Dependability evaluation of a computing system for traction control of electrical locomotives

Abstract

This article presents the dependability analysis of a computing system controlling the traction system and especially the semiconductor current-converters of a modern electric locomotive. Following special aspects of this application are taken into account: different degrees of performance reduction, lurking errors, periodic tests, short repair times. The dependability evaluation started from a simple “symptomatic model” showing the stages of performance degradation. It turned out to be a very helpful visualisation aid when discussing the failure modes with the experts for the components concerned. The detailed knowledge about hardware failures and their effects was collected in one large FMEA-table. For the subsequent mathematical analysis an elaborate “FMEA-oriented Markov model” was automatically constructed from the FMEA-table. This approach proved to be efficient and straightforward, giving clear results and hints on which components have most influence on MTTF, which must possibly be redesigned or planned redundantly. The customer's special requirements could be taken into account by arbitrarily varying the sets of up- and down-states. The approach is assumed to be applicable to many similar problems of dependability evaluation.