铀在花岗岩中吸附性能预测仿真研究

Research on Prediction Adsorption Capacity of Nuclide Uranium in Granite Rocks

针对高放废物处置库预选场花岗岩对铀的吸附性能受到温度、PH值、铀浓度等因素的影响巨大,具有明显的非线性和不确定特性,传统的数理统计预测方法很难得到有效的预测结果。对铀在花岗岩中吸附性能预测问题,提出了人工神经网络预测方法,对处理预测问题具有很好的泛化、学习和映射能力。以铀在花岗岩中吸附性能的影响因素作为神经网络的输入信号,花岗岩对铀的吸附比作为输出信号,建立神经网络预测算法,并用Microsoft Visual Studio为开发工具,研制了"核素吸附仿真软件系统"。以铀在处置库预选场花岗岩中的吸附实验数据为研究对象,用所研制的软件进行了铀在花岗岩中吸附性能预测仿真。仿真结果表明,预测效果好、精确度高。将仿真模型与传统多元线性回归预测结果进行对比分析结果表明,人工神经网络预测结果具有更高的准确定和精确度。

The adsorption capacity of nuclide uranium in granite rocks is influenced greatly by temperature, concentration of uranium, sample particle diameter, PH value and other factors. A prediction method based on neural network method was proposed in this paper. The neural network method is of great capabilities of generalization, learning and mapping in dealing with the nonlinear issue, which is established by adsorption capacity of nuclide ura- nium in granite rocks（ as neural network output ） and its influence factors（ as neural network input ）. Furthermore, the simulation software system of the adsorption capacity of nuclide uranium in granite rocks has been developed using Microsoft Visual Studio as a developing instrument. Using the experimental data of adsorption capacity of nuclide ura- nium in granite rocks as study object, the adsorption capacity of nuclide uranium in granite rocks prediction in future has been realized successfully based on the simulation software system. The simulation results show that this method has good effect and high precision. The comparison and analysis of predicted results of neural network method and multiple linear regressions method show that the prediction result of neural network method has higher accuracy and precision.