Seismic Design Optimization of Steel Moment Frames by Neural Networks

Abstract

This paper aims to optimize steel moment frames within a performance-based design framework by implementing a neural network-integrated metaheuristic algorithm. Performance-based optimization of steel structures poses a complex and nonlinear optimization problem, characterized by numerous local optimum solutions. Consequently, an effective search technique is crucial for addressing this class of complex structural optimization problems. Additionally, to reduce the substantial computational load of this process, it is essential to approximate the necessary nonlinear structural responses during the optimization. This paper employs the center of mass optimization (CMO) algorithm as the search engine. Furthermore, a cascade-forward back-propagation (CFBP) neural network model is trained to evaluate the nonlinear responses of steel moment frames within the seismic optimization framework. Two design examples of multistory steel moment frames are presented to demonstrate the effectiveness of the proposed CMO-CFBP algorithm through several independent optimization runs. The results of this paper highlight the superior performance of the proposed technique in comparison to other algorithms in the literature while maintaining a reasonable computational cost.