Abstract:Accurately predicting the transient temperature field variation of an electric drive gearbox can effectively monitor its future abnormal temperature state, and ensure vehicle safety. However, establishing an accurate thermal network model is critical due to the complex spatial–temporal coupling and transformation relationship of the transient temperature field. This paper proposes a spatial–temporal merge relational graph convolutional neural network model, based on least squares method, for transient temperature field prediction. The model leverages dynamic learning of thermal diffusion rules from sparse monitoring data across diverse electric drive gearboxes. First, lumped parameter thermal network method is used to establish the thermal network model of electric drive gearbox, yielding a static binary adjacency matrix by mining the correlation between temperature nodes via loss analysis. Then, Euclidean distance is used to learn spatial similarity of each node and the adjacency matrix is dynamically updated. Finally, static and dynamic graph features are weighted, combined, and fed into an improved graph convolutional neural network to extract spatial–temporal correlation characteristics across temperature nodes. Simultaneously, the proposed prediction model was validated through both real-vehicle test data and experimental data from different test bench. Compared to traditional gated recurrent units and graph convolution neural networks, when the prediction time was 10 s, the global maximum prediction error of the proposed prediction model decreased by 5.6 %∼30.4 %, controlled within 3.84 ∼ 5.81°C. The root mean square error declined by 15 %∼36.5 %, within 1.04 ∼ 1.30°C. Therefore, the proposed prediction framework is capable of effectively capturing the dynamic changes in the thermal network topology and accurately predicting future transient temperature field changes based on information from multiple sensors. This demonstrates the model's adaptability to various types of electric drive gearboxes.

GCG: Graph Convolutional network and gated recurrent unit method for high-speed train axle temperature forecasting

GA-GRGAT: A Novel Deep Learning Model for High-speed Train Axle Temperature Long Term Forecasting

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

The Temperature Prediction of Hydro-generating Units based on Temporal Convolutional Network and Recurrent Neural Network

Fault Diagnosis Method for Bearing of High-Speed Train Based on Multitask Deep Learning

Data Based Fault Diagnosis of Hot Axle for High-Speed Train

Fault detection of high-speed train axle bearings based on a hybridized physical and data-driven temperature model

High Speed Train Axle Temperature Prediction Based on Support Vector Regression

LSTM-Based Temperature Prediction for Hot-Axles of Locomotives

A Prediction Method Based on Stepwise Regression Analysis for Train Axle Temperature

A Multichannel MN-GCN for Wheelset-Bearing System Fault Diagnosis

Intelligent Health Assessment of Aviation Bearing Based on Deep Transfer Graph Convolutional Networks under Large Speed Fluctuations

Data driven prediction of hot-Axles: A case study of prognostics and health management for locomotives

GC-LSTM: A Deep Spatiotemporal Model for Passenger Flow Forecasting of High-Speed Rail Network.

Analysis of the Temperature Characteristics of High-speed Train Bearings Based on a Dynamics Model and Thermal Network Method

Semi-supervised fault diagnosis of wheelset bearings in high-speed trains using autocorrelation and improved flow Gaussian mixture model

Spatial-temporal gated graph convolutional network: a new deep learning framework for long-term traffic speed forecasting

Prediction of Train Station Delay Based on Multiattention Graph Convolution Network

Digital Twin-Driven Graph Convolutional Memory Network for Defect Evolution Assessment of Rolling Bearings

Application of Spatio-Temporal Attention Mechanism in Temperature Prediction of High-speed Train Bogie

Transient temperature field short-term prediction of electric drive gearbox based on thermal network static and dynamic graph construction