The Reserch Of The Analog Circuit Fault Diagnosis Based On Multiple Classifier

Along with the electronic scale is bigger and more bigger, the performance of the system and structure is more and more complicated, the urgent demands or testing and fault diagnosis of analog circuits are proposed, and artificial intelligence method based on neural network becomes one of research hot spots. The BP neural network is widely applied, and some achievements. But for BP algorithm, there are some shortcomings are difficult to overcome, such as generalization ability is weak, easily falling into local optimal solution.In view of the disadvantages of existing neural-network-based method for analog circuits fault diagnosis, a negative correlation algorithm of neural network ensemble for analog circuit fault diagnosis are proposed. Neural networks are trained by negative correlation algorithm can increase the integration degree of difference between the component networks, and improve the generalization ability of the neural network ensemble. Experimental results show that training the neural networks based on negative correlation algorithm, the ensemble of networks’ performance is not only better than a single network, but also than the traditional integrated neural network. This method can effectively improve the generalization ability of the ensemble of neural networks. Therefore, this proposed negative-correlation-based method for analog circuit fault diagnosis is practical and has reference value.