Search academic texts

Information × Registration Number 0219U003737, 0116U000893 , R & D reports Title Development of the theory of neural networks of adaptive resonance and associative memory for the creation of intelligent systems popup.stage_title Head Dmitrienko Valery, Registration Date 02-07-2019 Organization The Kharkov state polytechnical university popup.description2 The object of the study is the processes of determining the transformation functions that associate the variables of linear and nonlinear models of geometric control theory and the processes of development of neural networks (HM) associative memory with a new architecture that combines the possibilities of ordinary discrete HM bidirectional associative Memory (DAP) and HM, allowing to receive several associations on incoming information and to develop methods for solving problems of optimization of manufacturing processes of machining of precision parts and assemblies in machine building for the construction. intelligent decision support systems to optimize these processes. The subject of research is the methods for determining the transformation functions that associate the variables of linear and nonlinear models of geometric control theory and the neuronal networks (HM) of associative memory for modeling and synthesis of structures of technological processes of metal cutting by cutting at the stage of determining the structure of technological processes. The purpose of the work is to develop theoretical foundations of the methods for determining the transformation functions that bind variables of linear and nonlinear models of geometric control theory and the development of associative memory neural networks with a new architecture that combines the possibilities of ordinary discrete HM two-directional associative memory and NM, which allow to receive several associations on the input information and improve the accuracy of modeling and calculations of the initial parameters of machining processes by cutting and optimal control parameters t ecological operations and reduce the cost of modeling and manufacturing of complex parts. Research methods are geometric control theory, the theory of neural networks, the theory of decision support systems, the theory of blade processing of metals, the theory of data structures, the theory of problem solving in multicaterial formulation. The significance of the results obtained. An expanded sphere of application of the geometric control theory (GTU) to nonlinear objects, which are described by systems of ordinary differential equations, in which the right-hand sides of almost all differential equations contain more than one or two monomials, which is typical for known uses of the GTU and is associated with difficulty in the search for functions transformations that bind variables of linear and nonlinear models. To address these difficulties, it was initially proposed, with the help of a known method, to narrow the search space of the transformation functions and then use it to search for neural networks that are analogous to neural networks of the group method of taking into account arguments. The results correspond to the global level, because the scope of application of the method of control theory is expanded. For the first time, a new data structure based on neural networks of associative memory has been developed, which, unlike existing ones, which can only remember couple of associative images, enables the input image to match one or more associations and associations of trees and solve problems with several solutions, which allows to increase its flexibility and provides the opportunity to store in its memory a plurality of technological processes of mechanical processing of specific parts. Further development methods of synthesis of structures of technological processes of metal cutting by cutting at the stage of determining the structure of technological processes, which, unlike the known ones, take into account the possibility of interchange of metal cutting operations, and use production rules containing existing practical experience and expert knowledge of technologists, and at the definition stage Optimum control parameters for machining operations by cutting differ in the solution of the problem in the multi-criteria setting, which allowed taking into account the cont enthusiastic target functions, namely cost of operations, energy costs and performance of operations, and taking into account the effect of current depreciation accumulated by the tool on the back surface, which allowed to improve the accuracy of calculations of the initial parameters of machining processes by cutting and optimal operating parameters of control of technological operations and reduce the cost of manufacturing complex components. Product Description popup.authors Бречко Вероніка Олександрівн Гейко Геннадій Вікторович Главчев Дмитро Максимович Заковоротний Олександр Юрійович Заполовський Микола Йосипович Ліпчанський Максим Валентинович Леонов Сергій Юрійович Лимаренко Володимир Володимирович Мезенцев Микола Вікторович Носков Валентин Іванович Паржин Юрій Володимирович Серков Олександр Анатолієвич Скорняков Юрій Сергійович Угрін Дмитро Ілліч Хавіна Інна Петрівна Харченко Артем popup.nrat_date 2020-04-02 Close