摘要
在使用面源高斯扩散模型模拟铀尾矿库核素氡扩散浓度时,面源积分模型计算复杂,后退虚源模型模拟结果不准确且不具备保守性。针对上述问题,提出了基于信赖域算法的函数与参数寻优方法。通过构建近似后退虚源模型的非线性函数集,将面源积分模型与函数库中每个函数模拟值之间的误差作为目标函数,在信赖域策略下调用函数库并迭代寻优,输出函数库中最优函数与其最优参数值。通过浓度分布对比与拟合测验得出:优化函数在近源处的模拟值略大于面源积分模型,表明优化函数具备一定的保守性;模拟值与面源积分模型的拟合决定系数在近源处高于0.91,大部分区域高于0.98,说明其模拟结果可靠;同时,优化函数保留了后退虚源模型计算简便的特点。此外,根据误差的影响因素分析,优化函数对于面源面积和大气稳定度不变时,能适应其他参数的变化,大大简化了实际氡浓度评估时的计算过程。
When the area source Gaussian diffusion model is used to simulate the radon diffusion concentration of nuclides in uranium tailings ponds,the finite segment area source model is complex to calculate,and the simulation results of the virtual-point source model are neither accurate nor conservative.In response to the above problem,a function and parameter optimization method based on trust region algorithm has been proposed.Constructing a set of nonlinear functions that approximate virtual-point source model,and defining the error between simulated values of finite segment area source model and each function in the function library as the objective function.Then the function library will be called under the trust region strategy and iteratively optimized till the optimal function and its optimal parameter value are output.Through the comparison of concentration distribution and fitting test,it indicates that the simulated value near the source of the optimization function is slightly higher than the counterpart of finite segment area source model,which proves the optimization function partly has a conservatism;the goodness-of-fit near the source between the simulated values is higher than 0.91,and higher than 0.98 in most areas,showing that the simulation results are reliable.In addition,according to the analysis of the influencing factors of error,the optimization function can adapt to the changes of other parameters when the source area and atmospheric stability remain unchanged,which greatly simplifies the calculation process of the actual radon concentration assessment.
作者
丁玉恒
杨亚新
谢尚平
罗齐彬
吕文杰
黄培
王毅强
DING Yuheng;YANG Yaxin;XIE Shangping;LUO Qibin;LYU Wenjie;HUANG Pei;WANG Yiqiang(School of Nuclear Science and Engineering East China University of Technology,Nanchang 330013,China;School of Geophysics and Measurement-control Technology East China University of Technology,Nanchang 330013,China)
出处
《世界核地质科学》
CAS
2024年第2期386-395,共10页
World Nuclear Geoscience
基金
江西省教育厅落地计划项目(赣财教指[2012]211号)
江西省科技支撑计划项目(编号:20171BBG70033)联合资助。
关键词
高斯扩散模型
铀尾矿库
后退虚源模型
信赖域算法
Gaussian plume model
uranium tailings pond
virtual-point source model
trust region algorithm
