Artificial Intelligence (AI) is revolutionizing mining operations from prospecting to reclamation by optimizing processes, improving safety, and maximizing resource extraction. This article deliberates on AI techniques' role, specifically machine learning (ML), in the mining industry.
Image Credit: Scharfsinn/Shutterstock
Importance of AI in Mining
In mining, AI methods can be applied from the beginning of mining to the end of the mine life-cycle, from prospecting and production to closure and mine reclamation. AI and ML technologies could transform the mining industry's existing technological framework through big data collection, manipulation, handling, and analytics. AI provides accurate and fast on-site decisions, reducing errors in decision-making.
The technology represents a radically efficient and consistent method for effectively assessing potential risks as AI can process and manage huge amounts of data with high efficiency, accuracy, and speed. AI also ensures enhanced health and safety measures, greater resource throughput, lower operational costs, reduced energy demand, and better utilization of mining equipment and vehicles.
AI plays an important role in automated mining, which involves replacing human labor with fully or partially automated techniques or equipment in the mining process. Automation in mining has many benefits, such as a reduction in the overall cost without compromising safety and integrity, improved continuity and consistency of mining systems, and enhanced mine productivity.
Other benefits include increased resource throughput and quality, improved mining equipment performance, reduction in unscheduled maintenance, and enhanced security of personnel by reducing their exposure to risky and hazardous areas.
AI-driven Automation in Mining
Intelligent Drilling Systems and Automated Drillers: Blasting and drilling are the fundamental operations of all mining projects. Holes are drilled into any hard surface/rock to fill them with explosives, and then cracks are induced in the inner geology of the hard surface/rock by blasting the explosives. Using an AI-based computer vision system, the hole dimension perfection and drill hole pattern can be checked.
Similarly, an ML-based intelligent system can compile the entire drilling workflow with better accuracy and speed compared to human drillers. Fuzzy logic-based controllers have been used to develop intelligent drilling systems, while artificial neural networks (ANN) were utilized for over-break prevention and warning for drilling and blasting. ANN has also been used to predict the emulsion-based drilling fluid properties, and the rate of penetration, and identify bit malfunction during the drilling process.
Autonomous Haulage: Truck haulage is the common mode of material transportation from open pit mines. Surface mines/open-pit mines have limited haulage passage for haulage vehicles or trucks.
Using AI algorithms and automated vehicles for routing haulage vehicles, mining companies can achieve greater fuel efficiency, reduced spot creation time per truck, faster hauling, decreased production cost, reduced exception time per truck, and reduced wait time per truck.
Mine Automation Software: Mining companies design software that can administer overall processes to gain more control over the autonomous equipment and overall operation. AI-enhanced software assists and performs location tracking of machines and miners, predictive maintenance of equipment, operational decision-making, and analyzing sensor-achieved large data.
AI-enhanced software is also used for forecasting bottleneck activities in real-time, digital twinning projects, real-time adjustments in shift placements of operations to regulate overall mine efficiency, identifying patterns within productivity variance for organized planning and time management, and discovering correlations in events leading to accidents, fatalities, or perils.
ML Techniques Used in Mining
ML models used in the mining industry are classified into 12 types, including support vector machine (SVM), Gaussian process, instance-based, regression, decision tree, Bayesian, neural network (including subcategories like deep learning), genetic algorithms, dimensionality reduction, clustering, ensemble, and deep learning.
SVM is the most commonly used ML model type, followed by deep learning and ensemble methods. Among the deep learning models, ANN is the most frequently utilized model, followed by convolutional neural network (CNN), deep neural network (DNN), adaptive neuro-fuzzy inference system, extreme learning machine (ELM), autoencoder network, general regression neural network, and recurrent neural network (RNN).
Similarly, the naive Bayesian classifier is the most commonly utilized model among Bayesian models, while the other models used in mining applications are Narmax, Bayesian network, Bayesian logistic regression, relevant vector machine, and naive Bayes. In decision tree and regression model types, classification and regression tree (CART) and multiple linear regression are the most frequently used models, respectively.
Decision trees, M5 rules, logistic regression, linear regression, and multivariate regression are the other models in these segments used in mining. Density-based spatial clustering of applications with noise (DBSCAN) and k-means are the common clustering models employed in mining functions.
