No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Application of Fuzzy Logic to Diagnostics
Abstract
The basics of fuzzy logic as well as fuzzy modelling and control are described, for example, in the monographies by Czogała and Łęski (2000), Yager and Filev (1994), Drinkov et al. (1996), Rutkowska (2002), and Piegat (2001). An interesting overview of fuzzy logic application to fault detection and isolation can be found in Frank and Marcu (2000). This chapter presents the application of fuzzy logic to fault detection and isolation.
... In literature, the diagnosis problem is considered in different formulations, depending primarily on the models used to describe a system: deterministic 1,2,3 , stochastic, fuzzy 4 . The choice of a formulation is determined, as a rule, by the application and the problem to be solved by the dynamic system, as well as a priori information on the properties of the system and its possible faults available to developers of diagnostic tools. ...
Within a fuzzy approach, a method for dynamic systems diagnosis is proposed, based on the use of bank of interacting state observers. It is also analyzed whether interval observers can be applied for the diagnosis as well. Synthesis procedures for which both Hurwitz and Metzler conditions are satisfied are considered for them.
The decentralisation of the diagnostic tasks could be achieved when the automatic and technological equipment devices will be provided with class of smart autonomous diagnostic units. In this case the functions of the central diagnostic system may be simplified. This class may be represented by smart positioner with actuator fault diagnosis. In the paper an idea of smart positioner with actuator fuzzy based fault diagnosis is presented. The scheme of the residuum vector generation based on the fuzzy models is given, as well as with the scheme of diagnostic's decision making procedure for the positioner .
Fault diagnosis is the problem of finding out the root cause (the fault) of process malfunctions. In this work we present a new model based approach procedure for fault diagnosis in conventional chemical processes. The process model is a Signed Directed Graph (SDG). The SDG is used by a Qualitative Simulator. This stage allows us to know, for each potential fault, the possible process behaviour. This information is compiled into a set of IF-THEN rules. An Expert System evaluates them using information about the actual process state. Fuzzy Logic is used in this evaluation. By the way, the fault whose rule has the highest value of certainty should be the first one considered by the operator. Finally, the additional information provided by the Qualitative Simulation enables the Expert System to explain the diagnostic. Besides, this additional information can be used to improve the diagnostic.
This paper presents a Fault Detection and Isolation (FDI) method based on a fuzzy evaluation of residuals. The advantages of this non-Boolean evaluation are described. The ability of this approach to integrate a persistence index of the residual deviations and the sensitivity of the residuals with regard to faults is particularly studied. Some results issued from the implementation of the proposed approach on an urban water supply network illustrate its pertinence.
Fuzzy systems are represented as series expansions of fuzzy basis
functions which are algebraic superpositions of fuzzy membership
functions. Using the Stone-Weierstrass theorem, it is proved that linear
combinations of the fuzzy basis functions are capable of uniformly
approximating any real continuous function on a compact set to arbitrary
accuracy. Based on the fuzzy basis function representations, an
orthogonal least-squares (OLS) learning algorithm is developed for
designing fuzzy systems based on given input-output pairs; then, the OLS
algorithm is used to select significant fuzzy basis functions which are
used to construct the final fuzzy system. The fuzzy basis function
expansion is used to approximate a controller for the nonlinear ball and
beam system, and the simulation results show that the control
performance is improved by incorporating some common-sense fuzzy control
rules
A general method is developed to generate fuzzy rules from
numerical data. The method consists of five steps: divide the input and
output spaces of the given numerical data into fuzzy regions; generate
fuzzy rules from the given data; assign a degree of each of the
generated rules for the purpose of resolving conflicts among the
generated rules; create a combined fuzzy rule base based on both the
generated rules and linguistic rules of human experts; and determine a
mapping from input space to output space based on the combined fuzzy
rule base using a defuzzifying procedure. The mapping is proved to be
capable of approximating any real continuous function on a compact set
to arbitrary accuracy. Applications to truck backer-upper control and
time series prediction problems are presented
The architecture and learning procedure underlying ANFIS
(adaptive-network-based fuzzy inference system) is presented, which is a
fuzzy inference system implemented in the framework of adaptive
networks. By using a hybrid learning procedure, the proposed ANFIS can
construct an input-output mapping based on both human knowledge (in the
form of fuzzy if-then rules) and stipulated input-output data pairs. In
the simulation, the ANFIS architecture is employed to model nonlinear
functions, identify nonlinear components on-line in a control system,
and predict a chaotic time series, all yielding remarkable results.
