基于深度学习的永磁同步电机故障诊断方法

FAULT DIAGNOSIS METHOD OF PERMANENT MAGNET SYNCHRONOUS MOTOR BASED ON DEPTH LEARNING

针对永磁同步电机匝间短路和永磁体失磁故障因处理复杂、特征独立单一和样本稀少等因素引起的诊断偏差问题,提出一种基于深度学习变分自编码网络的故障特征样本快速扩展策略及融合稀疏自编码网络的故障诊断方法。通过组合永磁同步电机频域电流、磁通密度、电磁转矩特征,结合变分自编码网络的生成模型实现故障真实样本扩张,构建丰富、多样、更具鲁棒性的训练集合。将优化数据集输入稀疏自编码网络训练诊断模型,用测试数据验证网络的优劣。实验结果表明,相比传统故障诊断方法,该算法能更加高效快速地实现匝间短路及失磁故障诊断。

Aiming at the problem of diagnostic deviation caused by complex processing,independent single feature and few samples of interturn short circuit and permanent magnet loss of excitation faults of permanent magnet synchronous motor,this paper proposed a fast expansion strategy of fault feature samples based on deep learning variational auto-encoder network and a fault diagnosis method combining sparse auto-encoder network.By combining the characteristics of frequency domain current,flux density and electromagnetic torque of permanent magnet synchronous motor,and combining with the generation model of variational auto-encoder network,the real fault samples were expanded,and we constructed a rich,diverse and more robust training set.The optimized data set was input into the sparse auto-encoder network training diagnosis model,and the test data were used to verify the network's advantages and disadvantages.The experimental results show that compared with traditional fault diagnosis methods,our method can achieve interturn short circuit and loss of excitation fault diagnosis more efficiently and quickly.