Likhoedenko

The article studies a neural network approach to obtain the statistical characteristics of the input vector signal and noise implementations at ill-conditioned matrices of correlation moments to solve the problems to select and reduce the vector dimensions of informative features at detection and recognition of signals and noise on the basis of regression methods. It considers a problem concerning the matrix condition of the correlation moment of informative signal features and a possibility to apply the neural networks in theoretical studies. Neural network algorithms are used to obtain the necessary statistical data for ill-conditioned matrices of correlation moments of informative features. It is shown that the weights obtained at training the neural network with no zero weights are the regression coefficients of the initial parameters of informative features input.

The article presents a method to form the non-periodic selection functions in the short-range coherent pulse Doppler radar information systems with random signal phase-shift keying (manipulation) in transmitting and heterodyne channels. It studies statistical characteristics of Doppler signals received when reflecting from the point object that is located inside the first and arbitrary remote strobes at synchronized random carrier phase-shift keying in transmitting and heterodyne channels. The article shows an expansion of the relative bandwidth of energy spectrum of a Doppler signal received after reflecting from the point object from an arbitrary remote strobe, except for the first one. To form a non-periodic selection function is possible via the recognition method for the Doppler signal implementations using the relative bandwidth of energy spectrum and via the application of a regression tract to process intervals between zeroes of implementations.

The authors consider the main method of creating thermally isolated structures of sensors of physical quantities made according to MEMS technologies, that is electrostatic anodic bonding. It was shown that the main reason for a small percentage of production of suitable thermally isolated structures manufactured by a method of electrostatic anodic bonding is unevenness of physical and chemical properties across the whole surface of semi-conductor and corresponding glass plates. The authors propose improvement of the anodic bonding method; designs for adaptations were developed for implementing this improvement. The proposed designs were manufactured and probed. Results confirming the possibility of increase in percentage of production of suitable thermally isolated structures of sensors of physical quantities with the use of the method of electrostatic anodic bonding were obtained.

 Approaches to the problem of temperature compensation of measuring errors of the channel in pressure sensor for spacecraft operating under severe conditions were considered. It was shown that, for correction and, consequently, for temperature compensation of the pressure sensor output two different methods could be used. The first one is ancillary temperature measurements, the second one is ancillary measurements of reference signals. Technical solutions for creating  thermally isolated cells and electronic circuit for temperature compensation of sensors error were developed and tested. Methods of temperature compensation of zero drift and sensitivity of piezoresistive semiconductor pressure sensors were defined. Values of the temperature error were obtained; they allowed us to put manufactured measurement channel according to GOST 22250-93 into class 0.25 accuracy. 

The authors consider approaches to creation of security features for information and telecommunication equipment in an unmanned space vehicle (SV) with electromagnetic launch. It is shown that the main element of the microelectromechanical sensor element - the key - is a beam with a loaded core. The authors provide theoretical calculations and a table of geometric dimensions of sensor elements for two thresholds achieved with electromagnetic starting of SV - 2500 and 3000 acceleration of gravity of free fall (g). The authors consider MEMS key manufacturing technology and propose several ways to make them. The results of experimental studies under static and dynamic g forces are also presented in the article.

The authors propose an approach to construction of an adaptive system of automatic adjustment of excitement of synchronous electric generators. The authors consider optimization of transient processes characteristics in power supply systems, on the basis of indirect adaptive management with an identifier and neural network technologies. It was shown that accuracy of approximation of transient processes in a power supply system by the neural network identifier allowed to reproduce time length of the transient process to within the tenth fractions of a second.

The article covers realization of a device for synchronous generator-field control. It is shown that the proposed device is meant for transient response data optimization in electric generators due to adaptation of the field regulator parameters to the current state of power system. The device is able to adapt the parameters of the operational unit (regulator) to the parameters of electrical power network by means of continuous teaching during the operation.

The authors consider realization of an electric generator control unit for weakening adverse effects of overloads or transient processes. The authors review the work of the regulation characteristic calculator on the basis of using an ultra-fast annealing algorithm. It is shown that the device is capable of solving problems of decreasing the energy of transient processes (which appear during disturbances in power grids) due to the parameter optimization of the field regulator of the synchronous generator according to the variation of primary voltage at the installation stage.