Knowledge based systems are used in applications where an incorrect decision could put human life in jeopardy. A quick trawl through the World Wide Web is sufficient, these days, to locate such applications in design, analysis and testing; protection advice; operator decision support; signal monitoring; embedded systems and others. Depending on the type of system, these either give information which is not guaranteed to be correct (in many operator support applications) or which is imprecise (for example in fuzzy logic controllers).

Diagnosis was among the first subjects investigated when digital computers became available. It still remains an important research area, in which several new developments have taken place in the last decade. One of these new developments is the use of detailed domain models in knowledge-based systems for the purpose of diagnosis, often referred to as “model-based” diagnosis. Typically, such models embody knowledge of the normal or abnormal structure and behaviour of the modelled objects in a domain. Models of the structure and workings of technical devices, and causal models of disease processes in medicine are two examples. In this article, the most important notions of diagnosis and their formalisation are reviewed and brought in perspec tive. In addition, attention is focused on a number of general frameworks of diagnosis, which offer sufficient flexibility for expressing several types of diagnosis.

This paper reviews Machine Learning (ML), and extends and complements previous work (Kocabas, 1991; Kalkanis and Conroy, 1991). Although this paper focuses on inductive learning, it at least touches on a great many aspects of ML in general. In addition, incremental induction is also reviewed. Therefore, a general review of ML is presented, but specific detail which has been covered previously is omitted, although other relevant references are noted, and later material is commented upon.

Induced decision trees are an extensively-researched solution to classification tasks. For many practical tasks, the trees produced by tree-generation algorithms are not comprehensible to users due to their size and complexity. Although many tree induction algorithms have been shown to produce simpler, more comprehensible trees (or data structures derived from trees) with good classification accuracy, tree simplification has usually been of secondary concern relative to accuracy, and no attempt has been made to survey the literature from the perspective of simplification. We present a framework that organizes the approaches to tree simplification and summarize and critique the approaches within this framework. The purpose of this survey is to provide researchers and practitioners with a concise overview of tree-simplification approaches and insight into their relative capabilities. In our final discussion, we briefly describe some empirical findings and discuss the application of tree induction algorithms to case retrieval in case-based reasoning systems.

The main components of case-based reasoning are case retrieval and case reuse. While case retrieval mostly uses attribute comparisons, many other possibilities exist. The case similarity concepts described in the literature that are based on more elaborate structural properties are classified here into five groups: restricted geometric relationships; graphs; semantic nets; model-based similarities; hierarchically structured similarities. Some general topics conclude this survey on structure-based case retrieval methods and systems.

A major task of organic chemistry is the synthesis of compounds of interest (“targets”) from available starting compounds. We address here the generation by computer of the best synthesis routes, or sequences of reactions, to such compounds. The number of possible routes in the search space is enormous and so the reasoning is examined in some detail, leading to a protocol for the operation of our program, SYNGEN, which has been under development for some years. It requires a new representation of the data – molecular structures and their reactions – which is digital in format and quite unlike the usual graphical depictions of organic chemistry. This is presented as a problem in artificial intelligence, and every effort has been made to present it with a minimum background of organic chemistry required of the reader.

Belief revision is the process of rearranging a knowledge base to preserve global consistency whilst accommodating incoming information. Early approaches to belief revision used symbolic model theoretic methods, considering the problem as one of changi ng a logical theory. More recent approaches have adopted qualitative syntactic methods, taking them into the area of truth maintenance systems, and numerical mathematical methods, thus moving into the mainstream literature of uncertainty management. Multi -agent systems, in which information may come from a variety of human and artificial sources with different degrees of reliability, seem to be a natural domain for belief revision. The aim of this paper is to give a synoptic perspective of this composite subject from the clear air of the high theoretical peaks down to the muddy plain of practical algorithms.

