High-speed Maglev Train Suspension Gap Prediction Based on CNN-AGCRN

Suspension gap is a critical parameter for the stability control of high-speed maglev trains. Due to the nonlinearity and strong coupling characteristics of the high-speed maglev suspension systems, predicting their gaps is challenging. This paper proposes a joint prediction model based on CNN-AGCRN for predicting the suspension gap of high-speed maglev trains. It establishes a model integrating Gated Recurrent Units (GRU) and Adaptive Graph Convolutional Neural Networks to extract the time dependencies and variable correlations of multivariate time series, and models the prediction results with Gaussian noise. The loss function is improved to generate point estimates and interval estimates of the suspension gap, thereby enhancing the accuracy and robustness of the predictions. The results show that the joint prediction model of CNN-AGCRN improves the prediction accuracy by 21.8% and 5.7% compared to GRU and LSTM models, respectively, and 97.08% of the actual suspension gap values in the test set fall within the predicted 95% confidence interval, demonstrating that this model achieves excellent results in predicting the suspension gap of high-speed maglev trains.