【摘要】 针对小波神经网络在模拟电路故障诊断中存在"维数灾"这一致命缺陷,根据多小波空间中函数的多分辨率分解思想,构造了一种激励函数为具有紧支撑集、对称性和正交性的多尺度和多小波函数的多小波神经网络.仿真结果表明,多小波神经网络收敛速度比小波神经网络要快得多,有效地解决了"维数灾"问题.仿真实验和理论分析一致.更多还原

【Abstract】 Wavelet neural network has it’s fatal defect "dimension disaster" for fault diagnosis in analog circuit.According to the multiresolution analysis thought in the multiwavelet space,a multiwavelet neural network model is obtained.Some compact support,symmetric and orthonormal multi-scale functions and multiwavelets are adopted as the activation functions in the model.Simulation results indicate that the multiwavelet neural network converges more rapidly than wavelet neural network,and avoids the "dimension disaster".Experimental results support the result of theoretical analysis.更多还原