Heuristics for integrated blending optimisation in a mining supply chain. (July 2021)

• Record Type: Journal Article
• Authors: Haonan, Zhou; Samavati, Mehran; Hill, Andrew J.
• Is Part Of: Omega. Volume 102(2021)
• Publication Date: 2021-07
• Subjects: Scheduling -- Integrated blending optimisation -- Open-pit mine optimisation -- Decomposition-based algorithm -- Topological sorting; Management -- Periodicals; 658.4005
• DOI: 10.1016/j.omega.2020.102373
• Languages: English
• ISSNs: 0305-0483

Abstract

Highlights: Production scheduling across the mining supply chain. Integrated blending optimisation across the mining supply chain. Developing heuristics to address the scale and bilinear constrains of the problem. Abstract: In a mining supply chain, products from mines are blended at port terminals to ensure that a set of blending targets (such as grade and qualities) are achieved. The production scheduling problem of each individual mine and the blending problem for a network of mines and ports constitute the integrated blending optimisation, which involves modelling of material flows from mine-side pits to port-side stockpiles. Due to the problem scale and the bilinear constraints for blending behaviours, the problem is computationally hard to solve by any available optimisers. This paper extends upon a decomposition-based algorithm in the literature, which was first to solve the blending problem for a network of multiple mines and ports over multiple time periods. In our paper, a prune routine is proposed to progressively update the mixed integer program of the production scheduling problem for each mine during a rolling-horizon heuristic. Experiments have shown that this extension produces solutions of higher quality than the original algorithm. Furthermore, a ranking-based topological sorting heuristic is presented for selecting units of mineral deposits, known as 'blocks'. Experiments have shown that the average computation time can be reduced by 75.97% when this heuristic is implemented. On top of these extensions, an adaptive algorithm is adopted from the decomposition-based algorithm, featuring faster convergence and higher solution quality at the same time. Comparing our results to the literature, our adaptive algorithm, on average, yields an improvement in solution quality by 12.67% while reducing computation time by 65.09%.