350 rub
Journal Neurocomputers №7 for 2012 г.
Article in number:
The particularity development of the automatisation control system for the physical-mechanical properties based on a neural network on the production Al-profiles
Authors:
O.B. Gromov, S.Y. Dudnikov, A.D. Zhargalova, P.I. Mikheev
Abstract:
We described practice of construction of automated system on the base of neural network (NN) for control complex distributed system. As a control object we consider production of the profiled aluminum products (Al-profile) with key process as extrusion. One of the features of this control system is a set of separate, discrete-continuous processes with complex physical-chemical and mechanical transformations which imped the choice of best numerical value of the control parameters for required properties of the product. Approach for the correctly choice of NN as an element of the automation system and its application in calculation of the function properties of Al-profile described in detail. The main task is process input and transmits output on other elements of the system which is not directly related of NN. This is the comparison of the most common NN. To build an automation system of production Al-profile architecture of the NN was chosen. This is hybrid network that combines the organization of connections between neurons in the NN back propagation with the values of the output layer, expressed with the rules of fuzzy logic. It Is possible to interpret the result (vector of the physical-mechanical properties of the Al-profile) and compare it with parameters which require the customer. The described method of setting the NN consists in consecutive application of the following operations: selection of the optimal number of hidden layers of NN, calculation of the weight and minimization of a quadratic function of the error of NN training. To successfully solve the tasks in addition was considered requirements of the hardware architecture of the automation system using the NN. The proposed design features of the automated control system of physical and mechanical properties of Al-profile on the basis of the NN based on modern control theory allow creating effective, cost-based and reliable system operation and control of distributed productions.
Pages: 16-25
References
  1. Щерба В. Н., Райтбарг Л. Х. Технология прессования металлов. М.: Металлургия. 1995.
  2. Kohonen,T.,1984. Self-organization and associative memory. Series in Information Sciences. V. 8. Berlin: Springer verlag.
  3. Hopfield, J. J., Neural Networks and physical systems with emergent collective computational abilities // Proc. Nat. Sci. USA. 1982. V. 79. P. 2554-2558.
  4. Grossberg, S., Classical and instrumental learning by neural networks. Progress in theoretical biology, V. 3. New York: Academic Press. 1974. P. 51-141.
  5. Винер Н. Кибернетика, или управление и связь в животном и машине: Пер. с англ. Изд. 2-е. М.: Сов. радио. 1968.
  6. Колмогоров А. Н. О представлении непрерывных функций нескольких переменных суперпозициями непрерывных функций меньшего числа переменных. Докл. АН СССР. 1956. Т. 108. № 2. С. 179-182.
  7. Леоненков А. В. Нечеткое моделирование в среде MATLAB и fuzzyTECH. СПб.: БХВ - Петербург. 2003.
  8. Дьяконов В. П., Круглов В. А. Математические пакеты расширения MATLAB. Специальныйсправочник. СПб.: Питер. 2001.
  9. Rumelhart, D. E., Zipser, D., 1986. Feature Discovery by Competitive Learning. In D.E. Rumelhart & J.L. McClelland (Eds.), Parallel distributed processing (V. 1. P. 151-193). Cambridge. MA: MIT Press.