An appropriate use of neural computing techniques is to apply them to problems such as condition monitoring, fault diagnosis, control and sensing, where conventional solutions can be hard to obtain. However, when neural computing techniques are used, it is important that they are employed so as to maximise their performance, and improve their reliability. Their performance is typically assessed in terms of their ability to generalise to a previously unseen test set, although unless the training set is very carefully chosen, 100% accuracy is rarely achieved. Improved performance can result when sets of neural nets are combined in ensembles and ensembles can be viewed as an example of the reliability through redundancy approach that is recommended for conventional software and hardware in safety-critical or safety-related applications. Although there has been recent interest in the use of neural net ensembles, such techniques have yet to be applied to the tasks of condition monitoring and fault diagnosis. In this paper, we focus on the benefits of techniques which promote diversity amongst the members of an ensemble, such that there is a minimum number of coincident failures. The concept of ensemble diversity is considered in some detail, and a hierarchy of four levels of diversity is presented. This hierarchy is then used in the description of the application of ensemble-based techniques to the case study of fault diagnosis of a diesel engine.

Research in knowledge-based systems has shown that the use of multiple knowledge representation formalisms and reasoning mechanisms for achieving a specific task in complex domains could result in efficient and effective problem solving. This has led t o the development of a number of general architectures and application-specific systems integrating multiple problem solvers. A problem solver is defined to be an association between a knowledge intensive (sub)task, an inference mechanism and a knowledge representation formalism on which the inference mechanism is working to achieve the (sub)task. A knowledge-based system making use of different problem solvers should address a number of critical aspects of integration of the solvers, like interaction, in vocation, reactiveness, learning and expandability. The aim of this paper is to distinguish and discuss essential integration aspects, and to review a number of proposed general hybrid architectures and application specific hybrid systems on the basis of these aspects. The review shows that none of these general architectures or application specific systems directly addresses all the identified integration aspects. In general, the limitations exhibited by these systems are due to the naive form of interac tion and invocation of the integrated solvers. We give a high level specification for a competent hybrid knowledge-based architecture that supports the identified integration aspects.

The development of a knowledge-based design support system is a lengthy and costly process because various computational techniques necessary for intelligent design support are not readily available in a knowledge-based environment. The systematisation of design knowledge needs combined efforts from designers and knowledge engineers. Existing knowledge-based system development tools offer limited support to intelligent design support which require sophisticated knowledge engineering techniques in terms of knowledge representation, inference, control, truth maintenance and learning. In this paper, a knowledge-based architecture for intelligent design support is described. The existing knowledge-based design system architectures are reviewed first. Five key issues in intelligent design support using knowledge engineering techniques, i.e. design knowledge representation, structure of design knowledge base, intelligent control of design process, consistency and context management of design knowledge, and modelling of design collaboration are then discussed. These discussions provide a basis for a description of a knowledge-based design support system architecture which has been implemented in a Lisp-based environment and tested in two different domains. Current application of this architecture in the development of a design support system in the domain of mechanical engineering design at the Cambridge Engineering Design Centre is presented and evaluated.

The discipline of case-based reasoning develops techniques to retrieve and reuse old solutions for new problems. To reuse solutions, several case adaptation systems have been built and many are under development. They deal with different tasks in different domains, but a methodology is still lacking. From an analysis of up-to-date case adaptation systems, this article moves towards a methodology by providing a common framework and some guidelines. As key issues case adaptation techniques, global strategies, and the tailoring of cases are discussed.

The design and assessment of safety critical systems often involves broad and distributed teams of designers, suppliers and analysts who represent diverse areas of expertise and motivations. Accurate and effective communication between these groups is therefore an issue of primary importance. The formalisation of specifications and arguments of safety can be of significant benefit in ensuring the consistency of evidence in such cases, when it must be presented across many domains. However, a formal description of a safety critical system may be unconvincing unless it is presented in a form which is (or forms which are) accessible to the broad range of users and assessors of safety cases. This raises issues of human communication which include the tailoring of information to particular communicative tasks; the efficacy of differing media for communication and the cognitive impact that such differing media have. This paper draws together work in fields of knowledge engineering, knowledge based systems and human communication in an effort to address, from a sound theoretical basis, these and other communication issues raised by the use of formal descriptions in safety critical systems. Further, this paper argues that a primary role for knowledge based systems techniques in safety critical systems is in supporting the communication of information.

