Sign in Subscribe
Artificial intelligence

Researchers Develop Advanced AI Model for Automated Damage Diagnostics in Aerospace and Automotive Industries

Strategic Insights

— 2 min read

Investigating the importance of advanced artificial intelligence (AI) in the aerospace and automotive industries, researchers from the University of Bremen have recently developed a novel AI model for automated damage diagnostics in fiber metal laminates (FML) plates. According to the study, the impact damage is one of the major causes of structural failures in FML plates, which are widely used in these industries due to their superior mechanical properties. The researchers employed state-of-the-art deep learning models, including the Segment Anything Model (SAM) and the Mask Region-based Convolutional Neural Network (Mask R-CNN) implemented by the Detectron2 framework, to detect, segment, reconstruct, and characterize the damages in FML plates.

Key Takeaways:

  • The study proposes an automated approach to detect, segment, and characterize the damages in FML plates using deep learning models, which is crucial for improved safety and reduced maintenance costs.
  • The researchers applied a domain-adapted supervised learning process to the X-ray CT dataset of damaged FML plates impacted with energies of 5J, 7.5J, 10J, and 12.5J.
  • The Mask R-CNN model significantly outperformed SAM across all key performance metrics, offering around 8 times faster training and 80 times faster inference.
  • Mask R-CNN also proved to have superior explainability for end-users, which is essential for adoption in industrial inspection and quality assurance.
  • The study contributes to the area of damage diagnostics in composite materials and provides insights into the comparative performance and explainability of advanced deep learning models.
  • The researchers employed a domain-adapted supervised learning process, which is a novel approach for damage diagnostics in FML plates.
  • The study highlights the need for further studies due to the lack of absolute ground truth data for an absolute quantitative comparison.
  • The research was financially supported by the Deutsche Forschungsgemeinschaft.

Statistics:

  • The researchers applied a domain-adapted supervised learning process to the X-ray CT dataset of damaged FML plates impacted with energies of 5J, 7.5J, 10J, and 12.5J.
  • The Mask R-CNN model offered around 8 times faster training and 80 times faster inference compared to SAM.
  • The study proposes an automated approach to detect, segment, and characterize the damages in FML plates with an accuracy of 90% using deep learning models.

Sources:

  • NewsRx. University of Bremen Researchers Highlight Research in Artificial Intelligence (Investigation of deep learning approaches for automated damage diagnostics in fiber metal laminates using Detectron2 and SAM). Robotics & Machine Learning. September 8, 2025; p 646.
  • Investigation of deep learning approaches for automated damage diagnostics in fiber metal laminates using Detectron2 and SAM. Frontiers in Artificial Intelligence, 2025,8.
  • Frontiers in Artificial Intelligence. Publisher: Frontiers Media S.A. DOI: 10.3389/frai.2025.1599345.

Read more