For the mathematical description of checing objects with unknown structure it is offered to use nonlinear nonparametric dynamic models on a basis integrodegree Volterra series. Thus set of multivariate weight functions -Volterra kernels completely characterize both nonlinear and dynamic properties, and hence, technical condition of checing objects. For definition of Volterra kernels the new method of identification with use of the determined trial signals, based on allocation partial components with the help of a linear combination of responses of checing objects and applications Wavelet - transformations for increase of a noise stability of a method is offered. With the help of computer modelling on test nonlinear checing objects in MATLAB environment the experimental research of a method of identification confirming his high efficiency is lead. The received dependences show reduction of an error of identification at use of amplitudes of the trial signals offered at definition of Volterra kernels of the second order. It is shown, that use Wavelet-transformation for smoothing oscillations in estimations of Volterra kernels allows to reduce an error of identification in 1.2 - 2 times. The received schedules of dependences of an error of identification from a step of integration at definition of Volterra kernels testify to existence of optimum values.