石灰岩体中の不純物成分含有量の空間分布推定と影響因子解析

Abstract

Factors that determine the suitability of limestone for industrial use are contents of calcium oxide (CaO) and impurities. From 244 sample points in 18 drillhole sites in a limestone mine, the content data of four impurities, SiO2, Fe2O3, MnO, and P2O5 were collected. Since the spatial correlations of content data are not clearly shown by variogram analysis, a feedforward neural network was applied to estimate the content distributions. The network structure consists of three layers: input, middle, and output. Input data to the network are coordinates of a sample point, lithology such as conglomeratic limestone, and kind of fossil. Output data from the network are the contents of SiO2, Fe2O3, MnO, and P2O5. Numbers of neurons in the middle layer and training data vary with each estimation point to avoid overfitting of the network. Several important characteristics of the three-dimensional content distributions were detected through the network such as the continuity of low content zones of SiO2 along a Lower Permian fossil zone trending ENE-WSW. The neural network-based method was superior to a geostatistical method in spatial estimation accuracy and dealing with multivariate data. To evaluate uncertainty of the estimates, the method that draws several outputs by changing coordinates slightly from the target point and inputting them to the same trained network is proposed. The uncertainty differs with impurities, and is not based on just the spatial arrangement of data points. Influence-factor analysis of the network clarifies a strong effect of crystalline limestone on the P2O5 contents. Hydrothermal alteration, which could cause leaching and secondary concentration of phosphorus, is considered to have produced the effect.