论文部分内容阅读
研究了分形理论、小波变换与人工神经网络相结合进行故障诊断的机理与方法。利用小波包可进行多维多分辨率的特性,对振动信号进行分解与重构,提取频带能量特征分析。选用分形理论中的离散信号分形维数计算方法,提取分形维数的特征。以K-L变换作特征降维,然后用基于梯度符号变化的局部学习率自适应误差反传算法的小波神经网络对故障状态进行分类识别。并利用这种方法本文对风机转子故障进行了诊断,结果表明这种诊断方法是完全行之有效的。
The fractal theory, wavelet transform and artificial neural network are combined to diagnose the fault mechanism and method. The multi-dimensional and multi-resolution characteristics of the wavelet packet can be used to decompose and reconstruct the vibration signal and extract the energy characteristics of the frequency band. The method of fractal dimension calculation based on fractal theory is used to extract the features of fractal dimension. K-L transform is used to characterize the dimensionality reduction, and then the wavelet neural network adaptive error backtracking algorithm based on the gradient sign change is used to classify the fault states. And this method is used to diagnose fan rotor fault, the results show that this method is completely effective.