A Framework for the Optimal Selection of High-Throughput Data Collection Workflows by Autonomous Experimentation Systems
A Framework for the Optimal Selection of High-Throughput Data Collection Workflows by Autonomous Experimentation Systems
复制标题
DOI：
10.1007/s40192-022-00280-5
复制
发表时间：
2022-10-31
影响因子：
3.3
通讯作者：
Niezgoda, Stephen R.
中科院分区：
文献类型：
--
作者：
Casukhela, Rohan;Vijayan, Sriram;Niezgoda, Stephen R.
自主实验系统已被用来大大推进集成计算材料工程范式。本文概述了一个框架，使自主实验系统的数据收集工作流程的设计和选择。该框架首先搜索生成高质量信息的数据收集工作流，然后根据用户定义的目标选择生成最高价值信息的工作流。我们采用该框架选择最佳的高通量工作流程，使用基于深度学习的图像去噪器对增材制造的Ti-6Al-4V样品进行表征。所选择的工作流程减少了收集时间的背散射电子扫描电子显微镜图像的5倍的因素相比，案例研究的基准工作流程，并通过85倍的因素相比，在以前发表的研究中使用的工作流程。
Autonomous experimentation systems have been used to greatly advance the Integrated Computational Materials Engineering paradigm. This paper outlines a framework that enables the design and selection of data collection workflows for autonomous experimentation systems. The framework first searches for data collection workflows that generate high-quality information and then selects the workflow that generates the highest-value information as per a user-defined objective. We employ this framework to select the optimal high-throughput workflow for the characterization of an additively manufactured Ti-6Al-4V sample using a deep-learning based image denoiser. The selected workflow reduced the collection time of backscattered electron scanning electron microscopy images by a factor of 5 times as compared to the case study's benchmark workflow, and by a factor of 85 times as compared to the workflow used in a previously published study.