On the optimal parameters of a PSO-based algorithm for simulation of Endurance Time Excitation Functions

TL;DR

This paper introduces a particle swarm optimizer to generate more accurate endurance time excitation functions, improving the input motions used in endurance time methods for structural analysis.

Contribution

It presents a novel application of particle swarm optimization for simulating endurance time excitations, with optimized parameters determined through a parametric study.

Findings

PSO outperforms trust-region reflective method in accuracy

PSO achieves better excitation functions than imperialist competitive algorithm

Proposed method enhances the reliability of endurance time analysis

Abstract

This paper presents a particle swarm optimizer for production of endurance time excitation functions. These excitations are intensifying acceleration time histories that are used as input motions in endurance time method. The accuracy of the endurance time methods heavily depends on the accuracy of endurance time excitations. Unconstrained nonlinear optimization is employed to simulate these excitations. Particle swarm optimization method as an evolutionary algorithm is examined in this paper to achieve a more accurate endurance time excitation function, where optimal parameters of the particle swarm optimization are first determined using a parametric study on the involved variables. The proposed method is verified and compared with the trust-region reflective method as a classical optimization method and imperialist competitive algorithm as a recently developed evolutionary method. Results show that the proposed method leads to more accurate endurance time excitations.