Ultrasonic Phased Array Testing and Identification of Multiple-Type Internal Defects in Carbon Fiber Reinforced...

Area of Science:

• Materials Science
• Nondestructive Testing
• Artificial Intelligence

Background:

• Carbon fiber reinforced plastics (CFRPs) are prone to internal defects like delamination and impacts, compromising performance.
• Ultrasonic phased array (UPA) inspection is crucial for nondestructive testing (NDT) of CFRPs.
• Current UPA techniques struggle with accurate classification of diverse CFRP defects.

Purpose of the Study:

• To develop and validate a UPA-based NDT method for classifying internal defects in CFRPs.
• To leverage convolutional neural networks (CNNs) for enhanced defect identification and categorization.
• To improve the accuracy of detecting both manufacturing and impact-related defects in CFRPs.

Main Methods:

• A dataset of ultrasonic C-scan images featuring various internal CFRP defects was created.
• Defect features within C-scan images were analyzed.
• An autoencoded classifier network was designed and trained for defect recognition.

Main Results:

• The proposed CNN-based method demonstrated robust defect feature extraction capabilities.
• The technique achieved high accuracy in identifying both impact and manufacturing defects.
• The autoencoded classifier effectively recognized defects of varying sizes and types.

Conclusions:

• The CNN-based UPA approach offers a significant advancement in CFRP defect classification.
• This method enhances the reliability of NDT for ensuring CFRP quality and safety.
• The study highlights the potential of AI in addressing complex challenges in materials inspection.