基于1D-RSLBCNN的齿轮箱故障诊断

Gearbox Fault Diagnosis Based on 1D-RSLBCNN

针对传统故障模型参数多,训练和检测时间长的问题,提出了基于残差结构和局部二进制卷积(1D-RSLBCNN)的齿轮箱故障诊断方法。其利用局部二进制卷积层来替代传统卷积层,在减少模型参数的同时,加快了训练速度和收敛速度;同时在网络模型中引入残差结构,避免了由于网络深度的增加引起的正确率饱和甚至下降的问题。实验结果表明,局部二进制卷积层的参数量为传统卷积层的1/3,诊断准确率更是高达99.7%。与其他模型相比,具有更稳定、可靠的预测精度。

Aiming at the problems of traditional fault model with many parameters and long training and detection time,this paper proposes a gearbox fault diagnosis method based on residual structure and local binary convolution(1D-RSLBCNN).The local binary convolution layer is used to replace the traditional convolution layer,which accelerates the training speed and convergence speed while reducing the model parameters;At the same time,the residual structure is introduced into the network model to avoid the problem of saturation or even decline of accuracy caused by the increase of network depth.The experimental results show that the parameters of local binary convolution layer are one-third of those of traditional convolution layer,and the diagnostic accuracy is as high as 99.7%.Compared with other models,it has more stable and reliable prediction accuracy.