Acoustic performance prediction of fibrous materials using anisotropic and fundamental parameter distributions

Eco-friendly and recycled fibrous acoustic materials are emerging as key solutions for sound package interiors. Optimizing these materials for target acoustic performance requires predictive models that account for microstructural parameters. However, existing approaches often neglect inhomogeneity and anisotropy, relying on homogeneous assumptions and inconsistent parameter selections. This study introduces a deep learning model that leverages anisotropic parameter distributions to predict acoustic performance with significantly reduced computational cost. PET felt cross sections were analyzed via micro-computed tomography (micro-CT) to extract anisotropic parameter distributions, which were used to train the model. The proposed framework demonstrates superior predictive performance compared to conventional homogeneous models. Its efficiency in handling large datasets and reducing computation time makes it highly effective for optimizing and designing fibrous acoustic materials. By identifying critical structural parameters influencing acoustic behavior, this model enables data-driven optimization and tailored design. Furthermore, this research lays the foundation for practical applications, extending to diverse fibrous materials and advancing toward generating structural design images that enable multi-objective optimization. By incorporating acoustic performance and constraints like cost and sustainability, these models open possibilities for real-world applications, bridging design, and functionality.