Ensemble and decision tree methods are reused extensively for mine planning and evaluation in mineral exploration and targeting. SVMs are employed for mine hazard assessment and land cover monitoring. Deep learning is primarily utilized in equipment management and drilling and blasting, while ensemble methods are used in mine safety and geotechnical management. Moreover, deep learning, ensemble, and SVM methods are used in the exploitation stage, exploration stage, and reclamation stage, respectively.
AI Applications
Mineral Exploration and Prospecting: This is the first phase in the mining process and different AI and ML techniques have been implemented for this phase. For instance, an ANN-based model was developed for mineral resource evaluation for a bauxite deposit. The model performed competitively with the geostatistical kriging technique,
A multilayer perceptron neural network model has been built using geographic information system (GIS) data from diverse exploration projects to prepare a mineral prospectivity map for a gold deposit. Backpropagation ANN was utilized to effectively identify mineral-rich zones using GIS data consisting of exploration and remote sensing data for a goldfield.
A fuzzy logic model has been developed to identify favorable areas for iron oxide, gold, and copper deposits in Finland using geochemistry, geophysics, geological, and mineral occurrence data. The fuzzy logic model was also utilized to predict the mineral potential of multiple commodities in southwestern China. Specific fuzzy-based models were introduced to prepare a mineral prospectivity map to locate new targets in any area using geological and geochemical data.
Mine Planning: Fuzzy logic-based approaches have been extensively used in mine planning. For instance, the fuzzy set approach has been applied for mining method selection considering environment, economic, geotechnical, and geological data for an underground mine in Turkey. Similarly, a fuzzy set-based multi-criterion decision-making process was introduced to select mining equipment for a specific mining operation.
The fuzzy logic-based model was applied successfully for solving previously encountered problems using the conventional method for lignite basin reserves. Moreover, a study displayed that a fuzzy logic-based model is more effective compared to the conventional decision matrix methods for risk assessment and safety evaluation at an underground coal mine in Turkey.
Drilling and Blasting: ANN combined with factor analysis and mean value-based optimization effectively predicts the peak particle velocity for the blast-induced ground vibrations. ANN has also displayed acceptable performance while predicting rock fragmentation for iron ore and limestone mining operations.
An artificial bee colony algorithm was employed to improve the ANN model accuracy to optimize the blasting input parameters and forecast the rock fragmentation. In a recent study, the effectiveness of ANN over conventional statistical techniques and empirical models was demonstrated for predicting blast performance, which is characterized by the degree of resulting vibrations and rock fragmentation during a gold mining operation.
Ore Beneficiation/Mineral Processing: Froth flotation is a commonly used ore beneficiation method for processing metallic ores and/or producing cleaner coal. In this application, AI and ML can assist in designing an algorithm/model that includes optimal flotation conditions involving all crucial variables to maximize the mineral processing plant's objective.
These intelligent models must be implemented in real-time using microcontrollers at the plant to control and monitor the input conditions, ensuring that the plant's metallurgical performance is never hampered. Studies have shown the development of these intelligent systems using ANNs, hybrid neuro-fuzzy systems, SVMs, and random forests for cleaning coal and complex metallic ore beneficiation.
Overall, AI is transforming mining across its entire lifecycle, making every process more efficient, productive, and sustainable. However, challenges like inherent limitations of existing AI models, improvements required in terms of model development, and safe and secure collection of data must be mitigated to exploit the full potential of AI in the mining industry.
References and Further Reading
Ali, D., Frimpong, S. (2020). Artificial intelligence, machine learning and process automation: Existing knowledge frontier and way forward for mining sector. Artificial Intelligence Review, 53(8), 6025-6042. https://doi.org/10.1007/s10462-020-09841-6
Zelinska, S. (2020). Machine learning: technologies and potential application at mining companies. E3s web of conferences, 166, 03007. https://doi.org/10.1051/e3sconf/202016603007
Ghosh, R. (2023). Applications, Promises and Challenges of Artificial Intelligence in Mining Industry: A Review. Authorea Preprints. https://doi.org/10.36227/techrxiv.21493761.v1
Jung, D., Choi, Y. (2021). Systematic Review of Machine Learning Applications in Mining: Exploration, Exploitation, and Reclamation. Minerals, 11(2), 148. https://doi.org/10.3390/min11020148