Home

Project Overview

Cost efficient operation in transportation, manufacturing or other industrial activities, depends on the efficient co-operation between a diversity of companies, departments and hierarchical levels of personnel. This is reflected in modern maintenance, and fault diagnosis activities. Absence of accurate and up-to-date information leads to repeatedly diagnosing similar problems, resulting in an enormous wastage of costs, manpower, time and materials. Especially during malfunctioning of systems, not only the accurate status of the problem is required, but also a detailed conclusion on appropriate recovery actions in function of the responsibilities and authorisation level of personnel and within time restrictions. The use of artificial intelligence techniques is essential to provide only relevant and up-to-date information.

The current market situation makes efficient and reliable fault diagnosis systems, and associated tools for the maintenance of the diagnosis systems, imperative throughout the life-cycle. In practice, explicitly modelling behaviour and malfunctioning of such large technical systems is impossible. Case-based reasoning (CBR) provides, in principle, the integrative framework that avoids the explicit behavioural modelling. It enables companies to capture and manage the knowledge and experience of all parties; relevant information for each party can be automatically retrieved to generate specific fault diagnosis systems, maintenance of all diagnosis systems is limited to the case-base only, for which machine learning techniques can be applied, and unambiguous exchange of information. Existing CBR tools, however, have serious limitations that prohibit their use in large technical applications.

The project proposal builds upon the very promising results of several prototype fault diagnosis tools developed by some of the partners in a small industrial try-out-project. These prototypes are currently being tested by industrial partners. Based on the conclusions of the prototype testing, the following issues should be resolved during the proposed new project; extension of the approach to hard real-time expert system techniques, evaluate and redesign of the algorithms of the prototype, development and maintenance tools, and extend the applicability of the approach to other domains.

The project has the specific goal to develop a generic fault diagnosis tool for large technical systems. The principles of CBR are used to declare and maintain a unique formulation of the malfunctioning of the entire application. Uncertainty in occurrence of failures and faults enables the more realistic declaration of knowledge and operational experiences on malfunctioning behaviour. Development of a network structure makes the diagnosis search process deterministic, much faster and applicable for large case-bases. For real-time on-line diagnosis, the response time can be predicted and guaranteed.

The consortium consists of 6 partners. Three universities DUT, UWS and TTU are included in the consortium for their research achievements on fault diagnosis, real-time systems, real-time expert systems and case-based reasoning. The role of the universities is to further develop their research achievements into the generic fault diagnosis tool. MiTime Ltd will take a major role on the software development and will ensure high-level software standards. Bombardier-Eurorail participates with one of their customers for the development of a test application on rolling stock. DACS participates for the development of a second application.

The 2-year project will result in the beta release of the inference kernel, thoroughly tested on 2 industrial applications (BE, DACS) and ready for commercialisation, to be embedded in technical environments. Apart from this kernel, two man-machine interfaces for rolling-stock and remote access applications will be available for commercialisation, to be extended to meet the demands of other applications as well.