Abstract The fission matrix based radiation transport code RAPID is able to perform high fidelity and fast 3D, pin-wise whole core calculations. The code can quickly estimate the system fission matrix by combining pre-calculated database fission matrices, and solving for the system fission matrix provides the multiplication factor and a detailed fission distribution. In this work, RAPID acceleration and numerical techniques will be examined. These are the fission matrix collapsing or homogenization options, choice of the power iteration tolerance and estimation of iterative error. In old RAPID calculations, in order to achieve a fast power iteration convergence of the whole core fission matrix, the pin-wise fission matrix is collapsed into an assembly-wise one following a 2-D pin-wise core slice calculation. In addition to the radial collapsing from pin to assembly, other radial and axial collapsing options are tested to find out the best balance between accuracy and speedup. Applying the fission matrix collapsing technique, RAPID requires at least two eigenvalue calculations. Different tolerances of the 2-D and 3-D power iteration calculations are combined to investigate the convergence. A more sophisticated method to estimate the iterative error is provided to determine the convergence. Overall, a combination of all acceleration techniques speeds up the RAPID calculation on BEAVRS benchmark problem by three times, with only a negligible loss of accuracy.

中文翻译：

---

RAPID 中的裂变矩阵均匀化和迭代收敛

摘要 基于裂变矩阵的辐射传输码 RAPID 能够执行高保真度和快速的 3D、pin-wise 全核计算。该代码可以通过结合预先计算的数据库裂变矩阵来快速估计系统裂变矩阵，求解系统裂变矩阵提供倍增因子和详细的裂变分布。在这项工作中，将研究 RAPID 加速和数值技术。这些是裂变矩阵折叠或均匀化选项、功率迭代容差的选择和迭代误差的估计。在旧的 RAPID 计算中，为了实现整个核心裂变矩阵的快速功率迭代收敛，在 2-D 引脚方向核心切片计算之后，引脚方式裂变矩阵被折叠成组装方式。除了从销到组件的径向收缩外，还测试了其他径向和轴向收缩选项，以找出精度和加速之间的最佳平衡。应用裂变矩阵折叠技术，RAPID 需要至少两次特征值计算。结合 2-D 和 3-D 幂迭代计算的不同容差来研究收敛性。提供了一种更复杂的方法来估计迭代误差以确定收敛。总体而言，所有加速技术的组合将 BEAVRS 基准问题的 RAPID 计算速度提高了三倍，而精度损失可以忽略不计。RAPID 至少需要两次特征值计算。结合 2-D 和 3-D 幂迭代计算的不同容差来研究收敛性。提供了一种更复杂的方法来估计迭代误差以确定收敛。总体而言，所有加速技术的组合将 BEAVRS 基准问题的 RAPID 计算速度提高了三倍，而精度损失可以忽略不计。RAPID 至少需要两次特征值计算。结合 2-D 和 3-D 幂迭代计算的不同容差来研究收敛性。提供了一种更复杂的方法来估计迭代误差以确定收敛。总体而言，所有加速技术的组合将 BEAVRS 基准问题的 RAPID 计算速度提高了三倍，而精度损失可以忽略不计。