Abstract The Low Rank Approximation (LRA) and Unscented Transform (UT) are integrated to produce a new algorithm having the capability to decrease the time required for the uncertainty quantification during Loss of coolant accident (LOCA) in Pressurized Water Reactors (PWR). The LRA is an efficient technique used in reducing computational cost due to its ability to perform dimensionality reduction by revealing the active or important degrees of freedom and calculate the basis of the so-called active subspace basing on the Singular Value Decomposition (SVD). For further reduction in the computational time; the UT algorithm is also implemented to generate a set of sigma points, these sigma points are the representatives of the whole probability distribution (the UT is restricted to Gaussian distribution). The main safety parameter is the maximum cladding temperatures during the accident which are computed by ATHLET thermal-hydraulic code. The reactivity coefficients and the covariance matrix are calculated using the SCALE 6.2 code. The present calculation model has 14-dimensions, therefore the number of sigma points needed for the SVD/UT technique is 29, and can be minimized to 5 sigma points only if the LRA/UT is used where two singular values are sufficient to reproduce/span the space thanks to the strong correlations between the reactivity coefficients.

中文翻译：

---

冷却剂大断口损失事故不确定性分析中低秩逼近无迹变换的实现

摘要 低秩逼近 (LRA) 和无迹变换 (UT) 相结合，以产生一种新算法，该算法能够减少压水反应堆 (PWR) 中冷却剂损失 (LOCA) 期间不确定性量化所需的时间。LRA 是一种用于降低计算成本的有效技术，因为它能够通过揭示活动或重要的自由度来执行降维，并基于奇异值分解 (SVD) 计算所谓的活动子空间的基础。进一步减少计算时间；UT 算法也被实现来生成一组 sigma 点，这些 sigma 点是整个概率分布的代表（UT 仅限于高斯分布）。主要安全参数是由ATHLET热工水力代码计算出的事故发生时的最高包层温度。反应系数和协方差矩阵使用 SCALE 6.2 代码计算。目前的计算模型有 14 维，因此 SVD/UT 技术所需的 sigma 点数为 29，并且只有在两个奇异值足以重现/由于反应系数之间的强相关性，跨越空间。