Application of Multiple Linear Regression and Multilayer Perceptron Neural Network in Predicting Shale Content for Uranium Logging
The determination of shale content in uranium exploration is crucial for identifying the properties of underground rock formations and the distribution of sandstone-type uranium deposits.This study aims to overcome the limitations of traditional logging interpretation methods,which are affected by the accuracy of the selected GCUR coefficient,by proposing the use of multiple linear regression(MLR)models and multilayer perceptron(MLP)neural networks for analyzing and predicting logging data.By selecting logging data from a specific region,we constructed and validated shale content relationship models using MLR and MLP neural networks.The results indicate that the MLR model exhibits overfitting in low shale content intervals,whereas the MLP neural network demonstrates higher prediction accuracy.The MLP neural network outperforms the traditional MLR model in predicting shale content,providing an effective tool for accurate shale content prediction in uranium exploration.This approach promises to improve existing shale content evaluation methods significantly,enhancing the efficiency and accuracy of logging interpretation,and positively impacting subsequent uranium exploration and development activities.
uranium loggingshale contentmultivariate linear regression modelmultilayer perceptron neural network
万方数据
维普
