Fault diagnosis of ZDJ7 railway point machine based on improved DCNN and SVDD classification

Problems such as poor noise immunity and overfitting are prone to occur when convolutional neural network (CNN) is exploited in the fault diagnosis of ZDJ7 railway point machine. In addition, some features unbalanced have multiple tags, which lead low accuracy. Therefore, an improved deep (DCNN) support vector data description (SVDD) classification proposed. First, depthwise separable convolution Xception structure used optimize extraction features. Second, adaptive batch normalization processing (AdaBN) performed improve immunity. Meanwhile, global average pooling layer (GAP) instead fully connected generalization ability network. Aiming at machine sample, quantity learning algorithm for hypersphere coordinate mapping based on SVDD The realized under samples. experiment shows that accuracy DCNN 96.59%. It has a good anti-noise performance different kernels SNRs. When sample distribution unbalanced, indexes obtained by proposed model best.