ML-Based Identification of Nonlinear Dynamic Systems: Comparison of Black-Box and Grey-Box Approaches*

Traditional methods of system identification are based on solving linear regression problems. However, the application of this approach to nonlinear systems requires solving a complex procedure of nontrivial parameterization, which hinders their application. This article is devoted to the study of machine learning methods in the problem of identifying nonlinear systems. A mechatronic stand simulating the dynamics of a helicopter is used as an object for research. We compare the effectiveness of various machine learning models for solving this problem: in the first case, the parameters of a known mathematical model are identified; in the second case, the system is considered as a "Black-Box". A comparative analysis of the results is carried out.