Data-driven dynamic modeling and real-time wear prediction for nonlinear and strongly coupled suspended monorail wheel

In the field of suspended monorail dynamics, traditional dynamic models have limitations in handling strongly coupled and nonlinear vibration processes, resulting in significant deviations between the real-time wear prediction of suspended monorail wheels and the actual values. This article proposes a data-driven dynamic modeling method to solve the problem of real-time wear prediction of suspended monorail wheels. First, this article demonstrates that the prediction accuracy of suspended monorail wheel wear is strongly correlated with the accuracy of wheel vibration acceleration, and uses a data-driven method to train the vehicle vibration spectrum features. Second, an enhanced genetic algorithm is designed to address the convergence problem and long training time in the training of strongly coupled dynamic models, significantly improving the computational efficiency. Finally, through verification under different track radii, it is shown that the wheel wear prediction values of the trained dynamic model are highly consistent with the vehicle wear values of the virtual experimental vehicle, and the accuracy is significantly better than that of the dynamic model before training.