Recognition of geological rocks at the bedded-infiltration uranium fields by using neural networks

Interpretation geophysical data is one of the important factors affecting the economic indicators of mining process. The mining process depend on the speed and accuracy of geophysical data interpretation, but the process of logging data interpretation can not be strictly formalized. Therefore, computer interpretation methods on the basis of expert estimates are necessary. The method is based on expert opinion are widely used in weakly formalized tasks. Mention may be made of the system based on rules, fuzzy logic, Bayesian decision-making systems, artificial neural networks (ANN). ANN have already been used for solving a wide range of recognition problems. The paper analyzes the quality of network's data interpretation essentially depending on its configuration parameters, methods of data preprocessing and learning samples. About 2000 calculation experiments have been made, software and templates for preprocessing of data and interpretation findings have been developed. These experiments showed the effectiveness of neural network approach to solving the problem of geological rocks recognition in stratum-infiltration uranium deposits. Further research in this area will raise the recognition process automation and its accuracy.