Inventory Predictive Control for Batch Process Production Regulation

"A buffer’s ability to dampen production variations is directly related to its size. Small buffers cannot handle large variations in production, resulting in production losses when their inventory is depleted or saturated. This paper presents a predictive control algorithm used to regulate the inventory in a buffer of limited size, subject to large production variations caused by an upstream batch process. The algorithm was designed to manage the acid flow rates in an acid regeneration process, servicing batch leach reactors at Rio Tinto Fer et Titane, in Sorel-Tracy, Québec. The objective was to increase production by reducing equipment downtime caused by imbalances in the acid tanks servicing the leach reactors. The controller attempts to maintain acceptable acid tank volumes by optimizing the leach reactors acid/slag ratio and reaction time as a function of anticipated acid needs, all while maintaining product quality.INTRODUCTION Some beneficiation processes in extractive metallurgy require processing bulk material in batches. Typical examples of metallurgical beneficiation batch processes are smelting, leaching, electrowinning, elution and carbon-in-pulp. These processes are often situated midstream in an ore beneficiation flow sheet and either receive or transfer ore to continuous processes. Batch processes, by their very nature, introduce large variability in the flow sheet, because they process large finite quantities of material at nearly periodic time intervals. In order to dampen variations in production, various buffers–tanks and stockpiles–are utilized to accumulate ore upstream and downstream of batch processes. A buffer’s ability to dampen production variations is directly related to its size (Phillis & Gumede, 2011). Small buffers cannot handle large production variations, resulting in equipment downtime when their inventory is depleted or saturated (Everett, 2010, Berton et al., 2013). Consequently, the batch equipment production rates must be adequately managed to minimize production losses, attributed to depleted or saturated inventories, while maintaining product quality. Algorithms that anticipate or predict future production variations are often utilized to regulate buffer inventories, subject to large flow rate variations (Salvo et al., 2002; Bonvin, 2013). This paper presents a predictive control algorithm used to maintain an acceptable volume in a spent acid tank servicing several batch leach reactors. The algorithm is implemented at the acid leach plant of Rio Tinto Fer et Titane, in Sorel-Tracy, Québec."