一种基于多通道卷积神经网络的小齿轮轴裂纹诊断方法

A Method for Pinion Shaft Crack Diagnosis Based on Multi-Channel Convolutional Neural Network

分析现有轨道车辆小齿轮轴故障诊断的技术特点,提出一种基于多通道卷积神经网络的小齿轮轴裂纹诊断方法。对轨道车辆电机输出端附近的振动加速度信号进行短时傅里叶变换,得到二维时频复数矩阵。将二维时频复数矩阵拆解成多通道后,压缩到统一大小,输入到CNN中训练获得诊断模型。通过小齿轮轴实测信号验证了本文方法的有效性与泛化能力,诊断精度高达98%,优于单通道二维时频矩阵变换后输入到CNN模型。该方法为小齿轮轴裂纹故障诊断提供了新途径。

Based on the analysis of the existing technical characteristics of pinion shaft fault diagnosis for rail vehicles,this paper proposed a method of pinion shaft crack diagnosis based on multi-channel convolutional neural network(CNN).Short-time Fourier transform is performed on the vibration acceleration signal near the output terminal of the rail vehicle motor to obtain a time-frequency complex matrix,and two-dimensional time-frequency complex matrix was disassembled into multiple channels,compressed to a uniform size and put into CNN to train to obtain a diagnostic model.The effectiveness and generalization ability of the method are verified by the measured signals of the pinion shaft.The diagnosis accuracy is as high as 98%,which is superior to the single channel two-dimensional time-frequency matrix transformation and input into CNN model.The proposed method provides a new approach for the pinion shaft crack diagnosis.