Machine learning-aided multi-objective optimization of structures with hybrid braces – Framework and case study

Abstract

Hybrid brace is an emerging class of seismic resilient members capable of providing satisfactory energy dissipation with extra promising characteristics such as self-centering capability. Due to the multi-parameter nature of such braces, the design is often challenging. To address this issue, a general optimization framework for the optimal design of multi-parameter hybrid braced structures is proposed. The main steps include: selecting optimization parameters, establishing sample database, formulating machine learning strategy, conducting automatic design and cost calculation, and optimizing the multi-objective problem through the genetic algorithm. The framework is illustrated via a case study considering a viscoelastic self-centering braced steel frame. Various seismic responses, e.g., inter-story drift and floor acceleration, of the considered frame are mitigated simultaneously with no cost increase after the optimization. It is envisaged that the application of the proposed optimization framework can be readily extended to other complex systems which are difficult to design via traditional methods.