On the optimal parameter and noise identification on the basis of three parameter probability distributions

Abstract The optimality of the procedure of parameter identification is scrutinized in this paper. It was shown, with the relations between the mathematical theory of function approximation, three parameter probability distributions, which can adjust their shape, and the maximum-likelihood method, that the optimal expression of the distance between measured data and model fitting it can be established by using the three parameter probability distributions on the basis of iteration procedure, where the noise contained in the measured signal is extracted as well. The iterative method for optimal system/model parameter identification is presented and tested by the numerical experimentation. Four types of noise added to the simple single-degree-of-freedom system response are considered: Gauss, Cauchy, Laplace and Uniform. The method performs well for the noise types at relatively high noise content in the signal.