This study proposes an AutoML-based surrogate-modeling framework for crash box design optimization. A dataset consisting of 96 explicit finite-element crash simulations was generated for crash boxes with a wall thickness of 1.0–2.5 mm and six cross-sectional profiles: square, rectangular, circular, elliptical, hexagonal, and octagonal. AutoGluon-TimeSeries was used to benchmark machine learning models for predicting absorbed crash energy and crush displacement from geometric design variables. Among the evaluated models, Weighted Ensemble achieved the highest accuracy for absorbed energy prediction, with an R² = 0.975, while Temporal Fusion Transformer achieved an R² = 0.9863 for deformation-history prediction. The selected surrogate models were then used to support iterative sampling of the design space, significantly reducing the number of expensive finite-element evaluations required for optimization. These results show that AutoML-based surrogate modeling can accelerate crash box design and provide a practical foundation for data-efficient crashworthiness optimization.

This paper presents a sensitivity analysis of a reinforcement learning (RL) controller for stabilization of the Furuta pendulum under parameter variations. A high-fidelity multibody model of the system is first developed in Adams and used for policy training and closed-loop assessment. The RL agent is trained on the nominal model to stabilize the pendulum about the upright equilibrium. The trained policy is then evaluated by independently perturbing selected physical parameters within prescribed intervals around their nominal values. The resulting responses are used to assess the sensitivity of the learned policy to parameter variations. A trial is classified as failed when either the pendulum angle or the rotary arm exceeds its prescribed limit. The analysis identifies the parameters that have the greatest influence on policy performance and estimates the parameter ranges over which the trained policy continues to stabilize the pendulum. These results provide insight into the sensitivity of the trained policy to modelling errors and physical parameter variations.