Comparisons with artificial neural networks and earlier work on fuzzy
modeling are listed and discussed. Other extensions of the proposed
ANFIS and promising applications to automatic control and signal
processing are also suggested
A fault detection network is developed through principal component decomposition. It provides a model for the process under nominal operating conditions. The fault detection network provides several fault detection signals. The presence of a fault is detected when a fault detection signal exceeds its threshold. To locate the fault, a neuro-fuzzy network is developed from the normal and faulty process operating data. The neuro-fuzzy network is obtained by adding a fuzzification layer to a conventional feedforward network. The fuzzification layer converts the increments in on-line measurements into a number of fuzzy sets, such as "increase", "steady", and "decrease". In this way, training data is condensed and network robustness is enhanced. The neuro-fuzzy approach also provides a way to embed certain process knowledge into neural networks. The techniques are applied to a continuous stirred tank reactor.
For the automatic monitoring of technological systems, one has to develop diagnostic systems which are able to recognize the state the system is governed by, with quickness and confidence. Furthermore, since some states may be unknown, such diagnostic systems must be able to self adapt to new situations, i.e. to learn. Pattern Recognition methods may be used for that purpose. But nowadays, predictive diagnosis is required in order to prevent an undesirable evolution of the system. This paper presents an adaptive predictive diagnostic system which aims at reaching these requirements. A fuzzy approach is used for the learning, decision and prediction abilities of the system.
Oriented graphs are well suited to represent the causality of physical phenomena; therefore they can be used to better understand the propagation of faults through an industrial plant. Using this method, a framework for qualitative and/or quantitative reasoning in fault diagnosis is provided: the data enrichment of the causal structure leads to evolve from alarm treatment systems to refined model-based approaches.
This paper describes a monitoring and diagnosis process for the detection of evolution of a system whose characteristics vary with time. This is an adaptive fuzzy pattern recognition algorithm that achieves progressive and on-line learning of the system states. A neural net based architecture associated with a real-time algorithm for the detection of abrupt changes provides a diagnosis of evolution that may also be predictive. The algorithm was applied to the monitoring of a car driver s behavior. Special attention was given to the detection of hypovigilance. Promising results on real data are reported.
This survey paper discusses the application of fuzzy logic to fault detection, isolation and supervision of technical processes. Emphasis is placed upon the fuzzy model- based residual generation using the concept of knowledge observer, and the rule-based residual evaluation using fuzzy relations with linguistic variables and fuzzy terms. The principles of threshold fuzzification and fuzzy rule-based threshold adaption are outlined, and the usefulness of fuzzy logic for man-machine interaction is pointed out.
This paper describes three ways of building a decision system for quantitative model-based fault diagnosis. Some clues are given to apply the method to qualitative model-based FDI. The aim is the management of uncertain and redundant information provided by a residual generator. Although any residual generation method may be considered, the input-output parity equation approach has been used. The key point is the fault sensitivity estimations of the residuals. which lead to define the decision rules. In the case of sensor and actuator faults, sensitivity estimates are easily obtainable from the parity equations. Three ways to solve the decision problem are described: fuzzy logic-based, direct weighting of symptoms and directional properties. The proposed methods have been tested, first on simulation and then on two laboratory control equipments: 3-tanks system and d.c motor system.
In this contribution a new approach for fault detection and diagnosis (FDD) for nonlinear processes is presented. A nonlinear fuzzy model with transparent inner structure is used for the generation of relevant symptoms. The resulting symptom patterns are classified with a new self-learning classification structure based on fuzzy rules. The approach is successfully applied to a electro-pneumatic valve in a closed control loop.
Reliability of the diagnostic part of the fault detection system which is in charge of isolation of a fault, is as important as the robustness of the residual generator because the isolation logic unit provides the information about the kind of fault, failed components and the severity of the fault. This paper deals with the rule base design problem for the observer based fault detection systems and main emphasis is placed on the method of rule modification using the excess redundancy that is inherent in the observer based FDS and employing fuzzy logic inference. It is shown that the combined use of compensatory operator and non-compensatory one provides an useful means of enhancing the reliability of the diagnostic decision process even in the cases that some residuals are perturbed by external disturbances and some other type of uncertainties.
