Project details

Approach: Analytics on downhole geophysics, drill machine sensing, on-rig XRF sensing, and correlations to core geometallurgical/structural data.

Sensing during drilling will be used to generate three-dimensional maps of P-wave velocity (1). These velocities are directly related to many rock mechanics properties, and this project establishes the relationships between these properties and the sensed data. In particular, the project establishes the quantitative relationships between the drill sensor data and the geometallurgical attributes of the ore in the resource model. Multi-modal data such as imaging data, on-rig X-ray fluorescence (XRF) spectrometric and X-ray diffraction (XRD) data are also inputs for analysis. These data include information on morphology/grain size/textures/structures, chemical composition and mineral speciation.

The ultimate aim of this project is to obtain, integrate and analyse useful data on minerals in real time during resource drilling using multivariate statistics and/or statistical machine learning techniques. These data is being converted to information that can be used to understand the behaviour of the mined product in the crushing-grinding-concentration circuit, and is providing knowledge essential for optimising plant performance and maximising recoveries.

A particular family of machine learning techniques, namely, Deep Learning, has demonstrated considerable potential for the analysis of complex images and audio data. They have recently been used in the analysis of mining data and show state-of-the-art performance (2). The project is exploring and developing deep learning models to achieve accurate close-to-real-time processing of drill sensor information of various types, and at the same time enabling the retrieval of relevant information from off the-shelf resource knowledge databases.

The project is delivering breakthrough technologies that will, in turn, enable new, close to real-time approaches to the determination of geometallurgical attributes, with potential for widespread deployment across the minerals industry.

Image courtesy of Boart Longyear – Sonically drilled core samples


Xiaomeng Gu


University of Adelaide

Key Contacts

Principal Supervisor:
Professor Nigel Cook

Email: nigel.cook

Ph: +61 08 8313 1096

Professor Chris Aldrich
Associate Professor Andrew Metcalfe

Supervisor Profile


Gu, X., Cook, N., Metcalfe, A., Aldrich, C. 2024. A machine vision approach for detecting changes in drill core textures using optical images. Publication pending.


(1) Williams, P.K., Urosevic, M. Kepic, A. and Whitford, M. (2012). Recent experience with use of high definition seismic reflection for nickel sulphide exploration in Western Australia. 74th European Association of Geoscientists & Engineers Conference & Exhibition.
(2) Caleb Vununu, Kwang-Seok Moon, Suk-Hwan Lee, and Ki-Ryong Kwon (2018). A Deep Feature Learning Method for Drill Bits Monitoring Using the Spectral Analysis of the Acoustic Signals. Sensors (Basel). 18(8): 2634.

Industry Partners