Abstract:
Although the traditional neural network algorithm has been successfully applied to the inter-turn short circuit(ITSC) fault diagnosis of permanent magnet motors, it is difficult to extract enough fault features under the noise interference, and the model hyperparameters selected by relying on experience cannot reach the optimal performance, which restricts its wide application. Thus, a multiscale convolutional neural network algorithm using Bayesian optimization(BO) is proposed for early ITSC fault diagnosis of five-phase permanent magnet synchronous motors(FPPMSM). Firstly, sufficient fault features are collected using a multiscale convolutional neural network(MSCNN). Secondly, the Bayesian optimization algorithm is introduced to achieve global optimization of model hyperparameters and to save tuning time. Finally, the algorithm achieves early ITSC fault diagnosis in different short-circuit situations by extracting the actual current signal data. To prove the effectiveness of the proposed algorithm, it is compared with other four algorithms and the experimental results show that the proposed method is more accurate.