基于Dropout-多尺度空洞卷积神经网络的轴承故障诊断

Rolling bearing fault diagnosis based on Dropout-multi-scale dilatedconvolution neural network

为了提高故障诊断模型对故障轴承低信噪比信号的特征提取能力,使模型在强噪声环境下仍能发挥作用,提出了一种基于Dropout-多尺度空洞卷积神经网络(D-MDCNN)的滚动轴承故障诊断模型。首先,通过Dropout数据预处理,对训练数据进行“损坏”,强迫模型仅依靠少量特征便可进行故障诊断,以提高模型的抗噪声能力;然后,使用不同扩张率的空洞卷积扩充了多尺度信息,并利用CNN模块来完成对特征的提取与故障诊断;同时,在模型中加入批量归一化处理操作,用来加快模型训练的收敛速度,提高了模型的性能;最后,利用美国凯斯西储大学轴承数据集和东南大学齿轮箱数据集对基于D-MDCNN的模型进行了实验验证,并将实验结果与采用其他深度学习模型所得的实验结果进行了对比分析。实验结果表明:在无噪声至4dB的噪声环境下,D-MDCNN在西储大学和东南大学两个数据集上均可取得99%的诊断准确率;相比于其他同类模型,基于D-MDCNN的模型具有更高的诊断准确率和抗噪声能力。研究结果表明:基于D-MDCNN的模型是一种有效的轴承故障诊断模型。

In order to improve the feature extraction ability of the fault diagnosis model to the low signal-to-noise ratio signal of the fault bearing,and make the model still play a role in the strong noise environment,a rolling bearing fault diagnosis model based on Dropout-multi-scale dilated convolutional neural networks(D-MDCNN)was proposed.Firstly,the Dropout data preprocessing was used to"damage"the training data,which was forced the model to diagnose faults by relying only on a few features,so as to improve the anti-noise ability of the model.Then,multi-scale information expansion was completed by using the dilated convolution with different dilation rates,and feature extraction and fault diagnosis were completed by using CNN module.At the same time,batch normalization was added to the model to accelerate the convergence speed of model training and improve the performance of the model.Finally,the model was verified by the bearing data set of Case Western Reserve University(CWRU)and the gearbox data set of Southeast University(SU),and the experimental results were compared with those of other deep learning models.The experiment results show that D-MDCNN can achieve 99%diagnostic accuracy on the bearing data set of CWRU and the gearbox data set of SU in noise free to 4dB noise environments,and has higher diagnostic accuracy and anti-noise ability than other models.The results show that the model based on D-MDCNN is an effective bearing fault diagnosis model.