完备变分模态分解和多传感器卷积神经网络的轴承故障诊断方法

Fault Diagnosis of Rolling Bearing with Complete Variational Mode Decomposition and Multi-sensor Convolution Neural Network

针对传统滚动轴承故障诊断方法难以提取和辨识故障特征等问题,提出一种完备变分模态分解(CVMD)和工业多传感器卷积神经网络(MSCNN)相结合的轴承故障识别模型。在采集到的滚动轴承故障振动数据中加入2对符号相反但幅值相等的白噪声,并使用变分模态分解将故障振动数据分解为若干本征模态分量(IMFs)并进行集成平均;利用综合指标选择合适的IMFs分量并重构;针对多传感器结构,在卷积神经网络的基础上,提出MSCNN网络,并将重构后的振动信号输入MSCNN进行自动特征学习与故障诊断。结果表明:所提出的CVMD-MSCNN模型的故障诊断准确率达99.76%,标准差为0.16,相比于其他深度学习方法,其诊断准确率和稳定性较优。

Aiming at the problem that traditional rolling bearing fault diagnosis methods are difficult to extract and identify fault features,a method based on complete variational mode decomposition(CVMD)and multi-sensor convolution neural network(MSCNN)was proposed.The white noise pairs with equal amplitude and opposite signs were added to vibration signals of rolling bearing,and variational mode decomposition(VMD)was used to decompose the signals into several intrinsic mode functions(IMFs)and then integrated average;the comprehensive indexes were used to selecte the suitable IMFs and then be reconstructed;based on convolution neural network,a MSCNN network was proposed for multi-sensor structure,and then the reconstructed vibration signals were fed into MSCNN for automatic feature learning and fault diagnosis.The results show that the diagnosis rate of the proposed CVMD-MSCNN model reaches 99.76%and the standard deviation is only 0.16,compared with other deep learning models,the CVMD-MSCNN has great advantages in diagnosis accuracy and stability.