Solving complex mine optimisation problems using blend vectoring and multi-objective production scheduling

Abstract

Developing a large scale open pit mining project constitutes a complex process where decisions made at the planning level can affect the overall project economics and value in the scale of billions of dollars. In this paper, it will be demonstrated how the project reserve, and expected economic value can be increased substantially depending on the ability within both the optimisation and scheduling process in accounting for material process classification based on a blend vectored resource. This will be one of the first examples within the literature where the differences in ability to handle this fundamental concept is quantified as comparative analysis is provided to conventional approaches. These are applied within a practical implementation for a Western Australian iron ore case study. The elementary methodology of traditional cut-off optimisation to determine ‘ore’ is evaluated in comparison to results generated from applying metaheuristic blend vector techniques and derived mixed integer linear programming formulations. It will be shown how generating the results based on these techniques is computationally efficient in practice, with minimal engineering time consumed to generate solutions despite datasets generally being very large. Furthermore, post-schedule level results obtained from conventional scheduling and multi-objective genetic algorithms are provided. These explicitly demonstrate how the overall project value possesses a strong dependence on the mine planning conditions imposed by different approaches. A point of view on the open pit mine production schedule optimisation problem is also given that has allowed the development of high quality solutions in practical instances.