Optimal synthesis of an industrial fluorspar beneficiation plant using a jumping gene adaptation of genetic algorithm

Abstract

A modeling, simulation and optimization study of a complex industrial fluorspar beneficiation plant (Kadipani, Gujarat, India) with fourteen/ten flotation banks is carried out to find an optimum/near-optimum circuit for the given quality of feed ore. The phenomenological flotation model developed by Mehrotra and Kapur (1974) is used for circuit optimization. Two different optimization problems are formulated and solved. The first optimization problem involves a single objective function, which uses available plant data to estimate the feed-characterizing parameters, i.e., the flotation rate constants, by minimizing the weighted sum-of-square errors between the computed and plant values. The second optimization problem involves two objective functions to obtain several simplified circuits. The number of non-linking streams and the overall recovery of the concentrate, i.e., the acid-grade product, are maximized simultaneously in this problem. The single-objective variant (SGA-II-mJG) of the binary-coded elitist non-dominated sorting genetic algorithm with the jumping gene adaptation, NSGA-II-mJG, is used for the first problem, while NAGA-II-mJG (Guria et al., 2005 a) is used for the second problem. Simplified circuits with a few split streams are proposed on the basis of the present optimization study, which enhance the recovery of the acid-grade product without affecting its quality, for a given total volume of the flotation cell-bank. Even though the optimal solutions have certain drawbacks, they suggest a meaningful direction in using an iterative fitting of parameters to such optimal circuits, followed by re-optimization.