The decentralisation of the diagnostic tasks could be achieved when the automatic and technological equipment devices will be provided with class of smart autonomous diagnostic units. In this case the functions of the central diagnostic system may be simplified. This class may be represented by smart positioner with actuator fault diagnosis. In the paper an idea of smart positioner with actuator fuzzy based fault diagnosis is presented. The scheme of the residuum vector generation based on the fuzzy models is given, as well as with the scheme of diagnostic's decision making procedure for the positioner.
This paper describes a method to build a fuzzy model of a Fault Diagnosis System using fuzzy implications and reasoning. The identification mechanism is a neurofuzzy adaptive model. We use this method to construct a Sugeno fuzzy model of a plant Finally we use this model for building a decision system for model-based fault diagnosis, using parity equations fault diagnosis method. A simple simulated case has been used to prove its performance.
This paper presents a Fault Detection and Isolation (FDI) method based on a fuzzy evaluation of residuals. The advantages of this non-Boolean evaluation are described. The ability of this approach to integrate a persistence index of the residual deviations and the sensitivity of the residuals with regard to faults is particularly studied. Some results issued from the implementation of the proposed approach on an urban water supply network illustrate its pertinence.
Fuzzy and neural process modelling methods for purposes of fault detection have been discussed in the paper. Fuzzy neural networks and TSK models (Takagi-Sugeno-Kang’s models) have been used for modelling. Research results of modelling as well as fault detection algorithms have been presented. Research for the chosen evaporator of evaporation station at “Lublin” sugar factory has been performed. The features of methods mentioned above have been also discussed.
The paper presents diagnostics of power boiler system performed with use of the DIAG system. Fuzzy and neural local models of particular systems have been applied for fault detection. Obtained results of control valve and attemperator modelling have been presented. Fault isolation has been performed with use of the F-DTS method that bases on fuzzy appreciation of residuals and fuzzy reasoning. An example of fault isolation in a steam superheater and attemperator has been presented.
System identification is the task of constructing representative models of processes and has become an invaluable tool in many different areas of science and engineering. Due to the inherent complexity of many real world systems the application of traditional techniques is limited. In such instances more sophisticated (so called intelligent) modelling approaches are required. Neurofuzzy modelling is one such technique, which by integrating the attributes of fuzzy systems and neural networks is ideally suited to system identification. This attractive paradigm combines the well established learning techniques of a particular form of neural network i.e. generalised linear models with the transparent knowledge representation of fuzzy systems, thus producing models which possess the ability to learn from real world observations and whose behaviour can be described naturally as a series of linguistic humanly understandable rules. Unfortunately, the application of these systems is limited to low dimensional problems for which good quality expert knowledge and data are available. The work described in this thesis addresses this fundamental problem with neurofuzzy modelling, as a result algorithms which are less sensitive to the quality of the a priori knowledge and empirical data are developed. The true modelling capabilities of any strategy is heavily reliant on the model's structure, and hence an important (arguably the most important) task is structure identification. Also, due to the curse of dimensionality, in high dimensional problems the size of conventional neurofuzzy models gets prohibitively large. These issues are tackled by the development of automatic neurofuzzy model identification algorithms, which exploit the available expert knowledge and empirical data. To alleviate problems associated with the curse of dimensionality, aid model generalisation and enhance model transparency, parsimonious models are identified. This is achieved by the application of additive and multiplicative neurofuzzy models which exploit structural redundancies found in conventional systems. The developed construction algorithms successfully identify parsimonious models, but as a result of noisy and poorly distributed empirical data, these models can still generalise inadequately. This problem is addressed by the application of Bayesian inferencing techniques; a form of regularisation. Smooth model outputs are assumed and superfluous model parameters are controlled, sufficiently aiding model generalisation and transparency, and data interpolation and extrapolation. By exploiting the structural decomposition of the identified neurofuzzy models, an efficient local method of regularisation is developed. All the methods introduced in this thesis are illustrated on many different examples, including simulated time series, complex functional equations, and multi-dimensional dynamical systems. For many of these problems conventional neurofuzzy modelling is unsuitable, and the developed techniques have extended the range of problems to which neurofuzzy modelling can successfully be applied.
