第十二讲:一种新型二维故障特征提取方法

Chapter 12:A New Two-Dimensional Fault Feature Extraction Method

针对现有旋转机械轴承诊断方法存在故障特征提取不理想、一维数据作为神经网络输入数据时诊断精度低等问题,提出了一种基于能量频谱的二维旋转机械故障诊断新方法;该方法运用小波包分解对原始振动信号进行分解,并提取分解的每一节点信号的能量构建小波包能量频谱矩阵,在此基础上,基于凯西斯轴承数据集使用经典残差网络进行故障分类,结果表明,该方法对10种轴承故障的诊断精度达到了98%,说明基于能量频谱信息的二维旋转机械故障诊断方法提取深度特征的能力优越。同时通过故障信号分析表明,该方法能够从噪声干扰中有效提取到微弱故障特征,实现了轴承故障类型的准确判定,验证了该方法的有效性。

Aiming at the problems of unsatisfactory fault feature extraction and low diagnostic accuracy when onedimensional data is used as neural network input data,a new method for fault diagnosis of two-dimensional rotating machinery based on energy spectrum is proposed,which uses wavelet packet decomposition to decompose the original vibration signal,and extracts the energy of each node signal decomposition to construct a wavelet packet energy spectrum matrix.The diagnostic accuracy of 10 kinds of bearing faults in this method is 98%,which shows that the two-dimensional rotating machinery fault diagnosis method based on energy spectrum information has superior ability to extract depth characteristics.At the same time,the fault signal analysis shows that the method can effectively extract the weak fault characteristics from the noise interference,realize the accurate determination of the bearing fault type,and verify the effectiveness of the method.