Development of an artificial neural network for controlling motor speeds of belt weighers and separator in cement production

Abstract

An artificial neural network for controlling the motor speeds of belt weighers and a separator of a cement grinding unit intended for production of a three-component cement of various grades was developed and researched. The purpose of developing a neural network was to solve a problem associated with the large error in the amount of cement at the outlet relative to a given capacity, as well as to increase the speed of the control system and to increase its fault tolerance. As a result of the development, a two-layer unidirectional network with a sigmoidal function of hidden layer neurons activation and a linear function of output layer neurons activation was used. The network has trained on 50 examples over 120 epochs. The development of the neural network was performed in the Matlab environment using the Matlab Neural Network Toolbox. It is possible to use the obtained artificial neural network in conjunction with a SCADA-system using the OPC-server. The developed neural network can be used in control systems for dosing raw materials at cement manufacturing plants utilizing a dry process and a closed cycle.