Parameter Estimation of a Single Wheel Station using Hybrid Differential Evolution

Abstract

This paper investigates an application of a ‘discrete variable’ hybrid differential evolution (dvHDE) method to parameter estimation of a single wheel station. The parameters of the single wheel station represent a quarter of the suspension of a medium sized family car. The estimation method developed incorporated the dvHDE and use of a Kalman filter (KF). The KF provides estimates of the ‘unmeasured’ states of the system being studied. The dvHDE, which works as a function optimizer, provides a ‘best fit’ set of model parameters. The performance of the dvHDE method was examined and compared against the standard gradient-based (GB) method, downhill simplex (DS) method and original differential evolution (DE) method on simulated and experimentally obtained data. The normalized mean squared errors (MSEs) of the system outputs are considered as the fitting criterion in the optimization process. The identified model parameters gave an MSE of below 3.5 per cent. The dvHDE method performed better over the GB, DS and DE methods and has been shown to improve the convergence rate by approximately 19 per cent over the original DE method, without sacrificing ability to find the global minimum point.