This paper describes a monitoring and diagnosis process for the detection of evolution of a system whose characteristics vary with time. This is an adaptive fuzzy pattern recognition algorithm that achieves progressive and on-line learning of the system states. A neural net based architecture associated with a real-time algorithm for the detection of abrupt changes provides a diagnosis of evolution that may also be predictive. The algorithm was applied to the monitoring of a car driver's behavior. Results on real data are reported.
In this final chapter we consider several fuzzy algorithms that effect partitions of feature space ℝp , enabling classification of unlabeled (future) observations, based on the decision functions which characterize the classifier. S25 describes the general problem in terms of a canonical classifier, and briefly discusses Bayesian statistical decision theory. In S26 estimation of the parameters of a mixed multivariate normal distribution via statistical (maximum likelihood) and fuzzy (c-means) methods is illustrated. Both methods generate very similar estimates of the optimal Bayesian classifier. S27 considers the utilization of the prototypical means generated by (A11.1) for characterization of a (single) nearest prototype classifier, and compares its empirical performance to the well-known k-nearest-neighbor family of deterministic classifiers. In S28, an implicit classifier design based on Ruspini’s algorithm is discussed and exemplified.
FKBC has been proven to be a powerful tool when applied to the control of processes which are not amenable to conventional, analytic design techniques. The design of most of the existing FKBC has relied mainly on the process operator’s or control engineer’s experience based heuristic knowledge. Hence, the controller’s performance is very much dependent on how good this expertise is. Thus, from the control engineering point of view, the major effort in fuzzy knowledge based control has been devoted to the development of particular FKBC for specific applications rather than to general analysis and design methodologies for coping with the dynamic behavior of control loops. The development of such methodologies is of primary interest for control theory and engineering. In particular, stability analysis is of extreme importance, and the lack of satisfactory formal techniques for studying the stability of process control systems involving FKBC has been considered a major drawback of FKBC.
The main idea of this book is to present novel connectionist archite ctures of neuro-fuzzy systems, especially those based on the logical approach to fuzzy inference. In addition, hybrid learning methods are proposed to train the networks. The neuro-fuzzy architectures plus hybrid learning are considered as intelligent systems within the framework of computational and artificial intelligence. The book also provides an overview of fuzzy sets and systems, neural networks, learning algorithms (including genetic algorithms and clustering methods), as well as expert systems and perception-based systems which incorporate computing with words.
Classical Sets and Fuzzy Sets. Basic Definitions and Terminology: Classical Sets. Fuzzy Sets. Operations on Fuzzy Sets. Classification of t-Norms and t-Conorms. De Morgan Triple and Other Properties of t- and s-Norms. Parameterized t-, s-Norms and Negations. Fuzzy Relations. Cylindrical Extension and Projection of Fuzzy Sets. Extension Principle. Linguistic Variable. Summary.- Approximate Reasoning: Interpretation of Fuzzy Conditional Statement. An Approach to Axiomatic Definition of Fuzzy Implication. Compositional Rule of Inference. Fuzzy Reasoning. Canonical Fuzzy If-Then Rule. Aggregation Operation. Approximate Reasoning Using a Fuzzy Rule Base. Approximate Reasoning with Singletons. Fuzzifiers and Defuzzifiers. Equivalence of Approximate Reasoning Results Using Different Interpretations of If-Then Rules. Numerical Results. Summary.- Artificial Neural Networks: Introduction. Artificial Neural Networks Topologies. Learning in Artificial Neural Networks. Back-Propagation Learning Rule. Modifications of the Classic Back-Propagation Method. Optimization Methods in Neural Networks Learning. Networks with Output Linearly Depending on Parameters. Global Optimization Methods. Summary.- Unsupervised Learning. Clustering Methods: Introduction. Self-Organizing Feature Map. Vector Quantization and Learning Vector Quantization. An Overview of Clustering Methods. Fuzzy Clustering Methods. A Possibilistic Approach to Clustering. A New Generalized Weighted Conditional Fuzzy c-Means. Fuzzy Learning Vector Quantization. Cluster Validity. Summary.- Fuzzy Systems: Introduction. The Mamdani Fuzzy Systems. The Tagaki-Sugeno-Kang Fuzzy Systems. Fuzzy Systems with Parametrized Consequents. Summary.- Neuro-Fuzzy Systems: Introduction. Artificial Neural Network Based Fuzzy Inference Systems. Classifier Based On Neuro-Fuzzy System. ANNBFIS Optimization Using Deterministic Annealing. Further Investigations of Neuro-Fuzzy Systems. Summary.- Applications of Artificial Neural Network Based Fuzzy Inference System: Introduction. Application to Chaotic Time Series Prediction. Application to ECG Signal Compression. Application to Ripley's Synthetic Two-Class Data Classification. Application to the Recognition of Diabetes in Pima Indians. Application to the Iris Problem. Application to Monk's Problems. Application to System Identification. Application to Control. Application to Channel Equalization. Summary.
