Study on fault-diagnosis models of different neural networks and ensemble

Different diagnosis models, including multiplayer perceptron (MLP), radial basis function (RBF) and two types of support vector machines (SVMs), were designed, analyzed and compared based on the fault diagnosis of an analogue circuit instance. The experimental results show SVM model is of higher classification rate than MLP and RBF models, while MLP model has better ability to deal with uncertain signals. Considering different models correspond to different strategies, we combine four models of MLP, RBF and two SVMs to combine a diagnosis ensemble, which can achieve more accurate results than any individual model in the ensemble. The ensemble technique can provide a theoretical basis for further study on the fault diagnosis of analogue circuits.