Deep learning-based calibration method for material parameters of resilient wheel rubber components

In this article, the hyperelastic material of resilient wheels of subway vehicles is taken as the research object, a method system of identifying and calibrating the constitutive parameters of hyperelastic model based on joint experiment-simulation-deep learning is proposed. The theoretical analytical expression of the true stress-strain on stretch of the hyperelastic model Yeoh model under uniaxial loading condition is deduced, the accurate stress-strain curve data are captured by performing the compression experiment of cylindrical specimen of rubber component of urban rail transit system resilient wheel, and the initial values of parameters of the Yeoh hyperelastic model are fitted by the experimental data. The true stress-strain response samples of compression specimens under different Yeoh model parameter conditions were obtained by finite element numerical simulation. The Yeoh model's optimal parameters are obtained by training the data using a deep learning technique under the specified compression test stress-strain circumstances. Finite element numerical simulation is utilized to confirm the parameters' accuracy. The radial stiffness test of the resilient wheel rubber component was carried out, and the examination of the resilient wheel rubber component's stiffness analysis using the model parameters that were optimized. The study's findings indicate that the methodological system of identifying and calibrating the hyperelastic model constitutive parameters of the hyperelastic model proposed in this paper by combined experiment-simulation-deep learning has high prediction accuracy for the hyperelastic constitutive model parameters.