Ore Processing Profitability Optimization Using A Multiple Competing Model Computer Control "Smart" System

Optimizing ore processing requires a thorough understanding of not only the process itself but the economic constraints which influence business decisions. No single modeling technique is able to model all the different processes running in today's mines. The solution then is to use a number of different model­ing techniques, each competing with the others to best correlate the data. Among these modeling techniques are: Neural Networks, Genetic Algorithms, Fuzzy Logic, Crisp Logic, Statistics, First Principal Models, and System Identification models. Once a model has captured the behavior of the process, a numerical optimization determines the optimal operating point and set point parameters. A single step is then taken in the direc­tion of the optimal location by changing all the process parame­ters to the setting recommended by the optimization system. As the process begins to re-stabilize, the models are updated, the new optimal determined and another step is taken in the direction of the optimum. Through the use of multiple competing models integrated into a numerical optimization the process can achieve the optimal running condition even if the ambient conditions fluctuate dynamically.