Abstract
This paper describes a method of identification of a nonlinear system composed of a cascade connection of a zero-memory nonlinear element and a linear element. The characteristics of the zero-memory nonlinear element is represented by the stepwise approximation and the dynamics of the linear element is represented by the impulse response.
In this paper, the squared difference of the outputs of the unknown system and of its mathematical model is used as a performance index, and the estimated parameters are adjusted so as to minimize the performance index by an improved gradient method.
The convergence of the estimated parameters is studied theoretically and a method for choosing coefficient quick convergence and high accuracy of the estimated parameters is proposed.
Some examples are simulated by a digital computer and the results support the theory described in this paper.
