A knowledge-Based System for a Off-Line Optimization of Ball Milling Circuits

"Off-line optimization studies are often necessary to improve the efficiency of grinding circuits. These studies are primarily based on the mathematical modelling of processing units (e.g., ball mills and hydrocyclones) and computer simulations. To complete an optimization task, process engineers must possess skills and knowledge of different subtasks involved in such efforts, e.g., plant audit, laboratory tests, and data analysis. Such know-how can be in part provided by Knowledge-Based Systems (KBSs). One such system, being developed in C Language Integrated Production System (CLIPS), is presented here, with focus on the estimation of ball mill selection functions for the purpose of both diagnostic and simulation, and its use for the optimization of ball size. Some of the interpretation and diagnostic rules will be presented, as well as grinding kinetics and classification data likely to fire them.IntroductionSince the late eighties there has been a growing trend in applying Knowledge-Based Systems (KBSs) to mineral processing [Meech 1990, Bearman and Milne 1992). Table l lists some KBS (often referred to as expert systems) applications that have been used mainly as tools for improving on-line monitoring and control of grinding circuits. However, there has been fewer attempts to add the symbolic processing powers of KBSs to conventional off-line grinding simulators. Figure 1 shows a relatively standard procedure to perform an off-line optimization study [Napier-Munn and Lynch 1992]. In brief, a typical program includes taking representative samples from a number of streams, running laboratory tests on collected samples, estimating model parameters which describe physical processing units (e.g. ball mills and hydrocyclones) validating model predictions and using models for simulation trials. One inevitably needs to be familiar with concepts of process modelling and simulation as well as various computer programs to be able to complete all steps of the study. Much of this ""expertise"" can be captured by KBSs, given their inherent capabilities for symbolic knowledge representation and manipulation."