In October 1995, Takeshi Furuhashi and his collegues at the Bio-Electronics Laboratory of Nagoya University, Japan, organized the first of a series of on-line workshops, held entirely on the World Wide Web. The advertised advantages of the on-line format were to allow fruitful exchanges while avoiding physical travel, and to guarantee wide visibility of the discussion. The first two workshops in the series were devoted to evolutionary computation; they can be accessed on the web at http://www.bioele.nuee.nagoya-u.ac.jp. The third workshop, named “First On-Line Workshop on Soft Computing” (WSC1), had a broader scope, including all the techniques that go under the heading of “soft computing”, like fuzzy logic, neuro computing, genetic computing, and so on. WSC1 took place from August 19 to 30 1996, and it is accessible on the web at http://www.bioele.nuee.nagoya-u.ac.jp/wsc1/. Because the declared goal of an on-line workshop is to prompt discussion, the rules for submission were looser than in most traditional workshops: papers were not subject to peer review, and it was possible to submit already published papers. All the submitted papers were made visible on the web one week before the workshop, and people could send comments and questions by email during the two workshop weeks; all the questions, comments, and authors' replies are also visible at the WSC1 web site.

The KEML (Knowledge Engineering Methods and Languages) workshop which took place on 22–24 January at the Open University in Milton Keynes (UK) was the seventh in a series of workshops on methods and languages for knowledge engineering. Although t he KEML acronym suggests a broad knowledge engineering connotation, in practice the main emphasis of these workshops is on the construction, formalisation, verification and use of knowledge models. The term “knowledge model” originate s from the work of Allen Newell (1982), who proposed a level of description — the knowledge level — which abstracts from implementation-related considerations to focus on the actions, goals and knowledge embodied by a problem solving agent.

Increasing complexity of design in automotive electrical systems has been paralleled by increased demands for analysis of the safety and reliability aspects of those designs. Such demands can place a great burden on the engineers charged with carrying out the analysis. This paper describes how the intended functions of a circuit design can be combined with a qualitative model of the electrical circuit that fulfils the functions, and used to analyse the safety of the design. FLAME, an automated failure mode and effects analysis system based on these techniques, is described in detail. FLAME has been developed over several years, and is capable of composing an FMEA report for many different electrical subsystems. The paper also addresses the issue of how the use of functional and structural reasoning can be extended to sneak circuit analysis and fault tree analysis.

Over the past years, two main approaches to computational intelligence have emerged: the symbolic and the non-symbolic approach. The perhaps most prominent methods of the symbolic approach are based on logic. Logical methods exhibit a series of desirable properties:

[bull ] Transparent representation of meaning

[bull ] Precise understanding of the meaning of statements (semantics).

[bull ] Sound reasoning methods.

[bull ] Explanation capabilities.

A special session on logical methods for computational intelligence was held at the 3rd Joint Conference on Information Sciences. The field of computational logic is so broad that it is impossible to review the main developments in an article. Therefore, in the following we will restrict attention to two areas that turned out to be the focus of the special session: automated reasoning, and reasoning with incomplete and changing information.

Artificial intelligence research in temporal reasoning focuses on designing automated solutions to complex problems in computation involving time. TIME-97, the 4th International Workshop on Temporal Representation and Reasoning, held in Daytona Beach, Florida — like the three workshops that preceded it — had the objective of creating an international forum for the exchange of information among the many researchers and knowledge engineers who are developing and applying techniques in temporal reasoning.

Intelligent multimedia systems (IMMS) assist users in choosing the information content, and to design the information presentation, that satisfies their needs and idiosyncrasies. Towards this target, IMMSs posses and utilize knowledge concerning the means and function (purpose) of a presentation. ‘Means’ denote communication techniques (information carriers and modalities), as well as information content and information characteristics. The ‘function of a pr esentation’ denotes the communicative purpose of the presentation, methods for achieving that purpose, as well as a user's goals and tasks.

Ensuring reliability and enhancing quality of Knowledge Based Systems (KBS) are critical factors for their successful deployment in real-world applications. This is a broad task involving both methodological and formal approaches for designing rigorous Validation, Verification and Testing (VVT) methods and tools. Some of these can be adapted from conventional software engineering, while others rely on specific aspects of KBS.

Few would disagree with the assertion that safe engineering starts from the early stages of system design and should be maintained throughout the lifecycle. Different engineering domains have developed, mostly informal, frameworks with which they hope to promote this attitude. An interesting question for the KBS community is whether some of our methods for knowledge representation and reasoning can be used to assist in understanding, representing and interpreting such frameworks. This paper concentrates on what is (arguably) the area of greatest concern: relating system requirements to high level design. We highlight what appear to be the major difficulties which face us in this area, using examples from systems which have been built to tackle them.