基于遗传算法优化BP神经网络动车组塞拉门故障预测与诊断

Fault Prediction and Diagnosis of Sliding Plug Door in EMU based on Genetic Algorithms Optimized BP Neural Network

针对动车组塞拉门开关门故障预测与安全预警,提出了一种基于遗传算法优化BP神经网络动车组塞拉门故障预测方法,用于塞拉门开门时间均值和开门电流均值的预测;针对塞拉门开关门故障诊断和排除,提出了一种基于故障树分析的诊断方法。研究结果表明遗传算法优化BP神经网络的预测模型所预测的结果与真实值之间的相对误差均在±2%以内,通过故障树分析得出13个基本事件的关键重要度,其中机械锁闭装置卡滞(X5)、锁闭装置故障(X12)、解锁开关信号延时(X3)为影响塞拉门正常开关门可靠性的关键部件,研究结果为动车组塞拉门故障诊断和排除提供依据。

The fault prediction method based on genetic algorithm optimization and BP neural network was proposed for the fault prediction and safety early warning of sliding plug door in EMU,and was applied to predict the mean value of opening time and opening current of sliding plug door.A diagnosis method based on fault tree analysis was proposed for the fault diagnosis and elimination of the sliding plug door.The results indicate that the relative errors between the predicted results and the real values of the BP neural network model optimized by genetic algorithm are within±2%.Through fault tree analysis,the critical importance of 13 basic events is obtained,among which mechanical locking device stuck(X5),locking device failure(X12)and unlocking switch signal delay(X3)are the key components for affecting the reliability of normal switching door of sliding plug door.The research results can provide the basis for fault diagnosis and elimination of sliding door of EMU.