Peter Alan Dowd

Accurate and thorough detection of rock mass discontinuities is essential in geotechnical engineering as they have a significant impact on the stability of underground and surface excavations, yet they are difficult to automate. In the work reported here, an automatic algorithm, termed PhaseVote, is proposed for the extraction of discontinuity traces from the colour (RGB) attributes of 3D meshes which, combined with the region growing method, can help derive more accurate discontinuity modelling of a rock mass. Conventional edge detection methods degrade under low contrast and variable illumination, and geometry-based point-cloud methods overlook spectral features. The proposed approach uses contrast-invariant phase-based feature detection. The detected edges are filtered and then extended or pruned to form major discontinuity traces using the tensor voting method. The first algorithm, phase analysis, detects edges independently of local contrast, while the latter, tensor voting, links and denoises the resultant edge fragments into continuous traces using perceptual grouping. The detection capability of the proposed method is assessed by applying it to two case studies and comparing the results with an existing method. PhaseVote achieved detection rates of 94.85% and 84.72% and IoUs of 70.27% and 69.80%, compared with 76.12%/45.95% and 69.81%/43.81% for a conventional existing method. The results demonstrate that the proposed method is resilient to image contrast, robust to changes in the orientation of discontinuity planes, is less subjective and has a higher degree of automation.

This study investigates the influence of ore particle size on gravity-driven flow behaviour and dilution at the drawpoint in sublevel caving operations using a particle-based modelling framework implemented in the Blender rigid body physics engine. Two representative case studies with non-uniform particle-size distributions are analysed: the Ernest Henry Mine, supported by ring marker trial data, and the DeGagné dilution curve derived from discrete element simulations calibrated against Ridgeway mine data. For both case studies, a systematic calibration procedure is conducted to identify suitable modelling configurations and evaluate the sensitivity of key physics-engine parameters. Half-geometry simulations are adopted to overcome hardware limitations. The results show that particle linear damping is the dominant parameter controlling flow behaviour, while particle friction and geometry also significantly influence flow responses. Simulations using purely spherical particles underestimate inter-particle contact effects and produce less realistic behaviour. The most representative results are obtained using lower damping for ore relative to waste particles and non-spherical dodecahedral particles. Following calibration, increasing ore particle size by 10% and 20% leads to higher dilution and lower ore recovery, particularly at intermediate extraction levels, due to reduced ore mobility and earlier waste breakthrough.

As fuel consumption is a major operating cost in open-pit mining, identifying and quantifying controllable inefficiencies requires continuous monitoring under real production conditions. This study presents a comprehensive analysis of haul truck fuel consumption using an integrated onboard monitoring system installed on a Komatsu 785 truck at a copper mine. The system continuously measured fuel flow, payload, position and speed over 90 days and captured 1780 complete haul cycles over 150 shifts under varying operational, environmental and behavioural conditions. Specific fuel consumption (SFC) analysis revealed that road gradient was the dominant factor, with uphill waste-haul routes consuming 75–80 gr/t.km compared to 65–70 gr/t.km on crusher routes. An optimal payload range of 92–97 tonnes minimised the SFC, while both under-loading and over-loading reduced fuel efficiency. Spatial mapping identified specific high-consumption road segments that required maintenance. Driver behaviour significantly affected fuel use, with aggressive driving increasing the SFC by approximately 7–10% (75–77 gr/t.km) compared to normal driver behaviour (70–72 gr/t.km). Rainfall events increased SFC by 2–3 gr/t.km above the baseline due to elevated rolling resistance. Operational delays, particularly loading queues (45% of total idle time), contributed substantially to unproductive fuel consumption. The findings demonstrated that fuel inefficiencies were largely controllable through targeted operational improvements, including payload optimisation, road maintenance, driver training and dispatch co-ordination. The spatial mapping methodology provided a practical diagnostic tool transferable to other mine sites for identifying energy-intensive haulage segments and prioritising fuel-reduction interventions.

