Our PhD student Alice published a paper and presented her research at NAMRC 52 at the Knoxville Convention Center (Knoxville, TN) this summer. She also visited Oak Ridge National Laboratory, one of the world’s premier research institutions, for advanced manufacturing tools.
Alice’s paper titled “Deep Learning of 3D Point Clouds for Detecting Geometric Defects in Gears” has been published at NAMRC 52, and will be published in Manufacturing Letters soon.
Geometric integrity ensures the functionality and safety of manufactured products, making it essential in quality control. Recent advancements in 3D metrology allow fine-scale inspections but face challenges such as high dimensionality, unstructured nature, and sparsity in defective regions. To address these challenges, this paper introduces the “MFGNet-gear” dataset with 12 designs and four quality classes each. It also presents a deep learning model adapted from PointNet++ for automated 3D point cloud analysis, achieving up to 100% accuracy in design classification and 85% in quality inspection. Additionally, a systematic investigation was conducted on the impacts of measurement resolution and precision on classification performance through a series of case studies. The results highlight deep learning’s potential for 3D point cloud analysis in quality control beyond gear manufacturing, suggesting future research in new architectures and adaptive measurement planning.