Study on Novel Surface Defect Detection Methods for Aeroengine Turbine Blades Based on the LFD-YOLO Framework.

Area of Science:

• Materials Science
• Aerospace Engineering
• Computer Vision

Background:

• Surface defect detection on aeroengine turbine blades (ATBs) faces challenges with low accuracy and efficiency.
• Existing methods struggle to identify minute defects and meet real-time performance requirements.

Purpose of the Study:

• To develop a novel, highly accurate, and efficient method for ATB surface defect detection.
• To address limitations in current defect detection techniques for critical aerospace components.

Main Methods:

• Employs the LDconv model for dynamic convolutional kernel adjustment.
• Integrates the deformable attention mechanism (DAT) for effective capture of minute defect features.
• Utilizes Focaler-CIoU to optimize the bounding box loss function.

Main Results:

• Achieved a mean average precision (mAP_0.5) of 96.2%.
• Reached an F-measure of 96.7% and an identification rate (I) of 98.8%.
• Maintained a detection speed exceeding 25 images per second.

Conclusions:

• The proposed method significantly enhances precision and efficiency in ATB surface defect detection.
• The approach meets stringent requirements for accuracy and real-time performance in aerospace applications.