发表机构
RWTH Aachen University; Institute of Physical Metallurgy and Materials Physics (IMM)（亚琛工业大学; 物理冶金与材料物理研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
EBSDmagus通过自动化多阶段动态EBSD模拟流程，实现参数定义、计算复用、结果验证与中断恢复，并提供完整来源追踪，支持大规模FAIR数据研究。
AI 中文摘要
动态电子背散射衍射（EBSD）模拟日益支持对材料及实验参数如何影响衍射图案的系统性研究。将成分与晶体结构（包括原子有序度变化）同加速电压、温度以及多种取向或探测器条件相结合的研究，可能需要大量相关的计算集合。多个软件包提供动态EBSD模拟；我们使用开源的、可脚本化的EMsoft程序，因为它们非常适合在高性能计算系统上进行大规模研究。然而，一项EMsoft研究跨越了独立的配置、结构、散射和取向文件以及多个相互依赖的计算阶段。随着变体的增多，手动准备变得难以检查，中断使得哪些结果仍然可用变得模糊不清，单个图案的来源也变得难以重建。借助EBSDmagus，研究人员只需定义一次固定和变化的参数。该软件准备所需的EMsoft计算，重用兼容的中间结果，并在执行前检查生成的文件是否代表所请求的研究。执行后，它检查预期的输出，并在中断后仅恢复未解决的计算，同时保留已完成的工作。一个可移植的运行记录将每个输出与其输入和计算历史关联起来，提供了在结果作为可查找、可访问、可互操作和可重用（FAIR）数据存储时所需的来源信息。我们通过一个已完成的代表性计算、一个在提交前准备并检查的较大研究、两次观察到的集群中断以及一次受控作业取消来评估这种方法。在恢复过程中，EBSDmagus保留了已完成的计算，并仅重新提交了未完成的工作。
英文摘要
Dynamical electron backscatter diffraction (EBSD) simulations increasingly support systematic studies of how material and experimental parameters affect diffraction patterns. Investigations that combine composition and crystal structure, including changes in atomic order, with accelerating voltage, temperature, and many orientations or detector conditions can require large sets of related calculations. Several packages provide dynamical EBSD simulation; we use the open-source, scriptable EMsoft programs because they are well suited to large studies on high-performance computing systems. An EMsoft investigation nevertheless spans separate configuration, structure, scattering, and orientation files and several dependent calculation stages. As variations multiply, manual preparation becomes difficult to check, interruptions obscure which results remain usable, and the origin of individual patterns becomes laborious to reconstruct. With EBSDmagus, researchers define fixed and varying parameters once. The software prepares the required EMsoft calculations, reuses compatible intermediate results, and checks that the generated files represent the requested investigation before execution. After execution, it checks the expected outputs and, following an interruption, resumes only unresolved calculations while preserving completed work. A portable run record links each output to its inputs and calculation history, providing the provenance needed when the results are deposited as findable, accessible, interoperable, and reusable (FAIR) data. We assess this approach using a completed representative calculation, a larger study prepared and checked before submission, two observed cluster interruptions, and a controlled job cancellation. During recovery, EBSDmagus retained completed calculations and resubmitted only incomplete work.