This study compares the accuracy of empirical and regression models for predicting the size distribution of blasted material at drawpoints in sublevel caving. Field data were sourced from Ernest Henry mine (EHM). At EHM, full-scale blasting trials were conducted under controlled conditions by varying explosive density and burden size, with fragmentation measured using laser scanning at different extraction tonnages. To minimise scanning inaccuracies, scan results were combined to represent the particle-size distribution for the EHM data. Five models were evaluated: Kuz-Ram, extended Kuz-Ram, Two Component Model (TCM), Kuznetsov–Cunningham–Ouchterlony, and a regression-based Underground Ring Blasting Model (URBM) developed in our previous work. Models were assessed for predicting P20, P50, and P80 passing sizes. Results show the extended Kuz-Ram and TCM perform better for finer fragment sizes, with Mean Absolute Percentage Error (MAPE) of 8%. URBM was the most accurate for median and coarse sizes (MAPE 5% for P50 and 3% for P80) and showed the least variability in errors. A SHapley Additive exPlanations-based sensitivity analysis of the three most accurate models identified key variables for median size prediction, highlighting the importance of rock properties and ring design parameters.

This study compares the accuracy of empirical and regression models for predicting the size distribution of blasted material at drawpoints in sublevel caving. Field data were sourced from Ernest Henry mine (EHM). At EHM, full-scale blasting trials were conducted under controlled conditions by varying explosive density and burden size, with fragmentation measured using laser scanning at different extraction tonnages. To minimise scanning inaccuracies, scan results were combined to represent the particle-size distribution for the EHM data. Five models were evaluated: Kuz-Ram, extended Kuz-Ram, Two Component Model (TCM), Kuznetsov-Cunningham-Ouchterlony (KCO), and a regression-based Underground Ring Blasting Model (URBM) developed in our previous work. Models were assessed for predicting P20, P50, and P80 passing sizes. Results show the extended Kuz-Ram and TCM perform better for finer fragment sizes, with Mean Absolute Percentage Error (MAPE) of 8%. URBM was the most accurate for median and coarse sizes (MAPE 5% for P50 and 3% for P80) and showed the least variability in errors. A SHAP-based sensitivity analysis of the three most accurate models identified key variables for median size prediction, highlighting the importance of rock properties and ring design parameters.

The growing global demand for mineral resources is challenging mining operations to maintain productivity while processing lower-grade ores and increasingly complex deposits. This study presents an integrated framework that leverages machine learning (ML) and high-fidelity simulation to model and support scenario-based decision-making for the blasting–crushing–SAG (Semi-Autogenous Grindin) milling chain using a calibrated flowsheet. Using publicly available data from the Barrick Cortez Mine (Nevada, USA), more than three million operational scenarios were generated using the Integrated Extraction Simulator (IES) to capture system variability and sensitivity. Machine learning meta-models, built using Random Forest and XGBoost methods, were trained on the simulated data and achieved coefficients of determination (R²) exceeding 0.90 across all key outputs, including P20, P50, P80, and mass flow rates at different operational stages. The meta-models accurately reproduced plant-scale behaviour while reducing computational requirements by several orders of magnitude compared with full-scale simulations. SHapley Additive exPlanations (SHAP) analysis revealed that blast-hole diameter, explosive energy parameters, screen cut-size, crusher feed characteristics, and SAG mill operating conditions are the dominant factors impacting downstream particle size distributions. The proposed framework enables near-real-time evaluation of “what-if” operational scenarios and provides transparent, quantitative decision-support for integrated mine-to-mill optimisation.

This study presents an integrated, cost-aware artificial intelligence (AI) meta-modelling framework for mine-to-mill optimisation that couples high-fidelity simulation with data-driven predictive modelling. Using over three million scenarios generated in the Integrated Extraction Simulator (IES), the framework quantifies how upstream design parameters such as burden, spacing, hole diameter, and explosive density propagate through screening, crushing, stockpiling, and grinding to affect downstream costs and throughput. Random Forest-based meta-models achieved predictive accuracies above 90%, enabling the rapid evaluation of technical and financial trade-offs across the mining value chain. Stage-wise cost models were formulated for drilling, blasting, comminution, and material handling to link process variables to costs per tonne. The results reveal clear non-linear cost responses: finer fragmentation reduces the total comminution cost despite higher explosive expenditure, while SAG mill load and speed exhibit U-shaped cost relationships with distinct optimal operating windows. By combining physics-based simulations, machine learning, and cost integration, the framework transforms traditional stage-wise optimisation into a holistic, financially informed decision-support system. The proposed methodology supports real-time, AI-enabled digital twins capable of adaptive mine-to-mill optimisation, paving the way for more efficient and sustainable resource extraction.

