Fault Diagnostic Method Using a Series of Multi-Matrix

Abstract

We have developed a fault diagnostic method using a series of multi-matrix which have the knowledge of diagnostic objects. These knowledge are acquired by designers and well trained maintenance-crews coresponding with checking work of the diagnostic objects.
In this paper, we propose and evaluate the following items: the calculation method with the certain factors of the causes that uses the expanded Bayesian approach, the handling method of unknown relations of causes and results, the calculation method of certain factors in a series of multi-matrix, the order control method of answering the questions, and the hypothesis diagnosis for unidenified cause.
At the results, it is easily to acquire the knowledge and to reduce errors of the knowledge on relatively smaller size of matrix. Also, it is clarified that the restriction of the knowledge representation can be reduced by handling of the unknown relations, the unidentified observations, and “not yet known observations”. The calculation method with of certain factors of the causes is easy to use.
In the diagnosis system of using these method, when a malfunction happens, the system selects the matrix according to the malfunction, and the operator input answers for the system's questions. The system searches and shows the causes with certain factors and relations of causes and results. Also, the diagnostic objects diagram is shown on CRT, and then the parts of diagram twinkles, and colors to show the causes and the observative points.