Evaluating Physics-Based, Hybrid, and Data-Driven Models for Rubber-Metal Bushings

Rubber-metal bushings (RMB) are critical components in multi-body systems, such as vehicles and industrial machinery, due to their abilityto enable relative motion, dampen vibrations, and transmit forces. However,their nonlinear behavior challenges accurate modeling. Traditional physics-based models often fail to balance simplicity, accuracy, and computationalefficiency. The growing availability of experimental data offers opportunitiesto improve RMB modeling through hybrid and data-driven approaches. Thisstudy evaluates physics-based, hybrid, and data-driven methods based on predictive accuracy, modeling effort, and computational cost. Hybrid approaches,combining machine learning techniques with physics-based models, are investigated to leverage their complementary strengths. Results show that hybridmethods enhance accuracy for simpler models with a modest increase in computational time. This highlights their potential to simplify RMB modelingwhile balancing accuracy and efficiency, offering insights for advancing multi-body system simulations. Building on these insights, data-driven methods areexplored for their ability to provide surrogate models for dynamical systemswithout requiring expert knowledge. Experiments reveal that while simpledata-driven methods approximate system behavior when data has low variance, they fail with trajectories of widely varying frequency and amplitude.