Application Of Combined Algorithm In Identification Of Rock Mineral Types In Xiangshan Uranium Resources

Uranium mineral resources is not only an important part of natural resources,but also the survival of human society and the basis for material development.Its development and utilization in promoting the country's prosperity,regional economic growth,national economic security plays an important role in the process.At present,the ore deposits in China are dominated by middle and low grade,showing the phenomenon of less ore deposits in poor ore deposits,and the factors that affect the ore mineralization are also complicated.This not only brings difficulties to the mineral identification work,The development and utilization of mineral resources has affected.In order to improve the theoretical research and practical application of rock mineralization identification,it is important to study the classification of rock mineralization from the quantitative point of view by using mathematical algorithm to establish the combined model.In this paper,the following mathematical models were established and analyzed by using three deposits,55 samples and 20 trace element variables in Xiangshan uranium mineral resources as the analysis samples.First,20 variables were used to reduce the dimension of 20 trace element variables,and 14 variables with large influence on the three types of rocks(namely,rich ore,lean ore or alteration rock and surrounding rock)were found.: U,Th,Mo,Yb,Y,P,Sr,Pb,Zr,K,Zn,Sm,Rb and Ta,and establish a discriminant analysis model.In this paper,the first 45 samples were selected as the training samples and the last 10 samples were used as the test samples to obtain the experimental results.The results show that the classification accuracy is improved from 70% before the dimension reduction process to 80% after dimension reduction,and the model effect is better than that before dimension reduction.Second,the genetic algorithm based on the neural network combination model(GA-BP)on the rock mineralization to identify applications.U,Sr,Th,Ce,Hf,Mo,Zn,U,Pb were selected and the GA-BP model was established by using the variable selection method to select the eight variables which had the greatest influence on rock mineralization classification.The results show that the classification accuracy is improved from 70% to 90% after dimensionality,and the model effect is improved to a certain extent.Thirdly,the support vector machine model(PCA-SVM)based on principal component is used to identify the rock mineralization.Because of the complex correlation between the mineral composition of the rock,the six principal components with cumulative contribution rate greater than 85% are selected by PCA as input variables of SVM.The main parameters c and g of SVM are optimized by grid search(GS),particle swarm optimization(PSO)and genetic algorithm(GA).The results show that the accuracy of the classification is improved by using the GAA-GAM SVM model,and the accuracy of the PCA-GA-SVM model is better than that of the PCA-GA-SVM model.100%.The research and application of several combination model algorithms in the identification of rock mineralization in uranium deposits provide a new way and method to explore the problem of quantitative identification of rock classification in geological disciplines.