Improved accuracy and computational stability of the identification method of nonlinear systems based on Volterra models
65122, Odessa, Ukraine, Str. Architectorskaya, 12, Apt. 471 pp. (accepted)
A new method of constructing approximate model of Volterra nonlinear dynamical system (VAT) in the time domain using poliimpulsnyh and multistage test signals, which is different from the known use of a regularized least squares method and the choice of optimal value step amplitude test signals that can improve the accuracy in 2 4 times the computational stability and identification procedures as well as the application of wavelet filtering for smoothing estimates Volterra kernels, which increases the accuracy of 2-3.4 times and ensures the smoothness of the identification results.
It proposed and theoretically substantiated formalism, which is the universal expression for the experimental determination of multidimensional transfer functions (n-dimensional integrals of Volterra kernels) as a linear combination of the responses identified VAT multistage test actions to help simplify implementation of algorithms and software protsedry identification.