Managing the Reproducibility Level of Reported Metal Recoveries

"The content of production reports has its origins in raw measurements taken in harsh production environments. The ensuing measurement/estimation errors always affect reproducibility such that, regardless of the fact that they are balanced through statistical data reconciliation, reported numbers are simply the most probable values of inferred probability distributions. This has important implications for how metal recoveries should be monitored and reported. Indeed, in a context where many process optimisation activities are based on small improvements in recovery, there should be no doubt that inadequate handling of measurement/estimation errors can easily exaggerate or mask actual changes in process performance. However, estimates of the reproducibility of reported metal recoveries are rarely reported to plant management. This situation persists despite the fact that a code compliant metal accounting system must inform management about business risks related to metal accounting. To this extent, this paper focuses on the continuous management of reproducibility level of metal recoveries in compliance with the best practices. Sensitivity analysis of primary and secondary metal accounting models is performed to show the resulting probability distributions for metal recoveries. Some guidelines are given for designing and managing metal accounting systems capable of achieving and sustaining predefined levels for the reproducibility of metal recoveries.INTRODUCTIONEach and every mineral processing plant produces daily, weekly and monthly production reports in which metal balances for payable and penalty elements are presented. While such production reports are usually consulted in offices, underlying raw measurements are taken in harsh production environments where mechanical vibrations, humidity, corrosion and dust, to name a few, disturb measurement procedures. It is thus no surprise that several authors (Francois-Bongarcon & Gy, 2002; Gaylard, Randolph, & Wortley, 2014; Holmes, 2010; Spangenberg & Minnitt, 2014) have reported pervasive issues related to mass/sampling measurements in the context of mineral processing."