Scholarlink

In the constantly evolving field of additive manufacturing, it is crucial to ensure the quality of 3D-printed parts to guarantee their reliability and performance. This paper presents a new technique for using a convolutional autoencoder to identify visual anomalies in 3D-printed components. The technique was used to understand the essential visual features of high-quality prints by training it on a dataset comprising flawless and defected photos. After training, the model was evaluated on an independent set of images, revealing considerable reconstruction errors in the samples with defects, but enabling defect-free parts to be reconstructed with few deviations. The produced anomaly maps showed regions with notable variations, suggesting possible flaws. The findings of this study open the door for better quality control procedures in additive manufacturing processes by showcasing the potential of convoluted autoencoders as an efficient automated tool for evaluating 3D quality.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Developing an Artificial Intelligence Model for Automatic Error Detection in 3D Prints Using Photos of Correct and Incorrect Prints

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sara Samine,
  • Mohamed Karouchi,
  • Maria Zemzami,
  • Mohamed Gouskir,
  • Nabil Hmina,
  • Manuel Lagache,
  • Soufiane Belhouideg

摘要

In the constantly evolving field of additive manufacturing, it is crucial to ensure the quality of 3D-printed parts to guarantee their reliability and performance. This paper presents a new technique for using a convolutional autoencoder to identify visual anomalies in 3D-printed components. The technique was used to understand the essential visual features of high-quality prints by training it on a dataset comprising flawless and defected photos. After training, the model was evaluated on an independent set of images, revealing considerable reconstruction errors in the samples with defects, but enabling defect-free parts to be reconstructed with few deviations. The produced anomaly maps showed regions with notable variations, suggesting possible flaws. The findings of this study open the door for better quality control procedures in additive manufacturing processes by showcasing the potential of convoluted autoencoders as an efficient automated tool for evaluating 3D quality.