RESUMO

We consider the algorithm for identification of linear stationary systems by results of measurement of transient process. This algorithm can be treated as nonlinear analog of the Prony method. Similar to the Prony method the suggested algorithm makes it possible to realize decomposition of the initial problem of identification, however, it eliminates from consideration the intermediate linear model with discrete time, which is used in the Prony method. Generalization of the algorithm is adduced in the case when some roots of the characteristic polynomial of the identified system are known. By the before published examples we conduct comparison of accuracy of estimates obtained by the suggested method and other known methods for solving this problem.