论文部分内容阅读
针对BP(back propagation)神经网络搜索速度慢、容易陷入局部最小的缺陷,提出了经验模态分解(EMD)遗传神经网络方法,首先用对带噪的信号进行分解,得到信号的各阶本征模函数分量,每个本征模函数分量对应着一个能量不同的频段,即一种故障特征,将各频段能量的特征向量作为优化神经网络的输入样本;其次用遗传算法对神经网络的初始权值和阈值进行优化.利用EMD遗传神经网络方法对滚动轴承多类故障信号进行分析,可提高故障识别能力.
For BP (back propagation) neural network, which is slow in searching speed and easy to fall into local minimum, an empirical mode decomposition (EMD) genetic neural network method is proposed. Firstly, the decomposed signals with noises are obtained, Each of the eigenmodular function components corresponds to a frequency band with different energy, that is, a fault feature, and uses the eigenvectors of energy in each frequency band as input samples for optimizing the neural network. Secondly, the initial weights of the neural network Value and threshold.Using EMD genetic neural network method to analyze multi-class fault signal of rolling bearing can improve fault recognition ability.