Enabling Hybrid Modeling in Commercial MBS Software: A Force-Level Approach

Hybrid modeling aims to combine physical and data-driven models to increase simulation accuracy without losing physical interpretability. In the context of dynamic mechanical systems, this enables the compensation of modeling inaccuracies that arise from simplifications, missing effects, or uncertain parameters. In this work, a hybrid model is used as a starting point, in which the discrepancy between simulation and measurement is learned and compensated by a data-driven correction element. To integrate such models into commercial multibody system (MBS) software like MSC Adams and Simpack, the formulation is adapted to operate directly on the force level. This allows implementation via standard co-simulation interfaces without modifying the system’s differential equations or solvers. The method is demonstrated using a single-mass oscillator with synthetic measurement data. Results show that the coupled simulation works reliably and that the hybrid model significantly improves accuracy while remaining compatible with established industrial simulation workflows.