Deep learning-based approach for linking microstructural and macroscopic acoustic properties of sound-absorbing...

Area of Science:

• Materials Science
• Acoustics
• Data Science

Background:

• Sound absorption in porous materials is complex, depending heavily on microstructure.
• Quantifying the relationship between microstructure and acoustic performance is a significant challenge.
• Flexible polyurethane (PU) foam is widely used but requires optimized acoustic properties.

Purpose of the Study:

• To develop a deep learning framework for analyzing PU foam microstructure and predicting acoustic performance.
• To establish a quantitative, design-oriented link between microstructural features and sound absorption.
• To provide data-driven insights for designing advanced sound-absorbing materials.

Main Methods:

• Utilized a U-Net model for semantic segmentation of SEM images to extract microstructural parameters (cell size, pore size, pore shape, strut thickness).
• Developed an artificial neural network to model the relationship between extracted microstructural parameters and measured acoustic performance.
• Trained and validated the models on 210 flexible PU foam samples, including aged and non-aged materials.

Main Results:

• Successfully extracted key microstructural parameters from SEM images.
• Established a quantitative correlation between microstructural morphology and sound absorption coefficients.
• Validated the deep learning framework through experimental comparisons and parameter contribution analysis.

Conclusions:

• The proposed deep learning framework effectively links microscale morphology to acoustic performance in flexible PU foam.
• This approach offers practical guidance for material design, optimization, and fabrication of sound-absorbing materials.
• The findings support the development of tailored acoustic solutions for industries like automotive and construction.