Development of a Heuristic Methodology for Designing Measurement Networks for Metallurgical Accounting

A numerical study done to test the efficacy of heuristics developed for the design of measurement networks for high precision metallurgical accounting is presented in this paper. The heuristics considered were designed to maximise variance reduction of terminal stream measurements through data reconciliation for observable linear flow systems. A complex metallurgical process flow sheet from the mineral sands industry was used as a case study. Monte Carlo type simulations coupled with a non-linear programming based steady state data reconciliation algorithm were used as tools to investigate the heuristics. The results obtained suggested a close correlation between magnitudes of measured variances and the respective levels of variance reduction obtained through data reconciliation for both internal and terminal streams. It was observed that variance reduction through data reconciliation was maximised for terminal stream measurements in networks characterised by high precision measurement of internal streams and/or in flow sheets characterised by high ?stream to node ratios?. Terminal streams attached to highly interconnected parent nodes were observed to experience higher reductions in variance compared to terminal streams attached to less interconnected parent nodes. The results generally highlight the additional benefits of highly interconnected flow sheets and high precision internal stream measurement regimes on the maximisation of variance reduction of terminal stream measurements through data reconciliation. These findings are significant to metallurgical accounting and to process accounting applications in general in that they provide a conceptual basis for design decision making regarding the construction of measurement networks aimed at maximising variance reduction of targeted streams through data reconciliation.