Fault Diagnosis Method Based on Fuzzy Interval Optimization Using Fuzzy Inference

This paper presents a fuzzy rule extraction method based on optimal triangular equipartition.An optimization process based on the optimal criterion of fault identification rate is implemented to adaptively obtain the infimum of the number of fuzzy intervals,into which the input and output spaces of the given fault sample data should be divided.Subsequently,an improved DM algorithm is proposed to generate fuzzy rules from the given sample data based on the fuzzy intervals that compose the fuzzy inference system.Compared with factitious or random partition that the number of fuzzy intervals is lower than that of the infimum,the proposed method can generate more reasonable fuzzy inference system with higher fault identification rate and diagnostic accuracy.Moreover,the calculation burden in the process of fuzzy inference is significantly reduced because of smaller number of generated rules.As a result,the blindness in factitious or random partition is avoided.Application to fault diagnosis of track circuit shows the effectiveness of the new method.