New method for evaluating drill cores
CORESMART workflow enables quantitative determination of valuable metals in drill cores - Publication of results in the Australian Journal of Earth Sciences
With the current publication in the international Australian Journal of Earth Sciences, G.E.O.S. and Dimap Spectral GmbH report on the latest results of AI-based prediction of grades of gold, copper and iron in drill cores based on hyperspectral drill core scans. The article "Performance measurements of machine learning and different neural network designs for prediction of geochemical properties based on hyperspectral core scans" demonstrates the performance of the newly developed method, which is now also offered commercially as a CORESMART workflow.
It was impressively demonstrated that it is possible to quantitatively determine valuable metals in drill cores with the AI-based CORESMART workflow developed by G.E.O.S. and Dimap. In the process, an accuracy of predictions of more than 80 % is achieved. This clearly exceeds the results of previous AI-based approaches. The excellent performance is achieved on the one hand by the neural network structure used and on the other hand by the extremely extensive training data of more than 700 kilometres of drill cores with more than 110,000 chemical analyses.
Thus, this method can be used to minimise the costs of chemical analyses in exploration as well as in the re-evaluation of existing drill cores. In the investigations, the CORESMART workflow was tested for the elements gold, copper, iron, tin, tungsten, silver, nickel and uranium with prediction accuracies of more than 80 %.
