MACHINE LEARNING ENHANCEMENTS FOR KNOWLEDGE DISCOVERY IN MINERAL EXPLORATION AND IMPROVED MINERAL RESOURCE CLASSIFICATION
Contemporary issues of the mining industry, such as declining grades, increasing depth of deposits, cost per discoveries and demand for the raw materials are driving the mining industry to develop and adopt improved technologies. The main objective is to achieve success in exploring deeper parts of the earth and improve mining and processing technologies. These emerging technologies generate enormous amounts of data, therefore utilizing all the available data efficiently has become one of the biggest challenges for geoscientists and engineers. Machine learning (ML) algorithms are being deployed to handle the abundance of data in many traditional fields (e.g., computer vision, natural language processing and genetics) due to their fast processing capability. Despite its potential benefits, ML has not been fully adopted by the mineral exploration and resources sector yet. This thesis demonstrates that ML can assist in improving mineral exploration strategies and mining procedures by enhancing the understanding of processes from high dimensional data and by automating operations. An overview of state-of-the-art applications of different ML approaches and their relevance to the mineral exploration and resources sectors is presented by means of two original case studies. Furthermore, a generic workflow for ML application in geoscience is proposed to highlight good practices in their implementation. The first case study encompasses an example of a multivariate analysis applied to a lithogeochemical dataset from the Vazante-Paracatu District, Brazil, in order to provide insights about processes related to the base metal mineralizing system. The complex relationships between the data and the mineral occurrences are revealed to assist in finding new targets for zinc exploration. The second case study is an example of ML applied to resource classification, which normally relies on expert assessment of a qualified person to determine if the blocks of a 3D mineral resource model are classified as measured, indicated, or inferred. This study reveals that ML can assist to increase time efficiency of the task and improve consistency by automating the process. These applications demonstrate the relevance of ML methods in supporting knowledge discovery in geosciences and engineering and automating processes for improved consistency in the results.
URI for this recordhttp://hdl.handle.net/1974/28138
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item: