Optimizing the parameters of a biodynamic responses to vibration model using Particle Swarm and Genetic Algorithms

Abstract

Various local optimization techniques such as Interior Point Algorithm have been widely used to optimize the parameters of models representing biodynamic responses to vibration. The quality of the obtained solutions depends on the initial guesses. This paper presents a comparison between the performance of Particle Swarm Optimization and Genetic Algorithm in optimizing the parameters of a human body model, where these techniques do not require initial guesses. The model represents the vertical apparent mass and the fore-and-aft cross-axis apparent mass of the seated human body during vertical excitation. With both optimization methods, the model provided close fits to the moduli and phases for both median data and the responses of 12 individual subjects. However, it was noted that using PSO provided a better solution with less mean error than GA and a faster solution in most of the cases.