A Comparison Of Fuzzy And Neural Network Modeling Of Mineral Processing Equipment

Researchers at the University of Alabama are investigating artificial intelligence based computer models of mineral processing systems. Two approaches have been found to be particularly effective in developing data-driven computer models. The first approach employs neural networks which are computational paradigms based on the mechanics of the human brain. They are able to extract the basic relationships between input and output variables from large data sets. The second approach combines fuzzy mathematics with genetic algorithms to produce an adaptive modeling system. Fuzzy mathematics provide computers with the capability of manipulating abstract concepts like those which humans use so effectively in decision making. Genetic algorithms are search algorithms that mimic evolutionary processes apparent in nature. Together, these two techniques can be used to produce an effective modeling system. In this paper, the two approaches to modeling are compared and contrasted with respect to accuracy and computational time while used to model three mineral processing systems: (1) a grinding circuit, (2) a hydrocyclone separator, and (3) a flotation circuit.