An improved firefly algorithm for identifying parameters of nonlinear empirical models

Abstract

The nonlinear model is to describe the vortex-induced resonance of long-span bridges under the action of natural wind. The identification accuracy of its parameters directly affects people's understanding of vortex-induced vibration. People have been trying different algorithms to solve this parameter identification problem, but the efficiency and accuracy of algorithms are not satisfied. In this work, a firefly algorithm based on local chaos search and brightness variant (FACLBV) is proposed. The characteristics of chaos make FACLBV search the widely local scope and improve the accuracy of the solution. FACLBV modifies the fixed initial brightness, discards the absorption coefficient of light intensity, links the initial brightness of every firefly with the position of its solution space, and sets the attraction of every firefly as a simple linear function, which reduces the complexity of the algorithm and improves the efficiency. In order to better verify the superiority of FACLBV, the simulation experiment includes three parts: the comparison between FACLBV and other firefly algorithms, the verification of the parameters identified by FACLBV, and the nonparametric test between FACLBV and other intelligent algorithms. Simulation results show that FACLBV is better than other algorithms in performance.