Diagnosis and Fault-Tolerant Control

Transcription

1 Mogens Blanke Michel Kinnaert Jan Lunze Marcel Staroswiecki Diagnosis and Fault-Tolerant Control With contributions by Jochen Schroder With 228 Figures Springer

2 1. Introduction to diagnosis and fault-tolerant control Technological processes subject to faults Faults and fault tolerance Faults Requirements and properties of systems subject to faults Elements of fault-tolerant control Fault-tolerant control architecture Main ideas of fault diagnosis Main ideas of controller re-design A practical view on fault-tolerant control Survey of the book 23 Bibliographical notes Examples Two-tank system ' Ship steering and track control Models of dynamical systems Fundamental notions Modelling the system architecture System behaviour - basic modelling features Continuous-variable systems System structure Discrete-event systems Hybrid systems Links between the different models 57 Bibliographical notes Analysis based on components and architecture Introduction Generic component models Services ' Introduction of the generic component model Simple components 66

3 4.2.4 Complex components Building systems from components Faults in components and their consequences Fault propagation analysis Graph representation of component architecture Fault propagation in closed loops Cutting the closed fault propagation loop Assessment of the severity of the fault effects Decision about fault handling Fault tolerance analysis Relation between services and objectives Management of service versions Management of operation modes 96 Bibliographical notes Structural analysis Introduction Structural model Structure as a bi-partite graph Subsystems Structural properties Known and unknown variables Matching on a bi-partite graph Ill Definitions c Oriented graph associated with a matching Alternated chains and reachability Causal interpretation Matching algorithm System canonical decomposition Definitions Canonical subsystems Interpretation of the canonical decomposition Observability Observability and computability Structural observability conditions Observability of linear systems Graph-based interpretation and formal computation Monitorability Analytical redundancy-based fault detection and isolation Structurally monitorable subsystems Design of analytic redundancy relations Parity space and observer-based approaches Design of robust and structured residuals Fault propagation and alarm filtering 161

4 5.7 Controllability Structural analysis of fault tolerance Faults and the system structure Knowledge about faults Fault tolerance with respect to non-structural faults Fault tolerance with respect to structural faults Conclusions 170 Bibliographical notes Fault diagnosis of continuous-variable systems Introduction Deterministic model - parity space approach Fault detection Solution by the parity space approach Fault isolation Fault estimation Deterministic model - optimisation-based approach Problem statement Solution using the standard setup formulation Residual generation Stochastic model - change detection algorithms Introduction Sequential change detection: the scalar case Sequential change detection: the vector case Stochastic model - Kalman filter approach Model Fault detection Fault estimation Fault isolation 252 Bibliographical notes Fault-tolerant control of continuous-variable systems Fault-tolerant control architecture The fault-tolerant control problem Standard control problem Impacts of faults on the control problem Passive versus active fault-tolerant control Available knowledge Active fault-tolerant control strategies Supervision An optimal control approach to fault-tolerant control with actuator faults Control problem Control of the nominal plant Fault tolerance with respect to actuator faults 272

5 7.3.4 Fault accommodation Control reconfiguration Example Extension to a more general problem setting Model-matching approach to fault-tolerant control Reconfiguration problem Fault-tolerant control based on model-matching Model-matching control for sensor faults Model-matching control for actuator faults Control reconfiguration for actuator or sensor faults The idea of virtual sensors and virtual actuators Reconfiguration problem Virtual sensor Virtual actuator Duality between virtual sensors and virtual actuators Controller re-design in the general fault case System description Youla-Kucera parameterisation in coprime factorisation form Parametrisation in the state-space form Simultaneous design of the controller and the residual generator 309 Bibliographical notes Diagnosis and reconfigurable control of discrete-event systems Motivation Models of discrete-event systems Deterministic and non-deterministic systems Non-deterministic automata and Petri nets Stochastic automata State observation of stochastic automata Observation problem Consistent input-output pairs Solution to the state observation problem Recursive form of the solution Discussion of the results Observation algorithm State observation of non-deterministic automata Observability of stochastic automata Distinguishing inputs Diagnosis of stochastic automata Principle of consistency-based diagnosis Model of the faulty automaton Consistency-based diagnosis of stochastic automata

6 xiii Diagnostic algorithm Diagnosability of stochastic automata Control reconfiguration for stochastic automata Sensor and actuator fault isolation Automatic substitution of faulty sensors Automatic reconfiguration of diagnosis 382 Bibliographical notes Diagnosis and reconfiguration of quantised systems Introduction to quantised systems Supervision of hybrid systems The quantised system approach to supervisory control Quantised systems Continuous-variable system Quantisation of the signal spaces Example: two-tank system Behaviour of quantised systems Stochastic properties of quantised systems A behavioural view on supervision problems Discrete-event models of quantised systems The modelling problem Description of autonomous quantised systems by stochastic automata Extensions to systems with input and output Abstractions of faulty systems State observation of quantised systems Observation method Discussion of the result Observation algorithm Diagnosis of quantised systems Diagnostic method Discussion of the result Diagnostic algorithm Automatic reconfiguration of diagnosis in case of sensor or actuator faults Extensions and application examples Fault-tolerant control of quantised systems Reconfiguration problem Graph-theoretic formulation of the control problem A reconfiguration method 436 Bibliographical notes 437

7 xiv Contents 10. Application examples Fault-tolerant control of a three-tank system The control problem Generic component-based analysis of the three-tank system Solution of the reconfiguration task Diagnosis and fault-tolerant control of a chemical process System description and control aims Plant model for diagnosis Experimental diagnostic results Control reconfiguration in case of actuator faults Diagnosis and control of a ship propulsion system Structure of the ship propulsion system Models of the propulsion system Fault scenarios and requirements on the diagnosis Structural analysis of the propulsion system Fault diagnosis by using the parity space approach and state observation Diagnosis of the pitch control loop by means of the quantised systems approach Fault-tolerant propulsion Supervision of a steam generator Description of the process Modeling of the steam generator Design of the diagnostic system Structural analysis Fault signatures Experimental results Fault scenarios Conclusions Summary: A guideline for the design of fault-tolerant control Architecture Design procedure 520 Bibliographical notes 525 References 527 Appendices Appendix 1: Some prerequisites on vectors and matrices 537 Appendix 2: Notions of probability theory 540 Appendix 3: "H 2 and TiooController design 552

8 xv Appendix 4: Nomenclature 558 Appendix 5: Terminology 559 Appendix 6: Dictionary 562 Subject index 566