Abstract

The copper electrorefining is a rather complex process with multiple operational variables affecting its results. The tendency in recent years in the modernization of tankhouses has been the introduction of mechanization and automation of materials handling systems, with a low level of automatic process control in tankhouses. This study presents a Big Data analysis system for process control, developing operational models based on data analysis by Machine Learning techniques. The system is designed to maintain the copper electrorefining process at optimal and stable levels through operational recommendations allowing improvements productivity. Emerging technologies, such mathematical models derived from large historical databases of Chuquicamata copper refinery, Machine Learning and Industrial Internet of Things (IIoT) have been integrated with the current instrumentation, monitoring and automation existing capacities, without incurring in modifications of the local control systems. The developed system allows to make projections of production quality indicators that are used by a module of Support Decision-Making, optimizing in real time the performance of the operation, recommending actions to be taken by operators. In a first stage of this work, predictive models have been developed for the predictions of impurity concentrations in copper cathodes, physical rejection of cathodes and current efficiency, and validated with historical operation data.