Improved estimation with multiple data types for medium term mine planning

Abstract

Drilling is the primary source of geological information in the form of rock samples for geological logging and chemical assays. There are data from multiple drilling types in an open-pit mining operation, with unique technical features, costs, volume support, and sampling error and bias. The most common drilling types are diamond drilling and blast-holes. Drilling data is classified into primary and secondary according to the confidence and quality of the information. Primary data corresponds to the highest quality and confidence, and usually, diamond drilling is considered primary data. Secondary data presents lower confidence and quality, and usually, the blast-holes are considered in this classification. Access to real data is always quite limited. Besides, many times researching requires data with specific geo-statisticl features. Data simulation can provide a solution to these restrictions. This research has developed a procedure for simulating drilling data, with similar statistics and features to the real one and adjustable to the requirements of the user. The procedure simulates the distribution inside each drill hole, replicates drilling supports and sampling protocols to get data from multiple drilling types, specifically diamond drilling and blast holes. Simulated drilling data is used to estimate different resources models used in the ore control process, focusing on the medium-term model. The effects of dataset bias in the profits and ore/waste classification are also assessed for different resources models. Besides the ordinary kriging estimation, a cokriging outline is implemented to take advantage of the multiple simulated data types. The simulation method provides highly realistic data of multiple drilling types. They have been tested several times to check the correct reproduction of input parameters and distribution features. The simulated data have been used in resource estimation and assessing the performance of the different models used in the ore control process. The medium-term model estimated using cokriging and simulated drilling data provides better results and profits than models estimated using a single dataset type. The bias impacts negatively in the profits as expected, but under some conditions, it can diminish the loss in profits by misclassification.