Enhancing aircraft safety: Automated three-dimensional defect detection, localization and sizing in non-destructive testing

Abstract

In most cases, non-destructive testing (NDT) techniques typically rely solely on two-dimensional image data for defect detection, particularly in CT imaging. This limitation hindered the ability to accurately reconstruct the exact three-dimensional form of defects. In this study, we propose solutions for three-dimensional image reconstruction, which is crucial in industrial non-destructive testing applications and in the aircraft industry. We introduce a new, fully automated method for detecting, locating, and sizing defects in the context of non-contact quality control in industry, specifically focusing on aircraft-type equipment. Our method was applied to a confidential database containing over 120,000 images from Tassili Work Airlines Company. This database was curated and labeled by senior experts in the field of diagnostics and non-destructive testing, and we compare our results with theirs. Our combined approach, utilizing expectation maximization and fuzzy inference penalty, proves to be effective in addressing the challenging inverse problem of three-dimensional computed tomography defect detection, localization, and dimensioning. This contributes to enhancing safety in aeronautical transportation by enabling accurate diagnosis of parts.