Testing a predictive control with stochastic model in a balls mill grinding circuit

• Autor: Nieto Chaupis, Huber
• Formato: objeto de conferencia
• Fecha de Publicación: 2014
• Institución: Universidad de Ciencias y Humanidades
• Repositorio: UCH-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorio.uch.edu.pe:uch/322
• Enlace del recurso: http://repositorio.uch.edu.pe/handle/uch/322; http://dx.doi.org/10.1109/INDUSCON.2014.7059397; https://ieeexplore.ieee.org/document/7059397/citations#citations
• Nivel de acceso: acceso embargado
• Materia: Ball mills; Mining; Grinding (machining); Model predictive control; Particle size; Predictive control systems; Stochastic control systems; Stochastic systems; Circulants; Control system simulations; Mill-grinding; Quantitative measurement; Stochastic formulation; Stochastic variable; Stochastic models

In this paper, the formulation of a stochastic model and its subsequent incorporation into a predictive control of a balls mill grinding circuit, is presented. The apparition of stochastic variables is a consequence of variables interaction by which is impossible to know a well-defined determinist mathematical methodology. Thus, the perceived dynamics is simulated by emphasizing those possible scenarios of alarm situations in where overloading might collapse the system. Under this perception, the system identification is based on probabilities. Once the model is built, it enters in a based-model predictive control by taking into account the hypothesis that the circulant load and water are under interaction each other. Although the quantitative measurement of this interaction might be speculative, it is not discarded that this interaction might be actually the main source of disturbs on the the particle size evolution. The results have shown positive prospects of the proposed methodology as seen in the control system simulations in where the particle size acquires stability. Furthermore the dramatic reduction of alarms events supports the idea that the MPC is still robust even with stochastic formulations.