Development of control system by Studgard neural network simulator.docVIP

Acta Montanistica Slovaca Ro?ník 8 (2003), ?íslo 4 Development of control system by Studgard neural network simulator Andrea Bulnová and Karol Kostúr 1 1 Vyvoj riadiaceho systému ?tutgardskym simulátorom neurónovej siete The control of coking plant has been soluting by must diverent methods. From this examination its used neural networks and specifically was used Studgard Neural Networks Simulator from creating application into processing this problem. Key words: neural networks, neuron, heat flows, back-propagation networks. Introduction The coke plant is very complicated technological process. Today, the control is provided by manual way. It is non effective control because the operator decide on base more information. The aim of examination was creating application, which calculate heat flows by mathematical model. From calculate heat flows we can use: time series, fuzzy logic, differential equations, neural networks, etc. But optimal solution from this complicated calculating are neural networks. Neural Networks Research of neural networks was motivated by the endeavor to understand how human brain worked. Neurophysiological knowledge allowed us to create simplified mathematical models, which could be applied on practical problems. Neural networks consist of interconnected formal neurons – output of one neuron is input for several other neurons just like axons of biological neurons are connected with dendrits of other neurons via synapsis. Neural network is a universal tool, which is capable to calculate anything a personal computer can. The main advantage and the principal difference between neural network and the von Neumann’s architecture is the ability of the network to learn. Mathematical Model of Neural Networks Basic building block of the biological neural system is a neuron and analogically the most basic element of artifici