Simulation of Nonlinear Dynamic Processes with Volterra Polynomials and Continued Fractions

In this paper, we propose a technique for increasing the accuracy of modeling nonlinear dynamics of input-output systems using finite segments (polynomials) of Volterra integral power series. We highlight a case when the Volterra kernel identification problem is solved. The linear part of the Volterra polynomials of power N is formed as an equivalent continued fraction. Assuming that the corresponding Volterra kernel is known with some error, such an approach allows us to obtain a model with a new linear part that depends on the error. At the same time, prognostic properties of the integral model can be improved by the using the connection coefficient. The problem of indentifying such a parameter was solved using a special extremal problem. We analyzed the performance of the approach for some dynamic systems. Tests examples were used to compare our approach with the known modeling method that employs a series of Volterra polynomials whose transient responses are tuned to different test signal heights.