基于Rep-VGG的滚动轴承故障诊断

Fault diagnosis of rolling bearing based on Rep-VGG

为解决传统的轴承故障诊断过于依赖人为经验且耗时耗力的问题,文中提出一种基于Rep-VGG模型的故障诊断方法。首先,通过希尔伯特和小波变换对原始振动信号数据进行预处理,将其转化为可供Rep-VGG网络识别的时频图形式;然后,利用Rep-VGG模型进行训练和测试,实验数据来源于凯斯西储大学公开的轴承数据集,并与其他模型进行对比。实验结果表明,所提方法对于轴承故障的诊断准确率达到99.9499%,损失仅为0.0221%;通过混淆矩阵得到Rep-VGG模型将不同类型的故障进行分类的准确率达到99.3%,与VGG-16相比,准确率提升5.3499%,说明该模型具有广泛的应用前景。

In allusion to the problem that the traditional bearing fault diagnosis relies too much on human experience and is time-consuming and labor-intensive,a fault diagnosis method based on Rep-VGG model is proposed.The original vibration signal data is preprocessed by means of Hilbert and wavelet transform,and it is converted into a time-frequency map form that can be recognized by Rep-VGG network.The Rep-VGG model is used for training and testing,and compared with other models,with the experimental data sourced from the bearing dataset published in Case Western Reserve University.The experimental results show that the proposed method has a diagnostic accuracy of 99.9499% for bearing faults,with a loss of only 0.0221%.By means of the confusion matrix,the accuracy rate of Rep-VGG model to classify different types of faults is 99.3%,which is 5.3499% higher than VGG-16,indicating that the model has a broad application prospect.