A review of process fault detection and diagnosis: Part III: Process history based methods
Introduction
In contrast to the model-based approaches where a priori knowledge (either quantitative or qualitative) about the process is needed, in process history based methods, only the availability of large amount of historical process data is needed. There are different ways in which this data can be transformed and presented as a priori knowledge to a diagnostic system. This is known as feature extraction. This extraction process can be either qualitative or quantitative in nature. Two of the major methods that extract qualitative history information are the expert systems and trend modelling methods. Methods that extract quantitative information can be broadly classified as non-statistical or statistical methods. Neural networks are an important class of non-statistical classifiers. Principal component analysis (PCA)/partial least squares (PLS) and statistical pattern classifiers form a major component of statistical feature extraction methods. The different ways in which knowledge can be extracted from process history are schematically presented in Fig. 1. We review these approaches in this part of the review paper.
We also compare and evaluate the various methodologies reviewed in this three part series in terms of the set of desirable characteristics we proposed in Part I. This comparative study identifies the relative strengths and weaknesses of the different approaches. It also reveals that no single method has all the desirable features we stipulated for a diagnostic system. It is our view that some of these methods can complement one another resulting in better diagnostic systems. Integrating these complementary features is one way to develop hybrid methods that could overcome the limitations of individual solution strategies. We review some recent trends in this direction. The important role of fault diagnosis in the broader context of process operations is also outlined. Finally, we also discuss the technical challenges in research and development that need to be addressed for the successful design and implementation of practical intelligent supervisory control systems for the process industries.
Section snippets
Qualitative feature extraction
As mentioned earlier two of the important methods that employ qualitative feature extraction are the expert systems and trend modelling approaches. In this section we review these two approaches.
Quantitative feature extraction
In this section we will discuss the methods that are based on quantitative feature extraction. The quantitative approaches essentially formulate the diagnostic problem-solving as a pattern recognition problem. The goal of pattern recognition is the classification of data points to, in general, pre-determined classes. Statistical methods use knowledge of a priori class distributions to perform classification. An example is a Bayes classifier which uses the density functions of the respective
A comparison of various approaches
So far in this three part series, we have reviewed the three conceptually different frameworks for process fault diagnosis. In this section, we provide a comparative evaluation of these different frameworks against a common set of desirable characteristics for a diagnostic system that we proposed in part I. The evaluations are summarized in Table 2.
Quantitative model-based methods, such as parity space and observer-based approaches, have several desirable characteristics (Gertler, 1991). If one
Role of fault diagnosis in design and other process operations
Much interest has been shown in the concept of total process control (Garcia, Ramaker, & Pollard, 1991) with the realization that, due to the limitations of accurate model development, automation of decision-making requires the addition of an exception handling decision layer. These include such tasks as identifying the occurrence of events outside of normal operation, diagnosing the root cause and finally synthesizing and implementing a corrective action.
Fault diagnosis shares with other
Conclusions and future directions
The basic aim of this paper is to organize, classify, review and compare various approaches to fault diagnosis from different perspectives. Towards that goal, we classify the different methods into three categories: (i) quantitative model-based methods; (ii) qualitative model-based methods; and (iii) process history based methods. We also present a framework that shows how these different approaches relate to and differ from each other regarding the transformation of information from the
References (118)
- et al.
A model-object based supervisory expert system for fault tolerant chemical reactor control
Computers and Chemical Engineering
(1990) - et al.
An integrated neural network/expert system approach for fault diagnosis
Computers and Chemical Engineering
(1993) - et al.
Application of wavelets and neural networks to diagnostic system development. I. Feature extraction
Computers and Chemical Engineering
(1999) - et al.
Hierarchical decision process for fault administration
Computers and Chemical Engineering
(1992) - et al.
Representation of process trends part I. A formal representation framework
Computers and Chemical Engineering
(1990) - et al.
A qualitative physics based on confluences
Artificial Intelligence
(1984) - et al.
Framework for enhancing fault diagnosis capabilities of artificial neural networks
Computers and Chemical Engineering
(1994) Fault diagnosis in dynamic-systems using analytical and knowledge-based redundancy—a survey and some new results
Automatica
(1990)- et al.
Automatic generation of qualitative description of process trends for fault detection and diagnosis
Engineering Applications of Artificial Intelligence
(1991) - et al.
Representing bounded fault classes using neural networks with ellipsoidal functions
Computers and Chemical Engineering
(1993)