基于嵌入式多项式混沌算法的移动电话比吸收率SAR的不确定性分析研究

Specific absorption rate (SAR) is the most important parameter for the safety assessment of the electromagnetic radiation of mobile phone. The numerical simulation method is one of the effective methods to obtain SAR, and it can overcome the defects of measurement which unable to obtain the electromagnetic energy distribution in different tissues of human body non-destructively. However, the computational model of the SAR of mobile phone is complex, the uncertainties in input parameters induce the uncertainty of SAR. Presenting the calculated result SAR without quantifying its uncertainty is not enough to give authoritative evaluation conclusions. This project will analyze the uncertainty in the calculation of the SAR of mobile phone：First of all, to avoid the curse of dimensionality, it will apply sensitivity analysis method to find out the important parameters of the computational model and then reduce the dimension of data; secondly, it will apply non-parametric estimation method to estimate the probability density function of the uncertain input parameters, and select the corresponding type of polynomials based on the prior knowledge, which instead of simply assuming the input parameters from one of standard distributions and will avoid inducing errors in the calculation on account of the incorrect assumption; finally, an intensive study of the polynomial form on the different FDTD absorption boundary will be implemented to calculate the uncertainty in the calculation of SAR by intrusive polynomial chaos methods, thus, an uncertainty analysis method with high efficiency and high accuracy is formed. It will provide an important reference for the formulation of electromagnetic radiation safety standards in China and the international industry.

比吸收率（SAR）是移动电话电磁辐射安全评估最重要的参数。数值模拟方法是获得SAR的有效途径之一，它可以弥补实际测量不能无损地得到人体不同组织电磁能量分布等缺陷。然而，SAR的计算模型复杂，不确定的输入参数导致SAR的不确定性，只计算SAR而不量化其不确定性，不足以给出权威的评估结论。本项目针对移动电话SAR数值模拟计算中的不确定性进行分析：首先，通过敏感度分析筛选出模型中的重要参数，降低数据维数，避免维数灾难；其次，采用非参数估计得到输入参数的概率密度函数从而选择对应的多项式类型，避免因简单地假设输入参数服从某一标准分布，假设不成立而导致计算错误；最后，深入研究时域有限差分法（FDTD）吸收边界处多项式展开形式，应用嵌入式多项式混沌方法计算SAR的不确定度，从而形成一套效率高、准确率高的不确定性分析方法。将为我国以及国际电磁辐射行业安全标准的制定提供重要的参考依据。

本项目针对比吸收率（SAR）数值模拟计算中的不确定性进行分析：首先，建立了基于自动编码神经网络的模型，用于数据降维，避免维数灾难；其次，有效地结合辅助微分方程时域有限差分法（Auxiliary Differential Equation Finite-Difference Time-Domain, ADE-FDTD）与广义多项式混沌算法（Generalised Polynomial Chaos, gPCE）形成嵌入式多项式混沌算法，各向异性完全匹配层（Uniaxial Perfectly Matched Layer, UPML）作为吸收边界条件, 应用该方法对数值模拟结果进行不确定性分析，并将其结果与传统蒙特卡洛方法对比，该方法计算效率更高，显著降低计算成本。本项目形成了一套效率高、准确率高的不确定性分析方法。在项目研究过程中，发表期刊论文4篇，申请专利3项，培养硕士研究生4名。