Incorporating Spatial Variability of Lithological Units into Ore Grade Estimation of an Indian Limestone Deposit
The ore grade estimation requires accurate prediction of the grade at location from limited borehole information. It plays the dominant role in the decision-making process for investment and development of various mining projects and hence become an important and crucial stage.This paper evaluates the application of multi-layer perceptron neural network (MLP NN) architecture to improve the predictability in grade estimation from west coast limestone deposit, Chandrapur district, Maharashtra. The spatial variability of lithological information is incorporated as secondary information in the model for grade estimation. In this investigation the three dimensional spatial coordinates along with four underlying lithological units are taken as input variables and, the four grade attribute of limestone deposit such as CaO, Al2O3, Fe2O3, and SiO2 are taken as the output variable. The comparative analysis of these models have been carried out and the results obtained, are validated with geostatistical method ordinary Kriging (OK). The observed value of various performance criteria viz. regression coefficient and mean square error revealed that the MLP NN performed well as compared to OK in terms of generalization and predictability of ore grades.