【摘要】 基于不同点蚀模式的轴承振动信号的频域能量分布差异性,提出了基于小波包正交分解和BP神经网络的轴承点蚀故障模式识别技术。对轴承振动信号进行小波包正交四层分解,实现了信号空间完整拆分的同时得到了第四层由低频到高频的小波包分解系数,再分别进行单支重构得到各频段的成分。利用信号各频段的能量组成特征矢量作为神经网络的输入样本,对BP神经网络进行训练,获得模式识别网络;再用新数据进行网络的检验,结果证明网络的性能良好。更多还原

【Abstract】 Based on the difference of spectrum energy distribution of different spot erosion bearings,wavelet packet orthogonal decomposition and back propagation neural network technology were used in bearing spot erosion pattern recognition.Bearing vibration signal was decomposed orthogonally into four layers through wavelet packet transform,fourth layer decomposition coefficients was obtained from low frequency to high frequency and signal space were divided completely meantime.Component of each frequency band was gained through single branch reconstruction with the reconstruction formula.Applying energy of each frequency band to as the input vector of neural network,then network was trained to achieve the pattern recognition performance.New sampled data was input into the trained network,the result demonstrated the effective performance.更多还原