Drillhole infill has an important role in the mining industry, especially when its aim is to enhance the assessment of variance representativeness of a mineralized rock or any other measured characteristics. Some infill optimization methods propose the use of kriging variance, which is feasible when the goal is to search for sub-sampled regions, but those methods may fail in more complex situations given that a fundamental limitation of kriging variance is to only depend on neighboring samples nearby the estimate location. This paper proposes a method to infer the best location for new drillholes through optimization using as objective function the sum of simulated block variance (SBV), which does not have the same limitation as to the kriging variance. The SBV is reached by stochastic simulation (sequential Gaussian simulation) to compute the variance of each block along with the grid model, and the values are summed to attain the objective function. The objective function minimization is computed by three different methods of search: random search, simulated annealing, and particle swarm optimization. Due to smaller objective function values when applied to a synthetic deposit, simulated annealing with fast cooling schedule algorithm performed better than the others. Further tests led to the conclusion that simulated annealing had more representation of the population. These methods were also applied to a real sampled site, the Capanema Mine, and the simulated annealing with fast cooling also produced the best results with regard to representativeness.

钻孔填充在采矿业中具有重要作用，特别是当其目的是增强对矿化岩石的方差代表性或任何其他测量特征的评估时。一些填充优化方法建议使用克里金方差，这在目标是搜索子采样区域时是可行的，但考虑到克里金方差的基本限制是仅依赖于相邻样本，这些方法在更复杂的情况下可能会失败估计位置附近。本文提出了一种通过优化使用模拟块方差之和（SBV）作为目标函数来推断新钻孔的最佳位置的方法，该方法对克里金方差没有相同的限制。通过随机模拟（顺序高斯模拟）达到 SBV 以计算每个块的方差以及网格模型，并将这些值相加以获得目标函数。通过三种不同的搜索方法计算目标函数最小化：随机搜索、模拟退火和粒子群优化。由于应用于合成沉积物时的目标函数值较小，采用快速冷却计划算法的模拟退火性能优于其他方法。进一步的测试得出结论，模拟退火具有更多的总体代表性。这些方法也被应用到了一个真实的取样地点，即 Capanema 矿，并且快速冷却的模拟退火在代表性方面也产生了最好的结果。