A deep transfer learning model for inclusion defect detection of aeronautics composite materials. (15th November 2020)

• Record Type: Journal Article
• Authors: Gong, Yanfeng; Shao, Hongliang; Luo, Jun; Li, Zhixue
• Is Part Of: Composite structures. Volume 252(2020)
• Publication Date: 2020-11-15
• Subjects: Inclusion defect detection -- Aeronautics composite materials -- Transfer learning -- Feature extraction; Composite construction -- Periodicals; Composites -- Périodiques; 624.18
• DOI: 10.1016/j.compstruct.2020.112681
• Languages: English
• ISSNs: 0263-8223

Abstract

Highlights: A novel deep transfer learning model is proposed for aeronautics composite materials. An inspection method combining deep learning and sliding-window approach is explored. It's the first application of transfer learning in composites' defect detection. Abstract: Composite materials are increasingly used as structural components in military and civilian aircraft. To ensure their high reliability, numerous non-destructive testing (NDT) techniques have been used to detect defects during production and maintenance. However, most of these techniques are non-automatic, with diagnostic results determined subjectively by operators. Some deep learning methods have been proposed to identify defects in images obtained through NDT, but they need labeled image samples with defects, which can be expensive or unavailable. We propose a deep transfer learning model to accurately extract features for the inclusion of defects in X-ray images of aeronautics composite materials (ACM), whose samples are scarce. We researched an automatic inclusion defect detection method for X-ray images of ACM using our proposed model. Experimental results show that the model can reach 96% classification accuracy (F1_measure) with satisfactory detection results.