Case-based diagnosis of multiple faults

Abstract

In order to maintain complex technical systems e.g. a telecommunication network, a rapid and precise recognition of faults and critical situations is required. But the large number of different components, the high degree of interdependencies among the components and the permanent changes in these systems make the diagnosis of faults and critical situations difficult.

Case-based Reasoning (CBR) seems particularly well suited for such a diagnosis. The ability to reuse old experiences by adapting them to current needs offers a possibility to handle the large number of diagnoses as well as the changes in the systems. However, the large number of components results in the problem of diagnosing multiple independent faults which leads to serious difficulties in retrieval and adaptation.

This paper focuses on the specific problems of diagnosing multiple independent faults. An incremental retrieval and adaptation approach based on an hypothesis and test strategy is introduced which is able to cope with data of multiple overlapping problems. At the end of this paper is given a report about the experiences in using the described algorithm for handling multiple faults in telecommunication networks.