Home – Home – Jornals – Avtometriya 2006 number 2

A new method of identifying parameters for a high-speed adaptive system suppressing torque ripples in a permanent magnet synchronous motor is proposed. Simulation confirmed the high rate of convergence of the sought field parameters to the actual values.

Algorithms for synchronous motor excitation voltage control are proposed. They ensure control of the reactive power and suppress of rotor swings as the terminal voltage and the load torque fluctuate. The algorithms are based on estimating the internal rotor angle and on organizing the forced sliding movement of the system in the desired trajectory. The control algorithms are implemented in a fuzzy controller.

The second-order variable structure system with the non-ideal switching device is analyzed during its motion in the switching mode. A definition of this mode is given. Conditions of its existence and stability, and also conditions of falling the representative point into the switching area are obtained. The property of variable structure system robustness in the switching mode is formulated and proved. The maximum estimate of the system working accuracy compared with the ideal sliding mode is obtained. A method for reducing the auto-oscillation amplitude is developed.

The problem of synthesizing extremum control systems for the class of plants with a nonlinear time-varying dynamic part and a static quality function is discussed. It is proposed to organize two-contour control when the internal subsystem includes a dynamic plant part and the external subsystem contains a static extremum characteristic. The internal contour uses a controller based on the localization method. This makes it possible to preliminary stabilize the processes in the dynamic part of the system, and while synthesizing the external contour to take into account only the equilibrium value of the internal contour. Application of adequate gradient estimation along with the integral control law for the external subsystem allows organization of motion to extremum with the given dynamic quality. It is shown that the properties of such extremum control systems do not depend on the effect of the time-varying plant parameters and the action of uncontrolled external disturbances.

Adaptive stabilization of SISO systems with essentially nonstationary dynamic charac- teristics is discussed. Systems with parametric and (or) additive limited disturbances are considered. The adaptive algorithm design is based on the localization principle. It is proposed to reduce the system order due to decreasing the number of adjustable coefficients of the controller. Stability conditions for these systems are obtained.

An iterative method for solving nonlinear problems of minimizing resource consumption is proposed. It is a generalization of the method for solving linear minimum resource problems [1] to the class of nonlinear systems with the linearly controlled right-hand part separated in state and control.

Implementation of automated design of routes to go around geometry objects by means of artificial intelligence technologies is considered. A mathematical model of the task and a hybrid genetic algorithm for searching for effective trajectories by means of integrating the genetic programming procedure with traditional computing-retrieval procedures are presented. CAD software interacting with the graphics database of systems is designed.

A method for identification of the fact of voltage sags in a fuel cell stack and diagnostics of the causes is considered. The method allows one to state the event as soon as possible after it happens. A procedure of classifying the causes of voltage sage, such as membrane dryout from the cathode side, membrane dryout from the anode side, air starvation, or high CO concentration, is constructed using a certain set of observations of the process variables.

A fast nonparametric automatic classification algorithm for multispectral aerospace image segmentation is proposed. The algorithm is based on forming a cell-like data structure in the space of spectral features and using it in the iterative mean shift procedure for seeking local modes of the density function. Results of testing on model and real aerospace data are presented.

Two different nonlinear methods of multidimensional regression approximation, namely, neural networks and hierarchical approximation of function, are considered. The methods are applied to image filtering problems in the conditions of existing a priori information in the form of a pair of images interpreted as

A new method for reconstructing the internal structure of a 3D body from the known images of its sections, which are given as pictures or tomograms obtained from a computerized tomograph, is presented. The sections of a 3D body are constructed by an interflation operator in a system of three groups of intersecting planes, each plane being not necessarily parallel to the coordinate axes.

Application of invariant moments and information fields to comparing stereo fragments and constructing dense disparity maps is considered. Results of testing different stereo matching algorithms using these features are presented.