Modelling of an expert system for diagnosing the operational status of a diesel genset

Abstract

The modelling of an expert system for diagnosing the operational status of a diesel genset represents the final stage of the research undertaken so far on military gensets. The research was undertaken in several stages on a large number of gensets, and the results obtained at specific research stages were published at scientific-expert conferences. For the purpose of modelling an expert system, a knowledge base was compiled based on detailed observation of the genset operational status, interviews with experts with many years of experience in maintaining military gensets and also on a breakdown of the occurred faults archived in the overhaul documentation.The paper uses only that one segment of the knowledge base that is necessary for modelling a simplified form of the Bayesian network for fault detection in assembled condition. In addition to input probabilities, the results of diagnostic tests and simulations carried out in the Matlab Simulink program package are also entered in the Bayesian network. Fault detection represents a complex process so the application of an expert system significantly reduces the time needed for fault detection, resulting in optimized maintenance. It is especially significant in military and similar organizations which apply a large number of technical resources. The Bayesian network is processed in the GeNIe program package.