Automated design of self-centering shear walls using machine learning and genetic algorithms

Abstract

Self-centering shear walls (SCSWs) have demonstrated superior resilience compared to conventional shear walls in both numerical simulations and physical experiments. However, analytical design methods for SCSWs and other self-centering structural systems remain underdeveloped. This paper develops and validates a finite element model of SCSWs. Machine learning techniques are employed to evaluate seismic performance. The resulting models achieve high accuracy in predicting stiffness, peak shear capacity, and residual drift of SCSWs. Building on these predictors, an automated design tool is introduced to generate SCSW designs that satisfy resilience requirements while minimizing construction costs.