Automatic Detection of Defects in High-Reliability Components

Technical Report · Thu Sep 29 04:00:00 EDT 2022

DOI:https://doi.org/10.2172/1890067· OSTI ID:1890067

Potter, Kevin M. ^[1]; Famili, Soroush ^[1]; Garland, Anthony P. ^[1]; Jones, Jessica E. ^[1]; Pant, Aniket ^[1]

---

1. Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA (United States)

Disastrous consequences can result from defects in manufactured parts—particularly the high consequence parts developed at Sandia. Identifying flaws in as-built parts can be done with nondestructive means, such as X-ray Computed Tomography (CT). However, due to artifacts and complex imagery, the task of analyzing the CT images falls to humans. Human analysis is inherently unreproducible, unscalable, and can easily miss subtle flaws. We hypothesized that deep learning methods could improve defect identification, increase the number of parts that can effectively be analyzed, and do it in a reproducible manner. We pursued two methods: 1) generating a defect-free version of a scan and looking for differences (PandaNet), and 2) using pre-trained models to develop a statistical model of normality (Feature-based Anomaly Detection System: FADS). Both PandaNet and FADS provide good results, are scalable, and can identify anomalies in imagery. In particular, FADS enables zero-shot (training-free) identification of defects for minimal computational cost and expert time. It significantly outperforms prior approaches in computational cost while achieving comparable results. FADS’ core concept has also shown utility beyond anomaly detection by providing feature extraction for downstream tasks.

• Research Organization: Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States)
• Sponsoring Organization: USDOE National Nuclear Security Administration (NNSA)
• DOE Contract Number: NA0003525
• OSTI ID: 1890067
• Report Number(s): SAND2022-13025; 710366
• Country of Publication: United States
• Language: English

Related Subjects

47 OTHER INSTRUMENTATION