多工况条件下跨座式单轨列车齿轮箱故障振动信号趋势预测方法

打开文本图片集

关键词：单轨列车；趋势预测;双向门控循环单元；时序信息预测模型;空洞因果卷积中图分类号：TH165文献标志码：ADOI：10.7652/xjtuxb202511019 文章编号：0253-987X（2025）11-0198-11

Abstract： To address the low prediction accuracy of vibration signal trends in straddle monorail train gearboxes caused by low-frequency modulation interference， nonlinear coupled harmonics， and condition-dependent characteristics， this study proposes a vibration signal trend prediction model that integrates a bidirectional gated recurrent unit （BiGRU） and a temporal information prediction model （Informer）. This model utilizes the gating mechanism of BiGRU to suppress low-frequency interference， extracts nonlinear coupled harmonic features through its bidirectional structure，and enhances adaptability to various operating conditions through an adaptive learning mechanism. By combining the multi-head attention mechanism in the Informer encoder with dilated causal convolution，the model effectively expands its receptive field to capture long-term dependency features in time series data，thereby facilitating the trend prediction of gearbox vibration signals. Validation of the model is conducted using a dataset of gearbox fault data collected from a straddle monorail train gearbox test bench. The proposed model achieves average absolute error，mean square error，and root mean square error of O.264 8，0.116 O，and 0.339 1， respectively，across different prediction horizons，which are lower than those of the Autoformer， Informer，and Dlinear models. The prediction accuracy remains stable under four operating conditions： normal operation， assembly error， inner ring fault， and fatigue wear，with absolute erors ranging from O.21 to O.35. The results indicate that the proposed model demonstrates superior trend prediction accuracy for both short and long time series compared to other benchmark models，and the experimental outcomes under multiple operating conditions validate the model's adaptability and robustness.

Keywords：monorail train；trend prediction； bidirectional gated recirculation unit；temporal information prediction model; dilated causal convolutional

跨座式单轨列车是跨座在单轨道上行进的一种独特的公共交通形式，如图1所示。（剩余15295字）