Addressing the susceptibility of rail transit air supply system to faults under high-intensity and high-load daily operations,this study presents a fault warning framework tailored to the intermittent operation of the core component,the air compressor.The framework is designed to handle the uneven temporal distribution of monitoring data due to various influencing factors.Multidimensional temporal features are extracted from different types of data,and the Informer model is utilized to learn the characteristics of normal operation,with the TPE algorithm optimizing model parameters.A method for calculating a warning index is proposed based on the analysis of residuals between predicted and actual measurements.Simulation experiments on the MetroPT3 Air Supply fault dataset confirm the framework's can issue warning signals as early as 220 cycles prior to a fault and as late as 22 cycles before the fault occurs,and exhibiting its robustness against false alarms.
关键词
风源系统/空气压缩机/深度学习/故障预警
Key words
air supply system/air compressor/deep learning/fault prognosis
维普
万方数据