Effective solution of the long-term open pit production planning problem using block clustering

Abstract

Production planning of an open pit mine is defined as determining the extraction sequence of blocks in the final pit limit with the goal of maximizing Net Present Value (NPV) while considering the constraints imposed on the mining system. Mathematical programming models are used to optimize long-term open pit production planning problems. In real mines with several thousands of blocks and several scheduling periods, however, these methods lose their efficiency by increasing the number of decision variables, and it is almost impossible to solve and obtain a solution in a reasonable time using current solvers. In this paper, a new method based on block clustering was proposed to reduce the model size. In the first step, blocks are aggregated in clusters called “mother clusters” using the concept of clustering as a mathematical programming problem and solved using Genetic Algorithm (GA). In the second step, the mother clusters are converted to the mining clusters considering the practicality issues. The results of implementing the proposed method on a real deposit showed that block clustering could decrease the number of binary variables from 476,100 to 5,652. This led to a significant reduction in solution time from 372 hours to 23.4 seconds compared to the original block model. The clustering strategy was also able to realize more than 82% of the NPV obtained from scheduling original blocks.