Advancing IR Laser Thermography in Composites via Thermal Analysis-Informed Machine Learning.

Area of Science:

• Materials Science
• Non-destructive Testing
• Thermal Analysis

Background:

• Accurate characterization of material thermal properties is crucial for fatigue damage assessment using passive thermography.
• Traditional methods for measuring transverse conductivity (k_ct) and film coefficient (h) are often difficult and invasive.

Purpose of the Study:

• To develop a nondestructive method for simultaneously determining transverse conductivity (k_ct) and film coefficient (h) in composite laminates.
• To enhance the accuracy of machine learning-assisted passive thermography for fatigue damage characterization.

Main Methods:

• Utilized a near-infrared laser and spatial light modulator to create artificial temperature hotspots on a glass-epoxy composite.
• Employed Bayesian optimization (BO) with a 3D Finite Element Method (FEM) thermal model to iteratively refine k_ct and h values.
• Recorded ground truth temperature distributions using a midwave infrared thermal camera.

Main Results:

• Bayesian optimization converged within approximately 10 iterations, yielding k_ct = 0.40 W/m K and h = 5.78 W/(m² K).
• The loss objective decreased by a factor of approximately 3 within seven iterative rounds.
• Experimental validation demonstrated successful reproduction of thermal images with minor deviations using optimized parameters.

Conclusions:

• The proposed nondestructive method effectively and accurately determines key thermal properties (k_ct and h).
• This approach enhances the reliability of machine learning-assisted passive thermography for material characterization and damage assessment.
• The method shows robustness and applicability for real-world material analysis.