基于ISAM-Drsnet的故障识别模型及其应用

Fault identification modal and its application based on ISAM-Drsnet

针对滚动轴承故障诊断时网络模型在复杂环境下有效特征提取困难,无法充分挖掘具有周期性的滚动轴承故障数据时序特征的问题,提出了一种基于改进条纹注意力机制与深度残差收缩网络的滚动轴承故障诊断模型(ISAM-Drsnet)。首先,采用递归图(RP)编码方式生成了二维图像,使用ISAM和改进软阈值算法加强了Drsnet;然后,采取重叠采样的方式对数据集进行了增强处理,并将数据输入到ISAM-Drsnet中,实现了对不同故障类型的识别目的;最后,利用凯斯西储大学滚动轴承数据集进行了实验,选取了最佳数据截取长度,研究了改进软阈值、数据集规模、噪声对模型的影响;同时,将该模型与支持向量机(SVM)、反向传播神经网络(BPNN)、卷积神经网络(CNN)等进行了对比分析,并采用混淆矩阵等可视化方法对该模型进行了性能评估。实验结果表明:该模型(方法)的故障诊断性能明显优于SVM、BPNN、CNN等模型,其故障诊断精度可达99.79%,相比原始的Drsnet上升了1.60%;且在数据集规模有限和信号添加噪声的情况下,模型仍具有较高的故障诊断精度。研究结果表明:该轴承故障诊断模型不仅具有优秀的诊断性能,同时还具有较强的鲁棒性。

Aiming at the difficulty of extracting effective features of the network model in a complex environment during rolling bearing fault diagnosis,and the problem of not being able to fully mine the time series characteristics of periodic rolling bearing fault data,a rolling bearing fault diagnosis model of improved strip attention mechanism-deep residual shrinkage network(ISAM-Drsnet)was proposed.Firstly,recurrence plots(RP)encoding method was used to generate two-dimensional images,and the ISAM and improved soft threshold algorithm were used to strengthen Drsnet.Then,the data set was enhanced by overlapping sampling,and the data were input into ISAM-Drsnet to realize identification of different fault types.Finally,through experiments on the rolling bearing data set of Case Western Reserve University,the optimal data interception length was selected,and the influence of improved soft threshold,data set size,and noise on the model was studied.At the same time,comparative analysis was conducted with models such as support vector machine(SVM),backpropagation neural network(BPNN),convolutional neural network(CNN),and performance evaluation was conducted using visualization methods such as confusion matrix.The experimental results show that the fault diagnosis performance of this method is significantly better than that of SVM,BPNN,CNN and other models,and the fault diagnosis accuracy can reach 99.79%.The accuracy is 1.60%higher than the original Drsnet network model.And in the case of limited data set size and noise added to the signal,the model still has high stability.The research results show that the bearing fault diagnosis model not only has excellent diagnosis performance,but also has strong robustness.