Prediction of immeasurable variables using artificial neural networks

Abstract

One of the significant problems in process industry is real-time determination of laboratory values. Laboratory analysis is usually done periodically and can take significant amount time. The information about values between two laboratory analyses doesn't exist. Very often in that time big change happens and it is detected too late. Because of that it becomes necessary to undertake radical corrective actions to steer the process to desired performance. Our goal was to develop prediction system which would be able to calculate and predict laboratory values in the real time. In this paper system for real time prediction of Free Calcium Oxide contained in clinker is presented. This is one of the parameters which determine the quality of clinker, which is used in cement production. For prediction, artificial neural networks were used and findings are presented in this paper. The same approach can be used in development of similar prediction systems for real time prediction of laboratory measurements in agricultural industry, process industry, chemical industry ….

Keywords

• neural network
• model development
• free calcium oxide
• opc client