Faites connaître cette offre !
Reference : UMR7359-JULMER-001
Workplace : VANDOEUVRE LES NANCY
Date of publication : Thursday, October 7, 2021
Scientific Responsible name : Julien Mercadier
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 1 December 2021
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly
Description of the thesis topic
Context: The University of Lorraine and Orano Mining are working together under the GeomIn 3D Industrial Chair, sponsored by Orano and the French Agence Nationale de la Recherche, to develop new tools for the exploration of uranium deposits in the Athabasca Basin. The research strategy aims to establish a 3D Common Earth Model applicable in exploration and reconciling geological and geophysical approaches.
Mining companies working on the Athabasca Basin acquire during their annual exploration campaign a large amount of geo-localized data of various kinds (mineralogical, chemical and physical) from geophysical surveys and drilling on different types of objects: sterile zone without hydrothermalism, hydrothermalized but non-mineralized zone, weakly mineralized zone and giant deposits. The generated databases include thousands of marker data from these different environments.
Research plan: The project aims at analyzing these multivariate and georeferenced databases resulting from exploration and development programs via an operating mode using Machine Learning. This work should result in a representation of the imprints of the mineralizing processes of giant deposits directly transferable to exploration teams.The work will be organized in several phases aimed at providing proof of the operability of this multivariate analysis:
- State of the art. Since the dissemination of Machine Learning methods, a lot of test and case studies have been performed concerning different georessources and using a wide range of modelling tools and algorithms. In Canada, the Footprint consortium to which Orano was involved in from 2013 to 2018 has produced significant contributions that already provide examples of multivariate analysis focused on the McArthur River-Millennium area in the Athabasca Basin for uranium, on the Malartic mine for gold and Guichon Creek batholith region for copper.
- Integration and analysis of geophysical, geochemical and geological databases in a 3D Common Earth Model. The variables quantified continuously or at regular intervals are related to lithology, structures, spectral signature, radiometry, petrophysics, geochemistry of major and trace elements. It is proposed to analyze these variables and their signatures at the scale of the Athabasca basin in order to characterize the diagenetic and hydrothermal histories associated to the mineralizing phenomena and at the scale of an exploration project which is home to the giant Cigar Lake deposit. The information extracted from this analysis will provide knowledge and information that will be processed and feed into a library on the properties will be used to constrain geophysical modelling.
- Multivariate analysis of data sets using geostatistical methods and Machine Learning. Through an exploratory analysis of the data sets, by process of iterative tests and failures, the various inter-correlations between parameters will be established. This analysis will be done by supervised learning, aiming to detect the fingerprints already identified by factorial methods, essentially Principal Component Analysis and Multiple Correspondence Analysis, and by statistical learning, unsupervised (e.g. K-Mean, hierarchical ascending classification) and supervised (e.g. neural networks, Random Forest…). The results will be evaluated to establish potential axial variabilities for different areas of the project.
- Implicit modelling of the representation of physical properties. Synthetic logs and implicit models of the distribution of physical properties will be constructed. The realization of a pilot on the Waterbury Cigar Lake exploration project will be declined in the format of procedure in order to be reproducible. It will serve as a reference model for comparison with the model resulting from direct and inverse modelling of geophysical data.
The Ph.D. thesis will be performed at the University of Lorraine, Georessources. Missions to the Orano headquarters in Châtillon and to Orano Canada in Saskatoon will be carried out to ensure coordination with ongoing research actions on Machine Learning led by Orano experts and specialists. The data integration platform developed as part of the Footprint project (https://mirageoscience.com/mining-industry-software/geoscience-integrator/) will be used for the construction of the 3D Common Earth Model and the application of Machine Learning methods.
GeoRessources is part of the Université de Lorraine (http://welcome.univ-lorraine.fr/), which is one of the leading institutions for higher education in Europe with more than 55,000 students and 60 research laboratories. Université de Lorraine was ranking 13th in the 2021 Shanghai Ranking for the category “Mining and mineral engineering”, first European university in this field of activity. The department of Earth Sciences is one of the most important in Europe, with 4 laboratories hosting more than 300 researchers and 1000 students. Georessources is the French reference academic laboratory for the study of ore deposits, with almost 200 people (http://georessources.univ-lorraine.fr/), with long-term collaborations with major mining and petroleum companies. It is considered one of the reference centers for academic research on uranium deposits, with more than 40 years of research on this topic
Constraints and risks
We talk about it on Twitter!