ABSTRACT
Aiming at the problem of uncertainty in analog circuits, combining cloud model with probabilistic neural network, a fault diagnosis method for analog circuits is proposed based on n-dimensional normal cloud mode to optimize Probabilistic Neural Network (PNN). Firstly, the circuit voltage is used for Principal Component Analysis (PCA) to obtain sample set. Then the training sample under different failure modes through peak cloud transform to structure n-dimensional cloud model and it as PNN model neuron. Finally, weight between the model layer and the summation layer is obtained. The number of cloud model and threshold size is the key point of PCA-CPNN. The example shows that the method not only has a higher accuracy for analog circuit faults, but also can optimize three parameters of the PNN and greatly simplifies the neural network training process.
