Time-reversal-based ultrasonic non-destructive evaluation using convolutional neural networks.

Area of Science:

• Materials Science
• Mechanical Engineering
• Signal Processing

Background:

• Ultrasonic non-destructive evaluation (NDE) faces challenges in detecting defects in composite materials and complex structures due to signal scattering.
• Accurate defect identification is crucial for aerospace and space launch vehicle safety and reliability.

Purpose of the Study:

• To develop a highly accurate NDE method for composite materials and complex structures.
• To leverage convolutional neural networks (CNNs) and time-reversal (TR) signals for enhanced defect detection.
• To validate the proposed method on honeycomb sandwich panels and skin-stringer structural models.

Main Methods:

• A finite element model was validated against experimental data from an aluminum panel.
• A CNN-based classification framework was trained using time-reversal (TR) signals.
• The method was applied to honeycomb sandwich panels and skin-stringer structural models for defect classification.

Main Results:

• The CNN model achieved high classification accuracy when trained with TR signals.
• The finite element model demonstrated the algorithm's effectiveness on complex structural models.
• The time-reversal method successfully generated strong signals at defect locations, improving detection.

Conclusions:

• The proposed CNN-based NDE method using TR signals offers a robust and high-performance solution for defect detection.
• The study confirms the effectiveness of acquiring large, reflected signals from defects through time reversal for accurate classification.
• This technique is particularly promising for NDE applications in aerospace and complex mechanical structures.