基于VMD灰度图像编码和CNN的多传感融合轴承故障诊断

Multi-sensor fusion bearing fault diagnosis based on VMD gray level image coding and CNN

针对滚动轴承振动信号非平稳、非线性且易受噪声干扰的特点,以及单一振动信号对某些轴承故障识别率偏低的问题,提出一种基于变分模态分解(variational mode decomposition,VMD)灰度图像编码和卷积神经网络(convolutional neural networks,CNN)的多传感融合轴承故障诊断方法。首先,采用VMD对驱动端和风扇端振动信号分解,提取各阶本征模态分量与原始信号相关系数最大的分量;其次,将筛选出的本征模态函数(intrinsic mode function,IMF)分量依次排列并转换成灰度图像;最后,设计CNN结构,将训练集输入网络进行训练,测试集验证网络的有效性,实现滚动轴承故障识别。CWRU数据集和西安交通大学XJTU-SY数据集测试准确率分别达到99.90%和100%,结果表明:该方法能够准确识别变工况下轴承故障类别及损伤程度;对原始信号加入高斯噪声后的测试准确率分别达到99.75%和99.90%,证明该方法具有良好的泛化能力和抗噪性能。

Here,aiming at rolling bearing vibration signals having characteristics of non-stationary,nonlinear,susceptible to noise interference and lower recognition rate of a single vibration signal for some bearing faults,a multi-sensor fusion bearing fault diagnosis method based on variational mode decomposition(VMD) gray level image coding and convolutional neural network(CNN) was proposed.Firstly,VMD was used to decompose vibration signals of drive end and fan end of the target equipment,and extract components most strongly correlated to the original signal in various intrinsic mode function(IMF) components.Secondly,the filtered IMF components were arranged in sequence and converted into gray level image.Finally,a CNN structure was designed,data set to be tested was input into CNN for training to verify the effectiveness of CNN,and realize rolling bearing fault recognition.Testing correctness rates of CWRU dataset and XJTU-SY dataset could reach 99.90% and 100%,respectively.The results showed that the proposed method can correctly recognize bearing fault types and damage degree under variable operating conditions;after adding Gaussian noise into the above 2 datasets as original signals,their testing correctness rates reach 99.75% and 99.90%,respectively to reveal the proposed method having good generalization ability and anti-noise performance.