The growing global demand for mineral resources is challenging mining operations to maintain productivity while addressing lower-grade ore and increased extraction complexity. Despite the availability of vast datasets across mining stages, much of this information remains underused in decision-making. This study presents an integrated, knowledge-based framework that leverages artificial intelligence (AI) and high-fidelity simulation to model and optimise the full mine to mill process. Using publicly available data from the Barrick Cortez Mine in Nevada, USA, the mining chain from blasting to semi-autogenous grinding (SAG) was modelled using the Integrated Extraction Simulator (IES) from Orica. To mitigate the computational burden of full factorial simulations, three million scenarios were generated to evaluate performance sensitivity. Machine learning models, including linear regression, decision trees, random forests, and XGBoost, were trained and validated. The models achieved an accuracy of more than 90%, underscoring their reliability for predicting process outcomes. SHapley Additive exPlanations (SHAP) were applied to interpret model predictions and quantify feature importance. The findings confirm a strong alignment between simulation and real-world data and highlight key operational parameters that affect downstream process performances. This meta-model approach offers a powerful tool for real-time decision-making, enabling mining operations to improve efficiency, reduce costs, and support sustainable resource management.

The main purpose of this work is to illustrate a methodology for mapping (identification and delineation) mounds and mountains on the lunar surface. In doing so, a precise meaning is given to positive relief in relation to a local base level which is specific to each mapped feature. The procedure is useful for mapping very shallow volcanic domes and cones, shallow wrinkle ridges and any other positive relief as shown in different test sites. The mapping of lunar mountains has allowed their sizes to be determined and compared with mountains on Earth which are defined using the same methodology. This paper reveals which is the mightiest mountain on the Moon and, by comparison to two emblematic Earth mountains (Mount Everest and Mount Kilimanjaro), it is shown that lunar mountains are mightier than Earth mountains despite the fact that they are more rounded and do not have the steep slopes of mountains on Earth. When comparing their slopes, the mean slopes inside circles of 3 km radius around the summits are, respectively, 39º, 26º and 12º for Mount Everest, Mount Kilimanjaro and the mightiest lunar mountain. It is also shown that significant mountains on the Moon, in the form of volcanic cones, crater central peaks and mighty mountains, have the form of a three-sided pyramid. Given the DEM spatial resolution, the procedure is able to map any mound, which is illustrated by identifying very shallow volcanic domes and volcanic cones. Central crater peaks are also interesting mountains of medium altitude that may reach several thousand metres. The identification and delineation of mountains on the Moon is illustrated by using the LOLA digital elevation model but the methodology is completely general and the only limitation is the spatial resolution of the digital elevation model used to define the scale of the positive reliefs that can be identified and delineated.

Rapid resource model updating with real-time data is important for making timely decisions in resource management and mining operations. This requires optimal merging of models and observations, which can be achieved through data assimilation, and the ensemble Kalman filter (EnKF) has become a popular method for this task. However, the modeled resources in mining usually consist of multiple variables of interest with multivariate relationships of varying complexity. EnKF is not a multivariate approach, and even for univariate cases, there may be slight deviations between its outcomes and observations. This study presents a methodology for rapidly updating multivariate resource models using the EnKF with multiple data assimilations (EnKF-MDA) combined with rotation-based iterative Gaussianization (RBIG). EnKF-MDA improves the updating by assimilating the same data multiple times with an inflated measurement error, while RBIG quickly transforms the data into multi-Gaussian factors. The application of the proposed algorithm is validated by a real case study with nine cross-correlated variables. The combination of EnKF-MDA and RBIG successfully improves the accuracy of resource model updates, minimizes uncertainty, and preserves the multivariate relationships.