MCSIMus: Monte Carlo Simulation with Inline Multiphysics

在净零的背景下，各种形式的核反应堆日益突出。然而，严格的安全标准和先进的反应堆设计要求对反应堆物理和运行有更大的确定性和了解。随着物理实验变得越来越昂贵，核工程越来越依赖于反应堆的高保真模拟。传统上，由于计算能力的限制，解决反应堆中不同的物理现象(如中子传输或热-水力学)时，只假定对其他现象的弱依赖性。当反应堆设计中包含额外的保守性时，这种近似是允许的。然而，更经济或更复杂的反应堆设计使这种近似无效，反应堆设计者必须能够解决每一种物理现象之间的相互作用。这对反应堆物理学家提出了挑战，因为多物理计算的计算成本大大增加，以及数值不稳定性的风险--这些本质上是纯粹模拟的非物理行为。这一提议旨在为核工程中新的计算方法提供基础，这些方法比目前的多物理方法更便宜、更稳定。传统方法倾向于让一个工具完全解决一个现象，将信息传递给解决第二个现象的另一个工具，然后将更新后的信息传递回第一个工具并重复，直到(希望)结果收敛。这项提议希望探索一种稍微简单一些的方法，在不同的求解器之间交换信息，然后每个求解器都完全解决自己的物理问题，将这一方法扩展到反应堆设计者感兴趣的许多现象。初步分析表明，这种方法应该比以前的方法更稳定，计算效率更高。这些研究将使用剑桥大学自行开发的数字工具进行，这些工具旨在为新想法和算法的快速原型设计。预计最终结果将改变核工业对多物理计算的方法，并极大地提高我们探索和设计更先进核反应堆的能力。

Nuclear reactors in various forms are increasingly prominent in the context of net zero. However, stringent safety standards and advanced reactor designs necessitate ever-greater certainty and understanding in reactor physics and operation. As physical experimentation becomes more expensive, nuclear engineering relies increasingly on high-fidelity simulation of reactors. Traditionally, resolving different physical phenomena in a reactor (such as neutron transport or thermal-hydraulics) proceeded by assuming only a weak dependence upon other phenomena due to limits on computational power. Such approximations were allowable when additional conservatisms were included in reactor designs. However, more economical or sophisticated reactor designs render such approximations invalid, and reactor designers must be able to resolve the interplay between each physical phenomenon. This poses a challenge to reactor physicists due to vastly increased computational costs of multi-physics calculations, as well as the risks of numerical instabilities - these are essentially non-physical behaviours which are purely an artefact of simulation.This proposal aims to provide the basis of new computational approaches in nuclear engineering which are both substantially cheaper and more stable than present multi-physics approaches. Traditional methods tend to have one tool fully resolve one phenomenon, pass the information to another tool which resolves a second phenomenon, and then pass this updated information back to the first tool and repeat until (hopefully) the results converge. This proposal hopes to explore a slightly simpler approach, where information is exchanged between different solvers before each has fully resolved its own physics, extending this to many of the phenomena of interest to a reactor designer. Preliminary analysis suggests that this approach should be vastly more stable and computationally efficient than previous methods. The investigations will be carried out using home-grown numerical tools developed at the University of Cambridge which are designed for rapid prototyping of new ideas and algorithms. The final result is anticipated to transform the nuclear industry's approach to multi-physics calculations and greatly accelerate our ability to explore and design more advanced nuclear reactors.