Fundamental Study for AI-Based Impact Analysis of Structural Elements in Wooden Structures

This study presents a fundamental validation of an AI-based impact analysis framework for wooden structures, aiming to support efficient and automated engineering judgment in seismic design. Focusing on a single-story residential building, the proposed method quantitatively evaluates the influence of individual seismic elements and their spatial lo-cations on structural response. Numerical time-history analyses were conducted using a detailed three-dimensional nonlinear model, and parametric variations of stiffness and strength were systematically generated using an orthogonal array. Machine learning models were then trained to capture the relationship between these parameters and seis-mic responses, and explainable artificial intelligence (XAI) techniques were applied to in-terpret parameter influence. The results demonstrated that wall elements oriented parallel to the target inter-story drift consistently exhibited dominant influence, which is consistent with structural engineer-ing knowledge. In addition, model comparison revealed that linear regression achieved high accuracy in the elastic response range, while Gradient Boosting outperformed other models under strong excitation conditions involving plastic behavior. This difference re-flects the transition from approximately linear to highly nonlinear structural response. These findings suggest that a hybrid modeling strategy combining interpretable linear models and flexible nonlinear models is effective for impact analysis. Overall, this fundamental study demonstrates that the proposed AI-based framework provides a transparent, rational, and time-efficient tool for seismic performance evalua-tion of wooden structures, bridging data-driven analysis and practical engineering deci-sion-making.