Identification Method of Mathematical Model for Linear Dynamic System

Abstract

The paper presents materials and results on the development of a modified identification method for linear dynamic system as well as its application for identification of electric drive control system as a stage of its tuning or synthesis. The method as its conventional form is based on the approximation of the system transition function, that is, its step response in the input action, in accordance with the mathematical apparatus of real interpolation technique. The mathematical model of the system is identified in the form of a transfer function. When using the conventional identification method, there is the problem of determining the optimal form of identified transfer function, sufficient for description of the system. Its optimal form contains only significant coefficients. Usually this problem is solved by iteratively increasing the number of transfer function coefficients, if the result of identification is not accurate enough, which is evaluated by the corresponding of the experimental transition function and the model one. Compared to conventional identification method, proposed modified method is aimed to use the single iteration in determining the optimal form of math model. To achieve this, the conversion of the test input signal and the system of study response to it in time domain is applied, as well as the identification of the transfer function in the normalized form. Due to this conversion, actual time is reduced to dimensionless time. This solution makes it possible to introduce the simple criterion for the significance of the coefficients depending on their values in comparison with other coefficients.