Innovation mine optimization tailored to iron ore operations: long and short term planning with dynamic truck dispatching

加拿大的铁矿石储量约为63亿吨，居世界第八。铁矿石精矿（62%铁）价格的暴跌（从2011年2月的每吨187美元降至2016年初的每吨45美元）导致加拿大的一些业务停止，许多开发项目暂停。采矿业为当地社区和政府做出了重大贡献。因此，价格急剧下跌可能会产生深远的社会影响。应对这种负面经济氛围的唯一方法是找到创新的方法，以确保运营的效率、绩效、生产力、可持续性和/或盈利能力。本研究利用前沿的运筹学、决策、机器学习和不确定性管理工具，针对露天矿开采的具体特点，提出了有效的矿山规划和优化方法。首先，将实施多变量分析，以了解铁矿石变量之间的关系。考虑到一些低品位的材料可以以一定的混合成本包含在矿石中，以最大限度地提高矿石吨位，建立了一种新的矿石废物鉴别模型。该模型将作为一种投入用于长期规划。在参数化过程中，该方案应不断更新。将研究拉格朗日松弛法，以找到适当的乘数，使年产量与能力要求相适应。然后，结合Krige的关系，对短期规划和床型设计进行优化。当短期规划通过计算效率高的元启发式解决加工设计的等级要求时，床混合设计将使加工输入材料的等级波动最小化。短期规划的成功与否，在一定程度上取决于设备管理。在此范围内，将探讨一种高效的卡车调度策略。最后，将进一步研究与数据不确定性相关的敏感性和风险。因此，采矿业将有新的规划工具，侧重于铁矿石的具体特点。

Canada has approximately 6.3 billion tonnes crude iron ore reserves, which is ranked eighth in the world. The slump in iron ore concentrate (62% Fe) prices (from USD 187 per tonne in February 2011 to USD 45 per tonne at early 2016) led to cessations of some operations and suspensions of many development projects in Canada. Mining industry makes significant contribution to local communities and governments. As such, dramatic price decline may have profound societal effects. Only way to respond to this negative economic aura is to find innovative ways such that efficiency, performance, productivity, sustainability and/or profitability of an operation are ensured. This research proposes efficient mine planning and optimization approaches addressing to specific characteristics of open pit iron ore operations using cutting-edge operation research, decision making, machine learning and uncertainty management tools. First of all, a multi-variate analysis will be implemented to understand relationships among iron ore variables. A new ore - waste discrimination model will be developed in such a way as to maximize ore tonnage considering that some low grade material can be included in ore at a blending cost. This model will be used in long-term planning as an input. Throughout the parameterization, this ore - waste scheme should be updated continuously. A Lagrengian relaxation approach will be investigated to find appropriate multipliers such that annual productions are compatible with capacity requirements. Then, a short-term planning and bed-blending design will be optimized in connection to Krige's relationship. When short-term planning addresses grade requirements of processing design through a computationally efficient meta-heuristic, bed-blending design will minimize grade fluctuations of processing input material. The success of short-term planning, a certain extent, depends upon equipment management. In this scope, an efficient truck dispatching strategy will be explored. Finally, sensitivities and risks associated with data uncertainties will be further investigated. Thus, mining industry will have new planning tools that focuses on specific characteristics of iron ore.