Simulation And Optimization Of Copper Mine Grade Classification Device Based On PGNAA Technology And Research On Neural Network Spectral Solution

Ore is a non renewable resource and an important material foundation for human survival and development.The existing methods for analyzing copper ore,such as chemical analysis and atomic spectroscopy,do not have the ability to perform bulk analysis.The prompt gamma neutron activation analysis technology has the advantages of nondestructive,non-contact,volume analysis,and no impact on the dust environment,and has been successfully applied in the fields of cement,coal,and so on.However,this technology is still in the blank in metal ore detection.Based on the instant gamma neutron activation analysis technology,this paper uses the Monte Carlo method to optimize the design of the detection device,uses the neural network algorithm to analyze the gamma spectrum obtained,and finally realizes the grade classification of copper ore and the identification of associated beneficial components in copper ore,which provides a reference for the application of instant gamma neutron activation analysis technology to copper ore detection.The main work of this paper:(1)A copper ore grade classification system based on prompt gamma neutron activation analysis is established through MCNP simulation software,and the height,width,moderator,reflector and shield of the device are optimized.(2)Analyze the basic situation of Jiangxi copper ore,construct multiple samples with element content similar to real copper ore by combining real measurement data of various ores and gangue with different contents,establish a dataset for machine learning training based on MCNP simulation,and determine the two tasks of copper ore grade classification and associated beneficial component identification.(3)Use multiple machine learning methods to analyze the gamma spectrum of copper ore.A two-dimensional energy spectrum method based on piecewise Gram angle field transformation is proposed,which converts the one-dimensional energy spectrum into a twodimensional matrix that retains track address information and counting information.Combined with Res Net neural network,the accuracy rate of copper ore grade classification task on the simulation data set is more than 94%,and the accuracy rate of associated beneficial element recognition task is more than 93%,which is the best among all machine learning methods tried.(4)Build an experimental platform for actual measurement of copper ore.The accuracy of the two proposed neural networks in analyzing the measured gamma spectrum has decreased,and the reason has been analyzed and the accuracy has been improved by supplementing training data.