On globally optimizing a mining complex under supply uncertainty: Integrating components from deposits to transportation systems

Mining complexes are generally comprised of multiple deposits that contain several material types and grade elements, which are transformed in available processing destinations and transported to final stocks or ports as saleable products. These components, associated with a mining complex, encompass multiple sequential activities: (i) Mining the material from one or multiple sources; (ii) blending the material including stockpiling; (iii) transforming the material in different processing destinations considering operating modes; (iv) transporting the transformed material to final stocks or ports. Since these activities are strongly interrelated, their optimization must take place simultaneously. Conventional mining optimization methods suffer from at least one of the following drawbacks when optimizing mining complexes: some decisions are assumed when they should be dynamic (operating modes, destination of mining blocks, etc.); component based objectives are imposed, which might not coincide with global objectives; many parameters are assumed to be known when they are uncertain. Past research works have demonstrated that geological uncertainty is the main cause of the inability of meeting production targets in mining projects. This thesis presents methods to optimize mining complexes that simultaneously consider different components and account for geological uncertainty. A multistage methodology that uses simulated annealing algorithm to generate risk-based production schedules in mining complexes with multiple processing destinations is presented and implemented in Escondida Norte (Chile) copper dataset. Its implementation using Escondida Norte dataset generates expected average deviations of less than 5% regarding mill and waste targets, whereas a mine production schedule generated conventionally over a single estimated model generates expected average deviations of 20 and 12% for mill and waste targets respectively. An iterative improvement algorithm that considers operating modes at different processing destinations is developed and applied to a copper complex. The objective function seeks for maximizing discounted profits along the different periods and scenarios (orebody simulations). The implementation of the method at a copper deposit allows reducing the expected average deviations from 9 to 0.2% regarding the capacity of the first process while increasing the expected NPV by 30% when compared with an initial solution generated conventionally. A method that uses simulated annealing at different decision levels (mining, processing and transportation) is described and tested in a multipit copper operation. The implementation of the method in a multipit copper operation permits the reduction of the expected average deviations from the capacities at two mills from 18-22% to 1-3% and the expected average deviation from the targets regarding two blending elements from 7-1.8% to 0.3-0.6% when compared to an initial solution generated conventionally. The expected NPV also improves by 5%. The previous method is extended to mining complexes that combine open pit and underground operations and it is tested in a gold complex in Nevada. The implementation of the method at Twin Creeks gold complex in Nevada shows improvements in meeting the metallurgical blending requirements while increasing the expected NPV by 14%. The formulations described in this thesis encompass a large number of integer variables given the discretization of the mineral deposits. To solve the problems, efficient optimization algorithms are implemented with significant improvements when compared with conventional deterministic approaches. These algorithms outperform conventional methods regarding expected NPV and meeting targets at the different components of the value chain.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.