基于EEMD和CNN-SVM的滚动轴承故障诊断

Rolling Bearing Fault Diagnosis Based on EEMD and CNN-SVM

针对滚动轴承振动信号易受外界噪声干扰、传统的故障诊断方法难以提取故障特征以及准确率低等问题,提出一种基于集合经验模态分解(EEMD)和卷积神经网络结合支持向量机(CNN-SVM)的滚动轴承故障诊断方法。利用EEMD算法对原始振动信号进行分解得到本征模态函数(IMF)分量,再由相关系数筛选最佳分量进行信号重构,得到降噪后的振动信号。将重构降噪后的振动信号转换为二维特征图输入卷积神经网络进行训练提取特征。最后将提取到的稀疏代表特征向量输入到支持向量机进行故障分类。实验结果表明:所提方法能有效降低噪声干扰,便于提取故障特征,与传统的故障诊断方法相比准确率更高,诊断速度更快。

Aiming at the problems that rolling bearing vibration signals are susceptible to external noise interference,traditional fault diagnosis methods are difficult to extract fault features and have low accuracy,a rolling bearing fault diagnosis method based on ensemble empirical modal decomposition(EEMD)and convolutional neural network combined with support vector machine(CNN-SVM)was proposed.The EEMD algorithm was used to decompose the original vibration signal to obtain the intrinsic modal function(IMF)components,and then the correlation coefficient was used to filter the best components for signal reconstruction,and the noise-reduced vibration signal was obtained.The reconstructed and noise-reduced vibration signal was converted into a 2D feature map and inputted into the convolutional neural network to train and extract features.Finally,the extracted sparsely represented feature vector was inputted into support vector machine for fault classification.The experimental results show that the proposed method can effectively reduce the noise interference and easily extract fault features.Compared with traditional fault diagnosis methods,the accuracy is higher,and the diagnosis speed of the improved model is faster.