Update Process of Maintenance Knowledge for Diagnosis System

Abstract

n order to maintenance the equipment, we developed a diagnosis system using a Bayesian net based on FMEA of the gas engine. However, since the probability setting of the Bayesian net, failure mode and symptom which are not in the FMEA occur, the problem is that the estimated accuracy is low to 56%. To solve this problem, we have developed an update process of maintenance knowledge to update the probability setting of the Bayesian net and FMEA. In the update process, the symptoms and failure mode found from the Maintenance history are added to the FMEA after reviewing to a maintenance expert, and the probability setting of the Bayesian net updated by learning the maintenance history. As a result, the accuracy was improved to 95%.