Intelligent Control System for Process Parameters Based on a Neural Network

Abstract

A neural network for control systems of raw material dosing in dry cement production process with the use of a closed cycle has been developed. The artificial neural network is intended to control motor speed of belt-conveyor weighers and the separator of the cement milling unit intended for the production of three-component cement of various grades. The developed neural network is designed to solve the problem with an appreciable error in the output quantity of cement in comparison with the set capacity; in addition, there is a task of increasing the control system performance and increasing its fault tolerance. The control system uses a two-layer unidirectional network with a sigmoidal function of activating the neurons of the hidden layer and a linear function of activating the neurons of the output layer. The network was trained on 50 examples within 120 epochs. The neural network is developed in the Matlab environment using the Matlab Neural Network Toolbox application. The process control is carried out by SCADA-system through OPC-server intended to provide communication between the neural network and a controlled object.