A total of 1300 km of drill core scan data and 130,000 corresponding chemical analyses have been used to train the Artificial Intelligence.
Structured quality controlled database with all data as basis for training of the neural network.
A purpose-built and trained Convoluational Neural Network – 60% training data, 20% verification data and 20% test data – Accuracy between 85% and 95%
This is what we knitted together when applying Artificial Intelligence to mass data that is acquired during hyperspectral drill core scans.
Hyperspectral scans of drill cores produce massive amounts of data that go into customized databases. Analyzing these databases and consolidating the information, a special neural network was developed. This then processes the information into reliable quantitative predictions and forecast models of valuable metals.
This procedure enables accuracies of up to 85-95% which is unsurpassed by previous interpretation methods.
The predictor can be used to analyse your drill cores in your daily practice on the mine face and also for your large scale green- and brownfield exploration as well as for airbourne/UAV accquired data.
Results are showing a significantly better performance of the CoreSmart neural network compared to classical prediction methods.
The fully trained Predictor for Metal Grades - with Artificial Intelligence to a significant improvement of quality in hyperspectral data analysis. The CoreSmart Predictor improves productivity in interpretation, lowers costs and makes forecasts much more reliable.
