The Method of Building Neural Network Fault Diagnosis Expert System for Hydroelectric Generating Set

摘要

A method of building neural network fault diagnosis expert system for the hydroelectric generating set is discussed. Based on the fuzzy rules and its certainty factor transitive algorithm, the teacher samples of the neural network can be obtained. The weights and thresholds matrix of the neural network can be obtained by samples learning, so the neural network fault diagnosis expert system can be constructed and new fault samples can be diagnosed. This method can diagnose rapidly, it has strong generalization ability and high accuracy, it also can eliminate the conflict in the reasoning process. In this paper, the oil temperature of the thrust bearing is selected as fault diagnosis object.