Design and Additive Manufacturing of Porous Sound Absorbers-A Machine-Learning Approach

Area of Science:

• Materials Science
• Acoustics
• Machine Learning

Background:

• Additive manufacturing (AM), or 3D-printing, allows precise fabrication of porous sound absorbers.
• The relationship between micro-scale structure and acoustic behavior is not fully understood for AM materials.

Purpose of the Study:

• To develop machine learning models for predicting acoustic material parameters (Biot parameters) from micro-scale geometry.
• To bridge the knowledge gap between material structure and acoustic performance in 3D-printed absorbers.

Main Methods:

• Manufactured 50 test specimens using material extrusion (MEX) AM with varying bar designs.
• Used laboratory measurements (flow resistivity, absorption coefficient) to inversely identify Biot parameters.
• Trained artificial neural network and k-nearest neighbor models using micro-scale geometry and identified Biot parameters.

Main Results:

• Both machine learning models accurately predicted Biot parameters from micro-scale geometry.
• The models demonstrated the ability to predict material behavior beyond the initial training data.

Conclusions:

• Machine learning effectively computes acoustic parameters from micro-scale geometry.
• This approach facilitates tailored material design for specific acoustic applications and extends material behavior prediction capabilities.