The scientific fields of additive manufacturing and especially the extrusion-based technologies have gained immense attention in numerous commercial and research areas in the past decades. However, monitoring the manufacturing procedure and detecting errors during the process remain a technological challenge in the field. Generally, defect detection and dimensional accuracy inspection of the produced component is applied after the manufacturing has been completed and is accomplished via on-site manual monitoring. Hereupon, these approaches could affect the manufacturing production cost via the increase of feedstock material, waste parts, manpower, and machine rates. To overcome these issues, the present paper introduces a vision-based method to scan, filter, segment, and correlate in real-time the physical printed part with the digital 3D model as well as to evaluate the performance of the additive manufacturing process. More specifically, high-resolution point cloud data of the printed part are automatically captured, filtered, segmented, reconstructed, and compared with the corresponding digital 3D model in various stages of the procedure. Finally, the effectiveness of the suggested automatic monitoring and error detection methodology is experimentally validated.

在过去的几十年里，增材制造的科学领域，尤其是基于挤压的技术，在众多商业和研究领域得到了极大的关注。然而，监控制造过程和检测过程中的错误仍然是该领域的技术挑战。通常，在制造完成后，对生产的部件进行缺陷检测和尺寸精度检查，并通过现场人工监控完成。因此，这些方法可能会通过增加原料、废料、人力和机器费率来影响制造生产成本。为了克服这些问题，本文介绍了一种基于视觉的方法来扫描、过滤、分割、并将物理打印部件与数字 3D 模型实时关联，并评估增材制造过程的性能。更具体地说，打印部件的高分辨率点云数据在程序的各个阶段被自动捕获、过滤、分割、重建，并与相应的数字 3D 模型进行比较。最后，通过实验验证了建议的自动监控和错误检测方法的有效性。