Azerbaijan National Academy of Science

The first website of Azerbaijan (1995)

HOME PAGE  >>  Institutes and organizations  >>  Laboratory of recognition, identification and methods of optimal solutions

Laboratory of recognition, identification and methods of optimal solutions
Phone (+994 12) 5399231
Fax (+994 12) 5392826 
E-mail lab3.1@isi.az
Chief  Corresp. member of ANAS Aydazadeh Kamil Rajab oglu
Total number of employees  
Basic activity directions  

1.Mathematical modeling;

2.Parametric identification;

3.Development of numerical methods and software development for optimal control problems and intelligent decision-making systems;

4.Development of systems for recognition of authors of printed and handwritten texts, speech, texts in the Azerbaijani language.

Main scientific achievements 

1.A computer speech recognition system has been developed, consisting of various subsystems, each of which has a limited vocabulary. The division into subsystems increases the effectiveness and quality of recognition. Depending on the requirements for recognition, the system provides for the choice of various options for the degree of accuracy and effectiveness of recognition;

2.Optimal control of transient processes in oil and gas pipelines described by partial differential equations of hyperbolic type has been investigated. The regularities of the influence of the choice of technological constraints and the class of control actions on the obtained optimal modes of transient processes have been revealed;

3.For optimal control problems with lumped pulsed sources in systems with distributed parameters, necessary optimality conditions have been obtained for locations of source concentration and moments of action. Formulas have been obtained that allow using finite-dimensional first-order optimization methods for the numerical solution of the problem;

4.Using the example of control of thermal processes with feedback from point sources, an optimal synthesis of feedback parameters, an approach to optimizing the number and placement of state control devices have been proposed. Constructive necessary optimality conditions have been obtained that are used to construct a numerical scheme for solving the problem. Computer experiments have been conducted using the developed software;

5.A class of smooth membership functions that uses the idea of spline functions for fuzzy numbers and a generalizing class of many well-known membership functions has been proposed. The proposed membership functions have been used in the computer system of recognition of printed and written texts in the Azerbaijani language;

6.An approach to the numerical solution of the coefficient inverse problem for linear equations of parabolic and hyperbolic type with given nonlocal conditions has been proposed and substantiated. The approach is based on the method of straight lines and on a specially constructed solution of a boundary value problem. It includes the solution of auxiliary issues for a system of ordinary differential equations. The results of the study can be used for mathematical modeling of the processes of subsurface hydrodynamics, environmental and other processes, in cases where it is impossible to conduct instantaneous measurements of the state of the process or at different points of the object;

7.For parabolic and hyperbolic equations with nonlocal initial and boundary conditions, a numerical solution to the class of coefficient inverse problems for coefficients depending on spatial or temporal variables has been investigated. Two approaches have been proposed and studied, based on the direct method and the grid method. Unlike the known approaches, when using both methods, there is no need to build an iterative procedure. The first approach reduces the solution of the inverse problem to solving the Cauchy problem for ordinary linear differential equations, and the second approach to solving a system of linear algebraic equations;

8.An approach has been proposed for the formulation and numerical solution of the problem of synthesizing the optimal control of observation objects described by a system of nonlinear ordinary differential equations. The problem has been solved for different types of feedback in different classes of zonal control functions and different types of observations depending on the results of observations in the case of inaccurate presentation of information about the parameters of the object. The necessary optimality conditions have been obtained that contain the formula for the gradient of the objective function. Numerical experiments have been carried out on the example of solving model problems;

9.A hybrid approach based on a hidden Markov model and a neuro-fuzzy approach has been proposed for sentiment text analysis. A new weight function “Pruned ICF” has been proposed for calculating the characteristic vector in the process of meaning recognition. It has been shown that the weight function “Pruned ICF” when conducting a subjective analysis gives better results than other weight functions (TFIDF and IDF);

10.The problem of identifying the coefficient of hydraulic resistance of the linear part of the pipeline during transportation of hydrocarbon raw materials, reduced to the coefficient inverse problem for hyperbolic equations has been considered. It is assumed that there are results of measuring the modes of transportation of raw materials at the ends of the linear part and/or at intermediate points of the pipeline. The proposed method of identifying the resistance coefficient is based on the method of solving the problem of control synthesis in the class of zonal control actions previously considered by the authors. The results of the numerical solution of model problems have been obtained;

11.On the basis of parallel computing technologies, algorithms and software have been developed for the control system of the optimization application program package;

12.On the basis of Markov and neuro-fuzzy models, algorithms and software have been created for computer recognition of Azerbaijani texts;

13.A method has been proposed for the numerical solution of the problem of localizing leaks in the case of unsteady fluid flow in oil pipelines of a complex ring structure. Numerical experiments on model problems have been conducted; 

14.Algorithm and software for recognition of authors of texts in the Azerbaijani language have been developed;

15.The problem of optimization of sources in the right-hand side of nonlocal boundary conditions for dynamical systems with both lumped and distributed parameters has been solved. The necessary optimality conditions for the source parameters for the construction of numerical algorithms have been obtained;

16.Methods, algorithms and software have been developed for computer recognition and evaluation of the results of mass surveys and test examinations. The software has been implemented at the State Examination Center to evaluate the results of the final examinations conducted in the 9th and 11th grades of secondary schools of the republic.