基于多项式混沌展开法的涡流无损检测高效元模型辅助探测概率的分析
Anglysis of model-assisted probability of detection for eddy current nondestructive testing based on efficient polynomial chaos expansion metamodel
-
摘要: 探测概率对于量化涡流无损检测系统的检测能力非常重要。模型辅助探测概率的参数需要大量试验或仿真数据才可以确定,往往难以实现。应用基于退化核函数加速的边界元法的数值模型,提出使用基于普通最小二乘法的多项式混沌展开算法的元模型提升三维涡流无损检测问题探测概率的效率。通过有限截面线圈检测金属板面槽的算例,引入线圈位置和提离距离为不确定传播的参数,测试结果表明,该元模型预测的模型参数与基于退化核函数加速的边界元法物理模型计算的模型参数相对误差在1 %以内,极大降低了所需的计算开销。Abstract: Probability of detection (PoD) is very important to quantify the detection ability of eddy current nondestructive testing (NDT) system. In the study of Model Assisted PoD (MAPoD), large amounts of data are needed to accurately determine the model parameters which is time consuming and hard to achieve experimentally or by simulation. To overcome this issue, in this paper, the ordinary least square based polynomial chaos expansion (OLS PCE) metamodel was proposed, to improve the efficiency of the MAPoD study for 3D eddy current NDT problem with the application of kernel degeneration (KD). The case of placing a coil with a finite cross section above the thick plate with a surface slot with selecting the coil position and liftoff as the uncertainty propagation parameters were tested. The results showed that the differences of PoD parameters between the ones predicted by the OLS PCE metamodel and calculated by the KD-BEM based physical model were smaller than 1 % with less time required.