosity, index
number n = 3.14159 ...
p density
u standard deviation, u 2 variance
T time
angle
angular frequency, w = 2nj Tp; Tp period
change, deviation
parameter
product
sum
magnetic flux
magnetic flux linkage

Mathematical abbreviations

exp (x) =ex

E{} expectation of a statistical variable
dim dimension
adj adjoint
det determinant
Re real part
Im imaginary part
y dY j dt (first derivative)
var [] variance
cov [] covar1ance
:F Fourier transform

Abbreviations

ACF Auto Correlation Function

ARMA Auto Regressive Moving Average process
CCF Cross Correlation Function
DFT Discrete Fourier Transform
ETA Event Tree Analysis
FDD Fault Detection and Diagnosis
FDI Fault Detection and Isolation
FMEA Failure Mode and Effects Analysis
XVIII Contents

FFT Fast Fourier Transform

FTA Fault Tree Analysis
HA Hazard Analysis
LS Least Squares
MLP Multilayer Perceptron
MTBF Mean Time Between Failures
MTTF Mean Time To Failure = 1/'A
MTTR Mean Time To Repair
NN Neural Net
PCA Principal Component Analysis
PRBS Pseudo Random Binary Signal
RBF Radial Basis Function
RLS Recursive Least Squares
rms ( ...) root of mean squared of( ...)