Distal signatures and vectors of hydrothermal systems in carbonates (PI: Zhaoshan Chang)
Many hydrothermal deposits can be hosted in carbonate-rich wall rocks, including porphyry deposits (e.g., Grasberg, Indonesia; Cerro Corona, Peru), and skarns (e.g., Antamina and Uchucchacua, Peru), or have carbonates in the surrounding stratigraphy (e.g., Candelaria, Chile). Beyond the orebodies or massive replacement bodies the hydrothermal footprint continues in the carbonates, by spent fluids and producing weak signals, particularly along faults. The objective of this research is to define the distal signatures of mineralizing fluids in carbonates, including mineralogical and textural signatures, geochemical signals in whole rock, vein coatings and carbonate minerals, C-O isotopes, and fluorescence and cathodoluminescence features of carbonate minerals. Vectors in these parameters towards ores will also be identified. Seemingly similar carbonates are well known to have very different alteration intensity under similar conditions; the project will also develop methods to quantitatively assess carbonate reactivity.
Radar and Seismic Imaging for Mining Hazard Mitigation and Geologic Characterization (PI: John Hole)
This project is developing high-resolution and practical geophysical imaging methods that can be applied from within mines. Targets are 1) mine safety, ground control, and induced seismicity, and 2) delineation of geology for short- to medium-term mine planning. Ground-penetrating radar and seismic reflection are the highest-resolution geophysical imaging methods and complement one another in spatial resolution (sub-cm to 10’s of m) and depth of imaging (10’s of m to km’s). Current research is investigating i) imaging fractures and other geologic features that contribute to ground control (collapse) hazards, and ii) using dense seismic arrays to monitor and characterize active fracturing and earthquake hazards. Both studies illuminate rock mechanics for mine safety.
Machine Learning in Resource Modeling and Mine Planning (PIs: Yaoguo Li and Hua Wang)
Machine learning (ML) has the potential to revolutionize mining by advancing the technology of lofting in resource modeling and estimation. This approach can tap into the capabilities of ML algorithms to integrate big data, uncover hidden relations, and make predictions. Improved accuracy in predicting orebody shapes and quantifying uncertainties translates into monetary value through increased reserves or avoiding wasted mining operation based on false predictions. The aims of the project are to develop adaptive learning lofting methods using supervised and self-supervised ML algorithms. The method will adaptively integrate multiple geoscientific data with sparse drill hole intersections to construct 3D orebody shapes and quantify the spatial uncertainty of the shapes.
Mineralogy Across Scales (PIs: Katharina Pfaff and Thomas Monecke)
Knowledge of both deposit mineralogy and the physical and mechanical properties of rock units is critical at many stages of project development from early exploration to mining and remediation. This project aims to use hyperspectral core scanning data to determine the quantitative mineralogy of drill core and to predict rock physical and mechanical properties. This project is a multi-year project that will involve the following main tasks: (1) identify diagnostic features in hyperspectral spectra to identify the mineralogy using traditional automated mineralogy data for assessment, and (2) find relationships between the mineralogy derived from hyperspectral core scanning and petrophysical properties (i.e., density, hardness, abrasivity). The quantitative mineralogical and rock physical data obtained could then be used to build a 3D block model informing the cost of mining (rock blasting and comminution behavior). The project will make use of machine learning (ML) techniques.
Integrating sequential simulation and visual ensemble analytics for applications in the mining sector (PI: Ryan Pollyea)
Mineral resource exploration results in large quantities of spatially referenced, multi-variate data sets. This project implements recent advances in Bayesian Visual Analytics (BaVA) to discover common attributes (geochemistry or mineral abundance) among features of interest within these multi-variate data sets. The BaVA framework combines the computational efficiency of machine learning with human cognition and expertise to rapidly identify spatial and parametric relationships that may remain hidden in traditional univariate or bivariate geostatistical analyses. Through collaborative case study analyses with CASERM member-partners, this project introduces BaVA methods for data visualization and analysis in the mining sector.