Construction of Digital Twin Model for Train Simulation Operation Based on Gaussian Process Regression

The optimal design of a train scheduling scheme has been a focal point of research within the train dispatching domain. Given that the operation of the train is affected by multiple uncertain factors, this paper integrates digital twin technology into train simulation, adopts a physical train model as the foundation, and utilizes a virtual layer model approach. Based on a five-dimensional digital twin framework, the digital twin model of train simulation operation is constructed from three levels of elements, behaviors, and rules. Furthermore, a Gaussian process regression method is employed to perform in-depth machine learning on original training data, generating more precise driving data, which can be used to simulate train operations. Additionally, various charts are provided to facilitate real-time visualization of the simulation results. Theoretical analysis and experimental validation demonstrate the accuracy and usability of the proposed model, which exhibits high engineering application value and provides scientific support for optimizing train scheduling schemes.