Amelioration de l'optimisation des designs de fosses par l'approche stochastique

n open pit design consists in defining the ultimate pit shell beyond which the exploitation is not to be desired. The primary objective of open pit optimization is to find the pit design that maximizes the profit. This process is presented as a combinatorial optimization problem where the objective is to find a selection of blocks that maximizes profit.;The approach proposed in this thesis is to use conditional simulations in a stochastic optimization program to include the geological risk. The algorithm optimizes the expected profit of the blocks computed from the conditional simulations. The optimization algorithm used is a "maximum flow" type algorithm calculated on a directed graph. Because of the convexity of the profit function, the pit computed on expected profit differs substantially from the traditional one obtained with kriged grades.;In this thesis, it is shown that the stochastic optimal solution predicts a higher or equal profit than the traditional optimal solution (under the assumption that the average of simulated block grades corresponds to the kriged grade). This theoretical result is confirmed by two synthetic case studies.;A first case study was conducted on a fully controlled synthetic deposit with known characteristics in order to assess the quality of the predicted pit profits. The results show that stochastic optimization yields 8.66% more profit than the traditional optimization. Moreover, the amount of gold recovered is 73.81% higher with the stochastic pit.;The traditional approach, currently used in the industry, relies on kriged block grades from which a profit for each block is calculated based on a convex profit function. The profit function's inputs are the mining costs (drilling, blasting and hauling), the processing costs, the block economical classification (ore/waste) and the recoverable block value. The block profits are then subjected to an optimization algorithm that finds the selection of blocks that maximizes profit. Regardless of the optimization algorithm used, it still considers than the block profits are known and certain. This hypothesis is always invalid because it is impossible to perfectly estimate ore grades. The geological risk related to ore grades is known as a dominant factor in mine valuation and for possible project failure. The open pit optimization approach greatly benefits from the integration of grade uncertainty. One approach to this problem is to use geostatistical simulations to model grade uncertainty.;A second case study was conducted on Osisko's Canadian-Malartic gold deposit. The stochastic approach predicts a significant increase in recoverable gold and in pit profit compared to the traditional pit design. Furthermore, the stochastic pit provides an average increase of 4.04% (453.4 K oz) on the real amount of gold recovered and an average increase of 1.33% (US...