Fault Diagnosis Of Analog Circuit Based On The Information Entropy Feature Extraction Method

With the development of electronic technology, improved integration of electronic devices, circuit fault diagnosis is more and more important. In particular, fault diagnosis of analog circuits has become one of the important topic of study. Because of the advantages of artificial neural network about self-adaptability, learning ability, more efficient and accurate diagnosis, in the field of analog circuit fault diagnosis has a high practical value.In this paper, uncertainty of the circuit fault conditions are described with the Shannon entropy. And the diagnostic information of circuit fault conditions are defined by calculating the circuit difference of priori entropy and posteriori entropy. According to the diagnostic information, Particle Swarm Optimization(PSO) are adopted to search the best feature subset of the circuit fault condition. An experimental circuit has been picked out to verify this method, and neural network are used as a classifier. In this experiment, results show that the feature subset can be obtained using this method. And this feature subset can fully reflect the different malfunction conditions of the circuit and can effectively isolate different faults of the circuit under test.