Structural seismic design using hybrid machine learning and multi-objectives Particle swarm optimization algorithm: Case of Special moment frames in a high seismic zone

Abstract

The paper introduces a new methodology based on artificial intelligence (AI) techniques, including machine learning (ML) and multi-objective optimization (MOO), to design special moment frames (SMFs). The methodology aims to identify the optimal design concerning processing time, initial cost, fragility assessment, damage rate, and performance level. It provides a rapid alternative to create more resilient structures and simplify the design process. The methodology consists of two main components: a precise ML model that replaces the nonlinear time history analysis (NLTHA) and MOO algorithm to identify the Pareto frontier. This paper employs a hybrid ML model to predict seismic response regarding maximum Inter-story drift ratio (MIDR) to improve seismic prediction and enhance the generalization of the model. The proposed methodology will be applied to three SMFs with different heights (4-, 8-, and 12-stories), and the results will be compared to traditional force-based design (FBD). The results indicate that the proposed AI-based methodology achieved the desired performance, lower initial cost, and lower damage rate than the FBD designs. The AI-based methodology offers a fast alternative to performance-based design, resulting in structures that are more efficient in terms of performance and more cost-effective with reduced damage in high seismic zones.