【摘要】 针对神经网络诊断模拟电路故障中可能存在误诊这一不足,从多分辨分析理论出发,将一维小波的尺度函数和小波函数进行张量积,把得到的张量积小波的尺度函数和小波函数共同作为网络的激励函数,构造了一种小波神经网络,并用该小波神经网络诊断实例模拟电路.仿真结果表明,该小波神经网络不仅能够诊断出已训练的故障类型,而且还能对新故障进行正确的分类,避免了误诊情况的出现.更多还原

【Abstract】 The application of conventional neural network in analog circuit fault diagnosis might lead to false diagnosis.A wavelet neural network is proposed according to multiresolution analysis theory.Firstly,constructing tensor product scaling function and tensor product wavelet function according to the theory of tensor product;then the tensor product scaling function and wavelet are adopted as the activation functions for wavelet network.The proposed wavelet neural network is applied in analog circuit fault diagnosis.Experimental results show that this wavelet neural network can exactly diagnose those faults which have been trained.Furthermore,it can classify new faults exactly.Therefore,it can avoid the problem of false diagnosis.更多还原