A novel data-driven sampling strategy for optimizing industrial grinding operation under uncertainty using chance constrained programming

Multi-objective optimization of an integrated grinding circuit considering various sources of uncertainties has been targeted in this work using Chance constrained programming (CCP). Success of CCP depends on accurate transcription of uncertain parameter space for correct estimation of statistical measures, e.g. probability, which is challenging in practical scenarios, where the data available is sparse and difficult to fit using known statistical distributions. To tackle this situation, a novel Data based Intelligent Sampling strategies for CCP (DISC) has been proposed amalgamating the machine learning techniques with novel Fuzzy C-means algorithm. It identifies the data clusters in the sparse uncertain parameter space followed by sampling strictly inside those clusters using the Sobol scheme, which is often not accurately performed by the conventional techniques. Ten parameters depicting uncertainties in the model and feed stream have been considered for optimizing conflicting objectives of productivity, quality and energy savings. A comprehensive comparison displays 42 and 34% improvements over the conventional box and budget sampling techniques, respectively, demonstrating efficacy of the proposed technique. (C) 2020 Elsevier B.V. All rights reserved.

Automated summary

This study focuses on multi-objective optimization of an integrated grinding circuit considering various sources of uncertainties, using Chance constrained programming. A novel Data based Intelligent Sampling strategies for CCP has been proposed, combining machine learning techniques with a Fuzzy C-means algorithm to address sparse uncertain parameter space. The proposed technique demonstrates significant improvements over conventional sampling techniques in optimizing conflicting objectives.