Systematic calibration of a simulated semi-autogenous/ball-mill grinding circuit

Abstract

With no doubt, modeling and simulation are powerful tools improving the performance of current optimisation and control strategies. At least in the mineral processing field, however, one of the major gaps in literature is that authors frequently skip specific details on the model calibration methodologies, mainly those concerning data acquisition and processing, calibration sequence, cost function formulation (including constraints), optimisation problem solution, parameter uncertainty, etc. As a contribution to that matter, this paper presents a detailed calibration procedure for a grinding circuit model from plant and laboratory data. The methodology integrates both a global search routine to avoid cost function local minima, and a Jack-knife resampling and recalibration technique to estimate the parameter confidence. The simulator will serve in future work for the coupling of separation processes, and thus the development of plant-wide model-based optimisation and control strategies.