Seismic response prediction of asymmetric structures with SMA dampers using machine learning algorithms

Abstract

The dynamic response of asymmetric structures to seismic forces is challenging due to mass, stiffness, and damping distribution irregularities. Shape memory alloy (SMA) dampers have successfully dealt with these issues because of their distinctive super elasticity and energy dissipation characteristics. In this work, we study regression algorithms’ effectiveness in predicting the seismic behavior of asymmetric structures installed with SMA dampers. A numerical simulation produces a comprehensive dataset of structural parameters consisting of the structure’s varying periods, frequency ratios, and eccentricity ratios. The critical responses of structures, including lateral and torsional displacement, lateral and torsional acceleration, and stiff and flexible edge damper forces, are predicted using machine learning (ML) techniques, artificial neural networks, decision trees, support vector machines, ensemble bagged trees, and Gaussian process regression. The model is validated using performance metrics such as mean absolute error and root mean square error, mean absolute percentage error, coefficient of determination, and Shapley Additive explanations values, ensuring that predictions are robust and consistent. The results revealed that regression methods accurately model the nonlinear dynamic behavior of SMA dampers in asymmetric structures, providing exact and computationally efficient predictions of seismic response. This predictive paradigm facilitates optimal damper configuration, minimizing the computational complexity of iterative design methods. The proposed research integrates advanced materials with ML methods to create seismically resilient structural systems.