Optimizing an Industrial Grinding Operation Under Uncertainty

Handling uncertainties for parameters in nonlinear constraints using chance constrained programming (CCP) is not as straight forward as its linear counterparts. A simulation based CCP approach which can be thought as an alternative to handle such a situation, has been adopted in this work while solving a multi-objective optimization problem of an industrial grinding operation under various parameter uncertainties. Simultaneous achievement of conflicting objectives such as maximization of grinding circuit throughput and maximization of percent passing mid-size fraction, frames the ideal platform for multi-objective optimization that is studied here for an industrial grinding operational setup with upper bound constraints for various performance metrics (other size fractions, percent solids of the final outlet stream and circulation load of the grinding circuit). The deterministic multi-objective grinding optimization model taken from our earlier work forms the basis of this study on which various effects of uncertain parameters are shown and analyzed in a Pareto fashion. Nondominated sorting genetic algorithm, NSGA II, a popular elitist evolutionary multiobjective optimization approach, is used for solving the problem in hand. Keywords-multiobjective optimization, grinding, uncertainty, NSGA II, Pareto, simulation, chance constrained programming