An improved PSO algorithm for parameter identification of nonlinear dynamic hysteretic models

The nonlinear dynamic hysteretic models used in nonlinear dynamic analysis contain generally lots of model parameters which need to be identified accurately and effectively. The accuracy and effectiveness of identification depend generally on the complexity of model, number of model parameters and proximity of initial values of the parameters. The particle swarm optimization (PSO) algorithm has the random searching ability and has been widely applied to the parameter identification in the nonlinear dynamic hysteretic models. However, the PSO algorithm may get trapped in the local optimum and appear the premature convergence not to obtain the real optimum results. In this paper, an improved PSO algorithm for identifying parameters of nonlinear dynamic hysteretic models has been presented by defining a fitness function for hysteretic model. The improved PSO algorithm can enhance the global searching ability and avoid to appear the premature convergence of the conventional PSO algorithm, and has been applied to identify the parameters of two nonlinear dynamic hysteretic models which are the Leishman-Beddoes (LB) dynamic stall model of rotor blade and the anelastic displacement fields (ADF) model of elastomeric damper which can be used as the lead-lag damper in rotor. The accuracy and effectiveness of the improved PSO algorithm for identifying parameters of the LB model and the ADF model are validated by comparing the identified results with test results. The investigations have indicated that in order to reduce the influence of randomness caused by using the PSO algorithm on the accuracy of identified parameters, it is an effective method to increase the number of repeated identifications.