A DECISION-THEORETIC METHOD FOR THE CLASSIFICATION OF INCIPIENT-FAILURE PATTERNS WHICH ARE CHARACTERISTIC OF DETERIORATING MINE POWER-SYSTEM COMPONENTS
An integral aspect of mining-system automation is prediction of incipient failures in the mine power system. Previous research has established the existence of patterns which characterize incipient failures. The pattern elements include parameters which are computed from both the time and frequency domains. Utilizing concepts of artificial intelligence, the required measurements for developing these patterns are simple and economical to perform. A decision-theoretic method has been developed to classify incipient-failure patterns. Preprocessing transforms were developed for the pattern vectors to improve the probability of correctly classifying the failure patterns. The resulting algorithms were evaluated for failure modes involving portable-cable-connected motors, which are common in mining systems. The author's previous research, which provided the starting point for this thesis, is summarized. The experimental methodology is described. The theoretical development of a decision-theoretic method is presented. This dicussion is then followed by a description of the results of an experimental implementation and evaluation of the algorithms.
