Markets > Surface Inspection


Visual nondestructive inspections (NDI) for defects are critical for maintenance, repair, and overhaul (MRO) operations and are often the most economical method of detecting surface defects before they reach dangerous sizes. These inspections are conducted in a multitude of industries, including aerospace, infrastructure, and maritime. While technologies such as robotic borescopes and aerial drones have increased number and quality of available inspection data, humans remain the primary decision makers on the acceptability of a part or surface. This manual process is often costly, time-consuming, and subject to human errors that arise from mental fatigue or boredom. Computer vision (CV) and artificial intelligence (AI) algorithms have been shown to excel at image analysis tasks and have the potential to provide accurate and reliable defect recognition for automated visual inspection systems that can significantly reduce maintenance costs and improve inspection efficiency.


AIDL benefits include: (1) reduced maintenance costs, (2) improved inspection quality and consistency, (3) increased flexibility and adaptability of maintenance systems, and (4) improved efficiency and effectiveness of users within the system. AIDL reduces the number of images that require human review, reducing the overall time spent on inspections and increasing throughput to meet production goals. The AI algorithms used for defect recognition can be rapidly retrained and tuned for novel inspection tasks for flexibility in deployment. Errors are tracked to correct misclassifications through continuous model updates.

Example Application

AIDL was first developed under an SBIR Phase II contract for the automated borescope inspection of C-130 propeller blades and can be fine-tuned for a variety of novel visual inspection tasks. Presently, the borescope inspection of C-130 blades relies on human operators to analyze images of the propeller bore for indications of surface defects. Due to the high number of images, it often takes more than 2 hours to complete the manual inspection for a single blade and is a major rate limiting factor of the throughput of the overall C-130 repair and overhaul procedure. The AIDL system provided substantial resource savings by improving inspection efficiency through the use of convolutional neural networks (CNNs) to reduce the number of images that require manual review.

    Defect maps generated by AIDL detection algorithms.