Study on remodeling approach quality of dynamic objects with variable structure

Abstract

The paper introduces the procedure of constructing new models based on already existing ones. This process is called Mathematical Remodeling. The inertial torque transformer (ITT) is considered as an object which model is under consideration. The functioning process of ITT varies depending on a certain conditions. The initial model of ITT, built on the basis of physical laws underlaying its functioning, is a composite system of nonlinear non-stationary differential equations. Neurostructural models were selected as a class of models, to which the initial model was transformed. It was shown in many studies, that this class of models has a high approximation capabilities. The construction of neurostructural models was carried out on the data obtained by a numerical approximate solution of the original composite systems of differential equations showing ITT workflow. Due to the fact, that the approximate solution could be obtained by different methods, with different integration steps and different sets of initial parameters values, the paper conducted a study on the quality of obtained models, their sustainability (in the sense of changing model parameters with changing the input data).