A physical‒data-driven combined strategy for load identification of tire type rail transit vehicle

Abstract

Tire load is one of the basic input parameters required for vehicle design and safety evaluation. Identifying the tire load with high accuracy is of great significance. However, the direct measurement of tire load is often high-cost and complex. Meanwhile, load identification solely based on physical measurement or data has great limitations of low accuracy and low robustness. This paper proposes a physical‒data-driven combined load identification strategy that includes an extended Kalman filter (EKF) and a data-driven correction model. These two models are configured in series. Derived from a physical state-space model of the vehicle dynamics, the EKF is used for the preliminary identification of the load. The data-driven model extracts the spatial and temporal characteristics of signals through the convolutional neural network and bidirectional gated recurrent unit, and then predicts the errors of the extended Kalman filter and corrects the identified results. The presented strategy is applied to an APM300 rubber wheel vehicle for load identification. The results have shown that the physical‒data-driven combined strategy can reduce the influence of parameter perturbation and improve the identification accuracy (6.4%). Thanks to organically combining the physical- and data-driven methods and integrating the rules and experience of the entire system, the strategy has strong generalization performance. Compared with traditional algorithms, the presented strategy could effectively reduce the error of load identification, improve adaptability under different operating conditions, and handle the measurement error of different noise levels, which are of practical application value in the engineering field.