In the field of geological modeling, traditional implicit methods rely exclusively on data obtained from drill cores. In contexts of sparse or irregular sampling, these approaches require manual adjustments to incorporate structural trends, resulting in models with a high degree of subjectivity and elevated uncertainty. To address this challenge, this study evaluates an application-oriented Data-Driven Implicit Geomodeling (DDIM) workflow that combines established machine-learning classification, signed-distance-function-based implicit modeling, and geophysical proxy variables. The objective is to determine whether this operational integration can improve geological boundary delineation and model reproducibility under sparse sampling, rather than to introduce a new machine-learning architecture or interpolation algorithm. The DDIM workflow was evaluated in three application cases and compared with conventional implicit modeling using precision, recall, F1 score, Kappa, and accuracy. Model uncertainty was additionally examined through entropy calculations based on geostatistical simulation. Results from an initial synthetic two-dimensional (2D) lithological model showed an average 5
Gold-bearing Cu–Zn–Au ore from the Zlaté Hory–West deposit was investigated with respect to its response to centrifugal gravity concentration using a KC-MD3 Knelson concentrator. The effects of feed particle size distribution, fluidisation water flow rate, and centrifugal acceleration (G-value) on process performance were systematically evaluated. Process performance was assessed using standard technological indicators, including Au recovery, concentrate grade, mass pull, and upgrade ratio. The highest Au recoveries and separation efficiencies were achieved for the finest size fraction (< 125 μm), confirming the strong influence of particle size on gravity separation efficiency. The experimental results indicate that particle size influences the relative importance of density-related, hydrodynamic, and mechanical factors during centrifugal gravity separation. In particular, the behaviour of coarser particles suggests that density contrast alone may not fully explain their retention within the concentrating bed. The results highlight the importance of particle size and hydrodynamic conditions in determining separation efficiency and define an operational window for effective preconcentration of low-grade gold-bearing ores. The findings provide a basis for further optimisation and pilot-scale validation of centrifugal gravity concentration in polymetallic ore processing.
There are various modes of material transportation in mining industries, such as rope haulage, conveyor belts, and shuttle cars for mined ores.Among these, conveyor belts are the most important and frequently used mode of material transportation in the mining and industrial sectors. After excavation, continuous or uninterrupted transport of bulk materials is required, and belt conveyors are the preferred choice. The failure of one or more essential conveyor components, such as belts, motors, drums, rollers, and couplings, undermines system dependability. The current research provides an in-depth reliability assessment of conveyor belt systems by merging traditional models, including Weibull and exponential distributions, Markov Chain modelling, and sophisticated machine learning techniques to predict component failures. For the current research, data have been collected and analysed for fifteen major components of a typical conveyor belt system. The current research work clearly indicates that certain components, such as belting, exhibit low reliability, while components such as couplings demonstrate higher reliability. Machine learning validation showed highly precise predictions, with R2 scores above 0.91 for each component, MAE values between 0.037 and 0.044, and RMSE values ranging from 0.037 to 0.053. Failure of one component results in the failure of the whole conveyor system; therefore, feature importance analysis has been conducted to identify the key elements affecting system failure. The machine learning model identified load factor ( 0.55), lubrication quality ( 0.35), and temperature ( 0.20) as the key elements affecting failure of the whole system. The current study signifies the importance of merging traditional reliability methods with machine learning for accurate prediction of component life, leading to predictive maintenance.
Gold mineralisation at the Chagupani deposit in northwestern Ghana is structurally controlled by three deformational phases: a D1 east–west shear zone hosting quartz–carbonate–tourmaline veins; a D2 northeast-trending extensional boundary; and a D3 brittle fault network whose intersection with the D1 corridor localises the highest-grade ore shoots. Two supervised machine-learning classifiers, Gradient Boosting Machine (GBM) and Naïve Bayes (NB), were trained on eleven geo-evidential layers derived from airborne magnetic, airborne radiometric, and stream sediment geochemical datasets, and were calibrated on 344 balanced samples (172 gold occurrences and 172 non-gold points placed beyond a 500 m exclusion radius following standard absence-selection protocols) partitioned by spatial leave-out cross-validation. Three testable predictions were derived from the ore-system model: (i) high-prospectivity output should concentrate along the D1 shear corridor; (ii) arsenic should rank as the most influential layer, with the potassic-alteration signal also among the influential predictors; and (iii) GBM should distribute importance across alteration and structural layers while NB registers a univariate arsenic signal. The GBM-based MPM returned higher values than NB on all five evaluation metrics (AUC 0.88 vs 0.85, mAP 0.891 vs 0.860), with precision, recall and F1-score exceeding 0.7 for both models, though the AUC difference was not statistically significant. All three predictions were supported: arsenic ranked as the most influential predictor in both hierarchies, GBM distributed importance more evenly across structural and alteration groups than NB, and both models concentrated high-prospectivity classes within the northern and northeastern sectors, coincident with the D1 shear corridor and D3 intersection zones. The GBM-based classified map captured 87.8
Iron ore tailings (IOTs) are produced on a large scale during iron ore beneficiation and are commonly stored in tailings storage facilities (TSFs), where long-term accumulation can create geochemical and geotechnical liabilities. This structured narrative review links valuable-element recovery, post-extraction residue transformation, activation, cementitious utilization and TSF stabilization within a cascade framework. A structured literature search covering publications from 2000 to 2026 was combined with targeted quantitative synthesis of composition, recovery, activation, cementitious performance and environmental assessment data. Representative studies reported a 61.3
