CenterNet-based defect detection for additive manufacturing

作者：

Highlights：

• Surface defects on Additive Manufactured (AM) workpiece are characterized.

• A novel CenterNet-based defect detection model is developed for AM workpiece.

• Density map and count loss are merged into CenterNet-based detection model.

• The new detection model yields better performance than some well-known models.

摘要

•Surface defects on Additive Manufactured (AM) workpiece are characterized.•A novel CenterNet-based defect detection model is developed for AM workpiece.•Density map and count loss are merged into CenterNet-based detection model.•The new detection model yields better performance than some well-known models.

论文关键词：Additive manufacturing,Defect detection,Selective laser melting,Convolutional neural network (CNN),Density map estimation,Surface defects,Machine learning,Precision measurement

论文评审过程：Received 20 June 2021, Revised 28 September 2021, Accepted 28 September 2021, Available online 9 October 2021, Version of Record 13 October 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.116000