A dataset for defect monitoring in additive manufacturing, China

The dataset includes 14 sets of samples with fixed process parameters and 5 sets of samples with variable process parameters. For the image data collected from each sample, redundant black images (caused by the laser being turned off) were first removed. Then, using a sliding time-window strategy, local regions were treated as a whole, and the melt pool images within each region were aggregated into a single sample. Each melt pool image has a resolution of 224×224 pixels, and the data is stored in .mat format. In total, 1,434 fixed-parameter samples and 491 variable-parameter samples were obtained.本数据集包含14组工艺参数固定的样本与5组工艺参数可变的样本。针对每一样本采集到的图像数据，首先移除因激光关闭产生的冗余黑色图像。随后采用滑动时间窗口策略，将局部区域视作整体，将每个区域内的熔池图像聚合为单个样本。每张熔池图像的分辨率为224×224像素，数据以.mat格式存储。最终共获得1434组参数固定样本与491组参数可变样本。

搜集汇总

背景与挑战

背景概述

该数据集是一个用于增材制造缺陷监测的专用数据集，包含固定和可变工艺参数下的熔池图像样本，总计1,925个样本（1,434个固定参数和491个可变参数），每个图像分辨率为224×224像素，以.mat格式存储。数据集采用滑动时间窗口策略进行预处理，专注于机械工程领域，数据量为8.00 GB，访问受限，仅限公共大学和研究机构用于学术研究。