基于GRU与PCA-TL的热应力下IGBT故障预测

Fault Prediction of IGBT under Thermal Stress Based on GRU and PCA-TL

针对热应力下绝缘栅双极型晶体管(insulated gate bipolar transistor,IGBT)的性能随时间逐步退化的特性,将深度学习中的时间序列预测算法应用到IGBT故障预测中,提出了基于门控循环单元(gated recurrent unit,GRU)与主成分分析-迁移学习(principal components analysis-transfer learning,PCA-TL)的故障预测新方法。该方法以电参数集电极-发射极电压V_(CE)作为衰退参数,采用GRU模型构建衰退参数与故障时间的映射关系;利用PCA技术综合相异分布特征的IGBT故障指标,引入TL方法,通过微调GRU预测模型的参数完成从源域到目标域的迁移,实现目标域样本的故障预测。实验结果表明,基于GRU的故障预测模型具有较高的预测精度,与长短期记忆(long short-term memory,LSTM)算法相比,训练速度更快;PCA-TL方法可实现同类器件不同工况下的故障监测任务。验证了所提方法的可行性和正确性。

Aiming at the characteristic that the performance of insulated gate bipolar transistor(IGBT)degrades gradually with time under thermal stress,the time series prediction algorithm in deep learning was applied to IGBT fault prediction,a new fault prediction method based on gate recurrent unit(GRU)and principal components analysis-transfer learning(PCA-TL)was proposed.The method was taken the collector emitter voltage(VCE)as the decay parameter,and was used GRU model to construct the mapping relationship between decay parameter and fault time.The PCA technology was used to synthesize IGBT fault indicators with different distribution characteristics,and TL method was introduced.By fine tuning the parameters of GRU prediction model,the migration from source domain to target domain was completed,and the fault prediction of target domain samples was realized.The experimental results show that the fault prediction model based on GRU has higher prediction accuracy and faster training speed than long short term memory(LSTM)algorithm;PCA-TL method can realize the fault monitoring task of similar devices under different working conditions.The feasibility and correctness of the proposed method are verified.