PhD Studentship - Adaptive automated geological modelling for sustainable mining application

Project details:

Adaptive Automated Geological Modelling for Sustainable Mining Applications Sustainable mining requires a proper assessment of the risks involved in any decision in the mining value chain. Currently this possibility is hindered by the lack of automated end-to-end modelling workflows for risk and sensitivity assessment. Although many steps in the process can be currently automated, the construction of the geological model remains a significant challenge. The geological model is key because it constrains the volumes and data used when inferring the mineral resources, which in turn inform the mine plan and geometallurgical model of the operation. This project addresses this challenge by proposing an adaptive automated approach to geological modelling, to define geological domains and orebody volumes in a reliable, repeatable, and automated way. This project aims to establish a data-driven, adaptive framework that develops artificial intelligence tools, integrated with advanced geostatistics to deliver automated, auditable geological models that can be updated dynamically as new data becomes available.

The project involves four components to resolve this challenging problem:

(1) Inference of patterns from drillhole data. Drillhole loggings, elemental concentration assays, and mineralogical data derived from techniques such as Mineral Liberation Analysis (MLA) or hyperspectral core scanning provide large and complex datasets. By applying advanced pattern recognition and clustering algorithms, the aim is to automatically detect coherent spatial domains. These domains represent regions with similar mineralogical or geochemical signatures, serving as the basis for subsequent geological interpretation.

(2) Directional structural inference. Determining the shape and connectivity of the geological volumes can be dealt with using Discrete Cosine Matrices (DCM) and quaternion-based orientation modelling. Mining deposits are rarely isotropic, their geometry reflects geological processes such as folding, faulting, and intrusions. By inferring local directions and orientations directly from the data, this information can be fed into the modelling of the volumes that define these domains in space. This enables the modelling framework to account for locally varying directions in mineralogical or geochemical domains, improving the realism of spatial models, and ensuring that domains follow geological trends.

(3) Spatial extent of domains. Radial Basis Functions (RBF) can be used to infer the boundaries between different domains in a probabilistic framework, thus accounting for uncertainty in the domain definitions. This probabilistic approach reflects the uncertainty and variability of geological systems, providing, in addition, a measure of confidence.

(4) Calibration Ai architecture. The final component develops a calibration AI architecture designed to cross-validate and refine the definition of orebody volumes. This architecture functions as a self-correcting loop, integrating bagging and boosting iterations to continuously calibrate the model.

Together, these components create an adaptive, automated modelling framework that can help assess the impact of the uncertainty and variability in the extent of the geological domains in an end-to-end modelling framework for resource estimation and geometallurgy.

Please direct project specific enquiries to: Julian M. Ortiz (j.ortiz-cabrera@exeter.ac.uk) Please ensure you read the entry requirements for the potential programme you are applying for. To Apply for this project please click on the following link -https://www.exeter.ac.uk/study/funding/award/?id=5734

Funding details - 

Payment of tuition fees (Home), Research Training Support Grant £5,000 over 3.5 years

Advert information