IDW矿产资源/储量估算方法精细幂指数的智能优化

Although inverse distance weighting (IDW) is a kind of widely-used mineral resources and ore reserves (MROR) estimation method, the subjectivity and arbitrary in the determination of the power value and the associated weakness in IDW estimation has been uniquely concerned and studied in this project. The drafted solutions were considered from the following two aspects: for estimation to a certain unknown point by IDW method, (1) all the participated samples share a same power exponent, named IDW-OPPN (IDW with One Power to Samples in a Neighborhood) method; (2) all the participated samples have different power exponents compared with others, named IDW-MPPN (IDW with Many Powers to Samples in a Neighborhood) method. With respect to the efficiency, quality and intelligence, several utility techniques, such as genetic algorithm, cross validation and so on, will be employed in implement of the two methods. Thus, in theory the proposed method could not only intelligently select the optimal power values for IDW estimation, but also manage usual problems in IDW estimation using samples with non-uniform distribution in spatial variability or positions, which are very common situations in MROR estimation. As to the planned experiment process, a serial of theoretical model data and practical mineral data will be used for the performance test.Additionally, lots of measurements for global accuracy (e. g. , mean, variance, variogram and so on) and local accuracy (e. g. , average absolute error, average relative error, root mean square error (RMSE) and so on ) will be applied to evaluate the quality and performance of the results from the proposed method. If superiority of this project can be proved in the later experiment process as expected, the proposed method will be superior to many similar methods including geo-statistics ( in describing locally anisotropy, for instance ), with significant reality sense to enrich and perfect the theory of MROR estimation by IDW. What's more, the proposed method can be easily extended to work in other fields where conventional IDW method works well.

针对距离幂次反比（IDW）矿产资源/储量（MROR）估算方法的幂指数难以有效设置及由此造成的估值精度通常不高等问题，从两个层次提出基本解决方案：单次估值过程中，（1）所有参估样品使用相同幂指数（IDW-OPPN）的方法；（2）所有参估样品使用不同幂指数（IDW-MPPN）的方法。为了使这两种方法高质效、高智能，将采用遗传算法和交叉验证等技术进行实现。在自适应地动态计算出每个待估点最优幂指数的同时，所提方法还将能够较好地处理MROR估算过程中经常遇到的样品数据呈非均匀分布或属性值呈非均匀变化情况下的估值问题。将利用多个理论模型数据和实际矿山数据同时从整体精度和局部精度两方面对估算结果的质量进行验证对比。按预期，新方法将不再需要手工设置幂指数，并且能够较大程度提高传统IDW法的估值精度，在局部空间变异性的刻画方面甚至可能超越地质统计学方法。对丰富和完善IDW法MROR估算理论与方法有较大意义。

针对距离幂次反比（IDW）矿产资源/储量（MROR）估算方法的幂指数难以有效设置及由此造成的估值精度通常不高等问题，从两个层次提出基本解决方案：单次估值过程中，所有参估样品使用相同幂指数的方法和所有参估样品使用不同幂指数的方法。为了使这种方法高质效、高智能，基于交叉验证的技术框架下对其进行了实现。在自适应地动态计算出每个待估点最优幂指数的同时，所提方法可以较好地处理MROR 估算过程中经常遇到的样品数据呈非均匀分布或属性值呈非均匀变化情况下的估值问题。利用多个理论模型数据和实际矿山数据，从整体精度和局部精度两方面对估算结果的质量进行验证对比。这种方法能够自动设置幂指数，实验表明，特定情况下优化后的IDW法可以达到同等条件下的普通克里格法估计值的精度。同时，它在三维地学建模及不确定性评估的应用方面，相对于克里格法，也具有原理简单、计算速度快、插值效果好的优势。