基于卷积神经网络的时频图像识别研究

变速器作为汽车动力传递系统中的关键部件，其振动和噪声直接影响着汽车的性能。由发动机输入到变速器的转速很多情况下是变化的，这使得这种工况下的变速器故障诊断更加复杂。针对这个问题，提出了基于卷积神经网络（convolutional neural network，CNN）的变速器变转速工况下的故障分类识别方法:在变转速下，采集了变速箱多种故障状态下的振动信号，对各类信号进行时频变换得到时频矩阵，并利用CNN实现多类故障的分类。并研究了CNN结合不同时频方法时的识别性能，结果表明，连续小波变换(continuous wavelet transform，CWT)与CNN结合的方法对变转速下的时频图识别性能最好。

Time-frequency Image Recognition Based on Convolutional Neural Network

As a key part of the vehicle power transmission system, the transmission can directly affect the performance of an automobile by its vibration and noise. In many cases, the rotational speed of the input shaft of gearbox is changing, which adds to the complexity of the fault diagnosis. In response to this problem, the study presents a new method for the gearbox fault identification and classification based on convolutional neural network. The vibration signals of the gearbox under various fault conditions are collected, and all kinds of signals are transformed to time-frequency images by using the time-frequency analysis. Then the time-frequency matrices are input to the CNN for classification of different types of faults. And the recognition performance of CNN combined with different time-frequency analysis methods is studied. The results show that the method of CWT and CNN has the best performance in time-frequency image recognition with variable rotational speed of gearbox.