Parameter Identification Using an Evolutionary Algorithm and Its Performance under Measurement Errors

The study of engineering problems is usually made in two steps. A mathmatical model is first built to capture the physics of the phenomena, and then the identification of the parameters is made by fitting the model to numerical data. Such curve fitting is frequently done by the method of least squares, with no regard paid to previous knowledge concerning the values of the parameters, nor to the statistical nature of the measurement errors. In this approach, optimisation methods are used to find the parameter set by adjusting them until they provide the best agreement between the measured data and the computed model response [1][2]. Although various calculus-based optimisation methods have been incorporated, these techniques can fail in the actual situation, i.e., when the measured data are noisy and the model equations are inaccurate, since they can cause the objective function to be complex. These techniques are not thus practically useful unless some regularisation technique [3] is incorporated properly.