k10.1108/kKybernetes0368-492XEmerald Group Publishing Limited10.1108/03684920510595427e-reviewReviewcat-ENGGEngineeringcat-EEEElectrical & electronic engineeringcat-CSEComputer & software engineeringcat-SYSCSystems & controlcat-IKMInformation & knowledge managementcat-ISYSInformation systemscat-SMCSystems modelling & cyberneticsBook ReviewFault Diagnosis: Models, Artificial Intelligence, ApplicationsAlex M. Andrew010620053457427432005J. Korbicz, J.M. Kościelny, Z. Kowalczuk and W. Cholewa (Eds). Fault Diagnosis: Models, Artificial Intelligence, Applications. Berlin: Springer 2004. xxix +920 pp., ISBN: 3‐540‐40767‐7, hardback £154.00 Engineering Online Library series© Emerald Group Publishing Limited2005Modelling, Artificial intelligencepeer-reviewednoacademic-contentyesrightslinkincluded

This is an extremely full treatment of its topic, with a total of 23 chapters by different groups of authors, all of them including one or more of the editors. It is divided into three Parts, of which the first, with seven chapters, has the heading Methodology, the second, with ten chapters, is on Artificial Intelligence, and the third, with six chapters, is on Applications.

In a Foreword, Professor P.M. Frank (Duisburg University) confirms the current importance of fault diagnosis and observes that modern methods are model‐based. However, analytic or parametric modelling presents difficulties in setting suitable levels of deviation to indicate faults. The system must obviously be sufficiently sensitive to respond when needed, but will be useless if it gives numerous false alarms. Solutions are to be found among techniques coming under the heading of Artificial Intelligence and these are thoroughly treated in the large Part II of the book. At the same time, approaches coming from established methods in control theory are also included and there is mention of Kalman filtering and the Luenberger observer, and alternative methods due to Wiener and Hammerstein for modelling non‐linear systems. The book is intended as a reference for industrial control engineers and as a textbook for undergraduate and postgraduate students.

As Professor Frank goes on to say: “The book on hand is one of the few comprehensive works on the market covering the fundamentals of model‐based fault diagnosis, which has become an emerging discipline of modern control engineering, in a very wide context including both analytical and non‐analytical (fuzzy and neural) models as well as approaches based on artificial and computational intelligence. It is a multiauthored book, where the editors are well acknowledged experts in the field, not only in the theoretical domain but also with respect to industrial applications. The latter finds expression in the fact that a substantial part of the text is dedicated to practical applications”.

The importance attached to the topic is reflected in the fact that an earlier comprehensive work has appeared in the same series (Diagnosis and Fault‐Tolerant Control, by Blanke, Kinnaert, Lunze and Staroswiecki, 2003). The emphasis in the new work is somewhat different, especially in its attention to methods based on Artificial Intelligence. The earlier work is produced by workers directly sponsored by the European Science Foundation and taking part in its COSY and DAMADICS projects. The new work also contains acknowledgement, at several points, of support from the DAMADICS project.

The range of topics coming under the general heading of Artificial Intelligence is quite wide and includes evolutionary algorithms, artificial neural nets, fuzzy logic, genetic algorithms, pattern recognition and expert systems. Since most of these depend on machine learning, it seems likely that methods must depend on prior visualisation of possible faults, but perhaps such a limitation is inevitable in any application to a complex system.

Part 3 of the book is on Applications, and a number of specific cases are described, but there is also general discussion of how diagnostic and remedial measures should operate. It will often be possible to isolate a faulty part, temporarily, without closing everything down and a built‐in strategy has to be carefully devised and is again model‐based. Allowance must be made for multiple simultaneous faults. Specific application areas mentioned include detecting and locating leaks in transmission pipelines, and faults in processes in a sugar factory, as well as others within a steam power generator and a pneumatic actuator.

This is an impressively comprehensive and well‐prepared work. Each chapter begins with an introduction and ends with a summary, and is followed by an extensive list of references. There is no subject index, but in the initial contents list all section and subsection (and sometimes subsubsection) headings are shown and navigation for reference should not be difficult.