Adaptive model order reduction for high-fidelity off-road mobility simulation

For off-road mobility simulations, increasingly accurate models require increasingly long computation times. This increase in computation time has been prohibitive in applying these models to design processes in academia and industry alike. In order to improve the computation time while retaining the accuracy of high-fidelity models, this study explores the proper orthogonal decomposition (POD) as a method of model order reduction for nonlinear flexible multibody dynamics, specifically for high-fidelity off-road mobility models. Because the reduced order basis generated by the POD is highly situation specific, a process of mode adaptation using a method of interpolation on a tangent space of the Grassmann manifold is also investigated to make the modes more robust to changes in simulation parameters. The POD modes are found to be effective at improving computation time while retaining accuracy for a single tire soil bin mobility model and the adapted modes using the Grassmann manifold are found to be better than POD modes generated for different simulation parameters. The process is further applied to a full vehicle model to demonstrate the use of the POD-based model order reduction for off-road mobility prediction.