NSF Future Manufacturing Data Challenge: A Multimodal DED Dataset for Probabilistic Representation and Prediction of Laser-Track Geometry
收藏资源简介:
本数据集是由德克萨斯农工大学为美国国家科学基金会未来制造数据挑战赛创建的多模态定向能量沉积(DED)数据集,旨在支持激光轨迹局部几何变异的概率性预测研究。该数据集包含四种激光功率条件下在316L不锈钢基板上生成的单道扫描数据,核心由三种互补模态构成:来自Stratonics ThermaViz熔池传感器的原位热图像序列、利用蔡司EVO MA10系统获取的扫描电子显微镜(SEM)图像,以及通过布鲁克ContourGT-K白光3D光学轮廓仪采集的全场高度图,数据总量涵盖四个独立实验轨道,并提供了统一的20-100毫米物理坐标对齐框架。数据集通过精密实验平台采集,专注于剥离粉末流动影响,以探究激光-材料相互作用和重凝固行为,主要应用于先进制造领域,特别是增材制造的过程监控、质量预测与智能化控制,致力于解决激光沉积过程中局部几何特征随工艺参数和基板形态动态演变的量化建模难题。
This multimodal Directed Energy Deposition (DED) dataset was developed by Texas A&M University for the U.S. National Science Foundation Future Manufacturing Data Challenge, aiming to support research on probabilistic prediction of local geometric variations in laser trajectories. The dataset contains four single-pass scanning datasets generated on 316L stainless steel substrates under four laser power conditions. Its core is composed of three complementary modalities: in-situ thermal image sequences from the Stratonics ThermaViz molten pool sensor, scanning electron microscope (SEM) images acquired using the Zeiss EVO MA10 system, and full-field height maps collected via the Bruker ContourGT-K white-light 3D optical profilometer. The full dataset covers four independent experimental runs, and provides a unified physical coordinate alignment framework spanning 20–100 mm. Collected through a precision experimental platform, this dataset is designed to isolate the effects of powder flow to explore laser-material interactions and re-solidification behavior. It is mainly applied in the field of advanced manufacturing, particularly for process monitoring, quality prediction and intelligent control of additive manufacturing, and is dedicated to solving the quantitative modeling challenge of the dynamic evolution of local geometric features during laser deposition driven by process parameters and substrate morphology.




