CONICET | Buscador de Institutos y Recursos Humanos

Application of high-pressure grinding rolls (HPGRs) integrated with ball milling in size reduction of iron ore concentrates filled a prominent position in the minerals processing in the last 30 years. The company Vale S.A, being one of the pioners applying the technology on this circuit configuration [1], adopted the HPGR operation in most of the pelletizing plants from Complexo de Tubarão (Vitória, Brazil), where the HPGR often operates in the regrinding prior to pellet formation in the so-called pellet feed preparation step. Although it is a consolidated operation, several challenges can be stated when it comes to full process integration with HPGR pellet feed pressing, as this operation occupies the boundary between the end of the pellet feed preparation stage and the begining of the pelletizing process. Indeed, the current digital transformation in the mining industry is shifting traditional pro-cess operation towards new approaches able to correlate the multi-scale dynamic modelling and simulation with the main industrial demands, thus trying to make real-time decisions in order to improve the production capabilities. This new modelling approach, which requires a networking integration and real-time information between physical operation and digital models, allows predicting variations within the process besides being used as a robust model predictive control. For pressing iron ore concentrates, recent works by the authors [2; 3] succesfully applied phenomenological modeling to describe HPGR performance under several variations in operating conditions and feed characteristics. Nevertheless, these studies used the so called Modified Torres and Casali model which has only been used offline and in steady-state conditions and presents a lack of maturity on how to achieve a more robust process integration using online information. As such, the present work aims to propose a new HPGR modelling approach coupled with real-time information and applied as an online tool in an industrial-scale iron ore pelletizing plant. The model will be used to map the industrial operation in order to reduce the process disturbances and to propose optimal operational strategies to increase the HPGR through-put, thus reducing the energy consumption and improving the product Blaine Specific Sur-face Area (BSA). The work takes advantage on recent advances in HPGR mathematical modelling [2;3] using a phenomenological approach capable to give rapid responses from several variations in operating conditions, design variables and feed characteristics. A data-driven soft sensor model is proposed to predict the HPGR feed based on an Artificial Neu-ral Network (ANN) and developed from plant database. Application of the modelling ap-proach demonstrated capabilities to map the physical operation and give a realistic repre-sentation of the HPGR performance. The model was also capable to give support for the pellet feed production by given extended real-time information of the process, enabling the improvement of the operational strategies and process stability.

enviar mensaje