Artificial Intelligence is Helping Scientists Discover Mineral Resources

2023-07-13

Finding economically valuable mineral deposits requires a significant amount of resources. Scientists are turning to artificial intelligence to solve this problem. From lithium for electric vehicle batteries to iron for constructing new buildings, we have always relied on various hidden minerals in the Earth. The process of finding these minerals is resource-intensive, often involving satellite imagery, geochemical surveys, and knowledge gained from past mining projects. While it is sometimes possible to discover abundant natural resources, finding large mineral deposits is not easy. Therefore, scientists are turning to artificial intelligence to increase the chances of success. Researchers from the University of Notre Dame, the University of Arizona, and the Carnegie Science Institute have developed a model that predicts the locations of rare or valuable minerals by analyzing mineralogical data. The model is based on a mineral evolution database, which stores information on 5,477 types of minerals and their occurrences in 295,583 locations. As they wrote in their paper in PNAS Nexus, the vast amount of data, combined with the "complexity and inherent 'messiness' of our planet's geological, chemical, and biological systems," makes it extremely challenging for humans to use the mineral evolution database to find resource deposits. However, artificial intelligence can quickly evaluate this data. The key to the model lies in association analysis, which involves searching for patterns in the massive dataset. By matching the mineral evolution database with information on plate tectonics, atmospheric oxygenation, Earth's geospheres, and other relevant phenomena, the model identifies potential locations where minerals may occur and provides rough estimates of the quantities hidden there. The research team tested their model in October 2020, allowing it to determine the locations of deposits of rutherfordine, andersonite, schröckingerite, bayleyite, and zippeite. To improve efficiency, they set parameters to only display predictions with a "confidence" level of 70% or higher. The model provided predictions for four locations of rutherfordine, one of which has been confirmed in Italy; one prediction for andersonite, which remains unconfirmed; one confirmed prediction for schröckingerite in Colorado; two predictions for bayleyite, both of which were previously speculated in Utah; and seven predictions for zippeite, one of which has been confirmed in the Czech Republic. In another test concerning rare earth elements (REEs), the model predicted the locations of monazite, allanite, and spodumene, which are used in construction, radiation research, and batteries. Out of 15 predictions, 12 monazite deposits have been confirmed; out of 19 predictions, 13 allanite deposits have been confirmed; and out of 12 predictions, one spodumene deposit has been confirmed. The model is believed to have potential applications on other planets as well. With our understanding of the geological and astrobiological history of the Moon or Mars, the model may help locate minerals throughout the solar system. Source: https://www.extremetech.com/science/ai-is-helping-scientists-locate-rare-and-useful-minerals

RECENT AI NEWS

RECENT AI TOOLS