Fault tolerance of automatic control systems is gaining increasing importance. This is due to the increasing complexity of modern control systems and the growing demands for quality, cost efficiency, availability, reliability and safety. The use of knowledge based systems and of various“intelligent technologies” demonstrated significant improvements over the classic techniques. In this chapter, we review the state of this development along with the enumeration of some successful applications.
A mathematical tool to build a fuzzy model of a system where fuzzy implications and reasoning are used is presented. The premise of an implication is the description of fuzzy subspace of inputs and its consequence is a linear input-output relation. The method of identification of a system using its input-output data is then shown. Two applications of the method to industrial processes are also discussed: a water cleaning process and a converter in a steel-making process.
In this contribution a new approach for fault detection and diagnosis (FDD) for nonlinear processes is presented. A nonlinear fuzzy model with transparent inner structure is used for the generation of relevant symptoms. The resulting symptom patterns are classified with a new self-learning classification structure based on fuzzy rules. The approach is successfully applied to an electro-pneumatic valve in a closed control loop.
System identification is the task of constructing representative models of processes and has become an invaluable tool in many different areas of science and engineering. Due to the inherent complexity of many real world systems the application of traditional techniques is limited. In such instances more sophisticated (so called intelligent) modelling approaches are required. Neurofuzzy modelling is one such technique, which by integrating the attributes of fuzzy systems and neural networks is ideally suited to system identification. This attractive paradigm combines the well established learning techniques of a particular form of neural network i.e. generalised linear models with the transparent knowledge representation of fuzzy systems, thus producing models which possess the ability to learn from real world observations and whose behaviour can be described naturally as a series of linguistic humanly understandable rules. Unfortunately, the application of these systems is limited to low dimensional problems for which good quality expert knowledge and data are available. The work described in this thesis addresses this fundamental problem with neurofuzzy modelling, as a result algorithms which are less sensitive to the quality of the a priori knowledge and empirical data are developed. The true modelling capabilities of any strategy is heavily reliant on the model's structure, and hence an important (arguably the most important) task is structure identification. Also, due to the curse of dimensionality, in high dimensional problems the size of conventional neurofuzzy models gets prohibitively large. These issues are tackled by the development of automatic neurofuzzy model identification algorithms, which exploit the available expert knowledge and empirical data. To alleviate problems associated with the curse of dimensionality, aid model generalisation and enhance model transparency, parsimonious models are identified. This is achieved by the application of additive and multiplicative neurofuzzy models which exploit structural redundancies found in conventional systems. The developed construction algorithms successfully identify parsimonious models, but as a result of noisy and poorly distributed empirical data, these models can still generalise inadequately. This problem is addressed by the application of Bayesian inferencing techniques; a form of regularisation. Smooth model outputs are assumed and superfluous model parameters are controlled, sufficiently aiding model generalisation and transparency, and data interpolation and extrapolation. By exploiting the structural decomposition of the identified neurofuzzy models, an efficient local method of regularisation is developed. All the methods introduced in this thesis are illustrated on many different examples, including simulated time series, complex functional equations, and multi-dimensional dynamical systems. For many of these problems conventional neurofuzzy modelling is unsuitable, and the developed techniques have extended the range of problems to which neurofuzzy modelling can successfully be applied.
Application of fuzzy networks for fault isolation - Example for power boiler system
- J M Kościelny
- M Syfert
- JM Kościelny
Processing fault-trees by approximate reasoning on composite fuzzy relations in solving the technical diagnostic problem
- M Ulieru
Methods of fault isolation in industrial processes
- D Sędziak
Diagnostics of industrial processes with the use of partial models and fuzzy logic
- M Syfert
An overview of fuzzy modelling for control
- Babuŝka R
- H B Verbrugen