Fault Diagnosis Method of Gearbox Based on STFT and 2DCNN-SVM
In order to improve the effectiveness and the accuracy of gearbox fault diagnosis,a gearbox fault identifica-tion method based on the combination of short-term Fourier transform(STFT),two dimensional convolutional neural net-work(2D CNN)and support vector machine(SVM)is proposed.The JZQ250 fixed axis gearbox experimental platform is built.The acceleration sensor is used to collect the gearbox vibration signal,and the STFT is performed on the vibration sig-nal to obtain a two-dimensional time-frequency map.Then,the time-frequency map is input into the 2D CNN for feature in-formation extraction.The time-frequency map information of different types of faults is trained through 2D CNN forward propagation and backward propagation to establish a deeper relationship between different types of features.The training set and verification set loss curves,accuracy curve and t-Distributed Stochastic Neighbor Embedding(t-SNE)visualization are used to reflect the training degree of the model.Finally,SVM is used to identify the fault type.By comparing the fault identi-fication results of gearbox with those of FFT-2D CNN,1D CNN-SVM and 2D CNN-SVM,the proposed method has the highest accuracy of fault identification of 97.94%,as well as good robustness.
万方数据
维普
