卷 61, 编号 1 (2018)

We propose a method for the analysis of groups of unidirectional waves on the surface of deep water, which is based on spectral data of the scattering problem in the approximation of a nonlinear Schrödinger equation. The main attention is paid to the robustness and accuracy of the numerically obtained spectral data. Various methods of choosing the wave number of the carrier wave, which rely on the analysis of the local Fourier transform and the zero-crossing wave analysis, are considered. The most robust wave numbers have been chosen on the basis of two model examples. A method for improving the accuracy of the soliton amplitude prediction, which uses the “feedback” in solving the associated scattering problem, is proposed. In the wave steepness range from 0.15 to 0.30, the accuracy of determining the amplitude of the soliton group by this technique lies in a range of 10%.

We present the results of experimental studies of propagation of the low-frequency pulsed signals in the deep-water region of the Indian Ocean under conditions of a waveguide open from the bottom side when the signal from a near-surface source, which is received by the near-bottom receiver, is a sequence of reflections from the bottom and the ocean surface. The spatiotemporal diagram showing the amplitudes of the received pulses as functions of the distance and the relative time of their arrival is used for the data processing. It is shown that using a simple model of the stratified waveguide with a constant bottom and geometrical-optics approximation for calculating the tone-signal field, one can achieve good agreement between the experimental and theoretical results. In particular, this allows one to determine the effective waveguide depth for the frequency range 5–40 Hz and estimate the reflecting-boundary characteristics. The obtained results can also be used for predicting the parameters of the low-frequency signal propagating in the waveguides of this kind.

We use the auxiliary-source method to solve the problem of scattering of electromagnetic radiation by a thin cylinder made of a perfect conductor and a magnetodielectric. Some results of numerical calculations, which characterize the influence of the dielectric permittivity and length of the magnetodielectric part of the cylinder on its scattering cross section, are presented.

We discuss the possibility of recording and the efficiency of processing of information in freedynamics neural networks with context-dependent parameters when the data are presented in signal form. On the example of numerical stochastic experiments, we demonstrate the possibility of processing of information by free-dynamics neural networks. Closeness of the considered networks to structures of natural biological nature is discussed. Using particular examples, we study the stability of free-dynamics networks with context-dependent parameters in solving problems related to recording of signals in the presence of noise.