Diagnosis of Interturn Short Circuit of Permanent Magnet Synchronous Motor Based on Stacked Normalized Sparse Autoencoder

Abstract

Regarding the problems such as the traditional auto-encoders’ tendency to learn the similar features during the feature extraction and limited capability of feature learning in the shallow network models, the paper puts forward an interturn short circuit diagnosis method for Permanent Magnet Synchronous Motor (PMSM) based on the Stacked Normalized Sparse Autoencoder (SNSAE). The method is to guarantee the representativeness of the extracted features, acquire stronger anti-interference capability, and is more suitable for training diversified and large-capacity mechanical equipment failure data. Four types of Inter-Turn Short-Circuit (ITSC) fault data were collected on the PMSM fault diagnosis experimental platform for testing. The experimental outcome suggests that the SNSAE algorithm can effectively extract the fault features, and the fault diagnosis effect is better than the traditional intelligent diagnosis method, and its diagnosis accuracy is as high as 98.91%.

Online diagnosis for rolling bearings based on multi-channel convolution and transfer learning

“…However, the amount of network computation has also increased exponentially, which is not conducive to the construction of models. To solve this problem, this paper superimposes two convolutional layers with the kernel size of 3 × 3 to deepen the convolutional depth, which is equivalent to a convolutional layer with the kernel of size 5 × 5 [33]. It can greatly reduce the computing parameters and improve the expression capacity of the network.…”