Open AccessREVERSE ENGINEERING OF NONLINEAR SYSTEMS USING ANALYTICAL NETWORKS
Author(s) -
V. A. Rovinskyi,
O. V. Yevchuk
Publication year - 2020
Publication title -
metodi ta priladi kontrolû âkostì
Language(s) - English
Resource type - Journals
eISSN - 2415-3575
pISSN - 1993-9981
DOI - 10.31471/1993-9981-2020-2(45)-109-118
Subject(s) - nonlinear system , computer science , block diagram , artificial neural network , algorithm , control theory (sociology) , artificial intelligence , engineering , physics , quantum mechanics , electrical engineering , control (management)
A method for the synthesis of numerical models based on analytical networks for nonlinear systems ofincreased complexity using their known input and output signals taken synchronously is proposed. The functioning of the proposed analytical networks is based on the use of a modified genetic algorithm and a library of blocks of constant functionality. Genotype structure and mutation algorithms are proposed for describing an analyticalnetwork. In addition, methods for modeling complex non-linear systems using the Volterra, Wiener-Hammerstein series, adaptive filters, non-linear model of an autoregressive moving average with exogenous inputs, neural networks and genetic algorithms are considered, and the main problems that arise when using these models are identified. Apractical example of the possibility of using the analytical network is shown on the example of the resynthesis of asound synthesizer. A typical diagram of such a synthesizer is described. A possible scheme of a re-synthesized system based on an analytical network that is functionally as similar as possible to desired system is considered. The possibility of automatically constructing a numerical model of the reaction of a nonlinear mechanical system to inputdisturbances using known input and output signals recorded synchronously is shown. The main difficulties of theresynthesis of complex systems for sound reproduction are considered - the influence of psychoacoustic phenomena on the perception of synthesis results and the need to ensure high fidelity for obtaining adequate results are shown.The structure of typical blocks of the analytical network is proposed, which should include typical conversions used in digital signal processing, arithmetic and logical operations, correlation and comparison blocks, hysteresis components, and in addition, typical possible standard blocks of a system that undergoes resynthesis.