Case-Based Reasoning and Adaptive Learning in the MOLTKE 3 Workbench for Technical Diagnosis

Abstract

In this paper, we deal with the support of integrated knowledge acquisition workbenches by case-based reasoning techniques. We illustrate that the problem of test selection has to be taken into account for the diagnosis of technical systems. We offer a possible solution which enables both the utilization of all the well-known advantages of case-based reasoning systems and the avoidance of its also known drawbacks. We concentrate on a case-based reasoning system which has its well-defined role within a fully implemented knowledge acquisition workbench for technical diagnosis, namely the processing of temporary and absolute exception cases. It is able to utilize all the workbench’s qualities, such as knowledge about the underlying technical system, its analogy-based rule generator, and its heuristic classificator. To meet the requirements which evolve from real world applications, our case-based reasoner deals not only with the classification problem of technical diagnosis, but also with that of test selection. Additionally, it learns the relevances of the symptoms for the respective diagnoses which enable the realization of an adaptive similarity measure. Having all this characteristics in mind our case-based reasoning approach defines a new state of the art for case-based reasoning systems (with the restriction to the field of technical diagnosis).