Artificial neural network application for a predictive task in mining

Artificial intelligence research has produced several tools for commercial application. Some of the techniques that are widely used today include neural networks, fuzzy logic and expert systems. Artificial neural networks (ANNs) are excellent predictive, pattern recognition and data analysis tools. In the mining industry, ANN techniques are being used commercially for real-time process-control applications. Modeling of spatial data, ore-reserve estimation, tunnel design, long wall-stability prediction and geologic roof classification are additional applications in which neural networks have been applied successfully. In this study, a standard back propagation algorithm was used to train a series of neural networks for a real-world predictive task. After training and optimizing the neural network architecture, the performance of the network is measured on an independent validation set. Results indicate a mean error of less than 1 % between the actual and predicted values. A neural network model was developed for learning the spatial continuity of a mineral field and, consequently, for predicting sulfur values for given coordinates. The neural network not only performed satisfactorily, but in some cases performed even better than the kriging model.