Virtual-to-real mapping: an AE missing signal generation and assisted localization method for defects in CFRP plates

Area of Science:

• Materials Science
• Structural Health Monitoring
• Non-destructive Testing

Background:

• Structural health monitoring (SHM) of carbon fiber reinforced polymers (CFRP) is critical for safety and longevity.
• Acoustic emission (AE) is a key technique for defect localization in CFRP, but challenges remain in detecting and locating signals from undetected defects.
• CFRP's anisotropic properties complicate AE signal propagation, impacting localization accuracy and cost-effectiveness.

Purpose of the Study:

• To develop a cost-effective method for generating undetected AE signals in CFRP structures.
• To enhance the accuracy of defect localization and classification in CFRP using data-driven approaches.
• To validate the proposed method's applicability across different material systems and sensor configurations.

Main Methods:

• Virtual-to-real mapping was employed to generate simulated AE signals based on material properties and sensor configurations.
• An Auto-encoder model was utilized to map experimental and simulated signals, predicting missing waveforms.
• Generated signals were integrated into a multilayer perceptron model for defect classification and localization analysis.

Main Results:

• The virtual-to-real mapping method successfully generated missing AE waveforms, validated by quantitative and qualitative analyses.
• Incorporating generated signals improved multiple missing region classification accuracy by an average of 21.1% and independent missing region classification accuracy by 43.9%.
• The method demonstrated robust performance on 6061 aluminum alloy plates and was found to be largely insensitive to sensor placement variations.

Conclusions:

• The proposed virtual-to-real mapping offers a cost-effective solution for AE data collection and enhances defect localization in CFRP.
• This data-driven approach provides valuable insights for predicting and localizing defects in composite structures.
• The method's cross-material applicability and robustness to sensor placement suggest broad potential in structural health monitoring.