Title: Data driven models for cement grinding circuit

Authors: Venkatesh Sivanandam; Ramkumar Kannan; Seshadhri Srinivasan; Guruprasath Muralidharan

Addresses: School of Electrical and Electronics Engineering, SASTRA University, Thanjavur, India ' School of Electrical and Electronics Engineering, SASTRA University, Thanjavur, India ' Kalasalingam University, Krishnankoil, India ' FLSmidth Pvt. Ltd., Chennai, India

Abstract: Cement grinding in ball-mill consumes majority of the energy in cement industry. Current models in literature capturing the material flow are not suitable for designing predictive controllers for energy savings. This investigation proposes two data-driven modelling approaches for cement grinding process that relate material flow and energy. Data obtained from a cement grinding circuit in a cement industry located near Chennai, India, is used to build the models. The first modelling approach uses system identification techniques to identify a linear ARX model. A number of candidate models are developed using the collected data and the model that shows reasonable accuracy and less computational intensity is selected to model the cement mill process. The second approach uses the feed-forward neural network (FFNN) to develop a non-parametric model. As the number of neurons in the hidden layer influences the accuracy of the FFNN, different candidate models with various network parameters are tested. The FFNN model that gives better accuracy is selected as the most suitable model. Finally, validation test on the selected parametric and non-parametric models is used to infer the suitability of the models to capture the dynamics of the cement grinding mill process. Our results indicate that the FFNN-based non-parametric model shows better accuracy and computation simplicity than the linear ARX model. Energy saving predictive controllers for cement industries can be designed using the proposed model as it directly maps energy usage with cement production.

Keywords: cement ball mill grinding process; system identification; ARX model; neural network model.

DOI: 10.1504/IJAIP.2017.084994

International Journal of Advanced Intelligence Paradigms, 2017 Vol.9 No.4, pp.414 - 435

Received: 28 Apr 2015
Accepted: 26 Aug 2015
Published online: 10 Jul 2017 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article