Large mining environment benefits from accurate models of 3D geological domains for resource estimation and production. In current practice, exploration-based geochemical assays are used to autonomously develop three-dimensional (3D) spatial-models for grade estimation using Gaussian Process (GP) (Rasmussen and Williams, 2006). However, using only exploration data for 3D models limits the accuracy of the domains. This is a result of the lateral sparsity of the exploration holes. Furthermore, there is little additional exploration-hole data gathered by new drilling campaigns. Another source of data collected from mining is that of production data (blast-hole data). Production data is gathered with each new blast-hole pattern drilled. Blast-hole lateral spacing is in the order of 5m, while vertical sampling is coarse and of the order of 10m. The frequency of update for production data and the lateral spatial fidelity of production data contrast significantly to that of exploration data. Exploration models use only information from exploration data, while fused models integrate information from both exploration and blast-hole data sources. The integration of data and knowledge from several sources is vital for robust estimation of expected tonnages in mining and needs to be re-estimated in areas of change that have data or domain variations. Data fusion techniques have been extensively employed in multisensory environments with the aim of fusing and aggregating data from different sensors. The problem thus is in fusing these multiple and multimodal sensor data sets to obtain an integrated model. This study analyses potential biases between exploration-only and exploration fused with blast-hole geochemical-data using cross validation style testing. The dynamics of the models were evaluated through analysis of histograms and spatial scatter plots. The fusion modelling was performed on the banded iron formation (BIF) hosted iron ore deposits located in the Hammersley region of Western Australia.
CITATION: Balamurali, M, Zigman, J and Melkumyan, A, 2019. Analysis of model fidelity between exploration and blast-hole based models: a cross-validation style testing, in Proceedings Future Mining 2019, pp 2628 (The Australasian Institute of Mining and Metallurgy: Melbourne).