Data-driven impedance tube method for prediction of normal sound absorption coefficienta)

Area of Science:

• Acoustics
• Materials Science
• Machine Learning

Background:

• The two-microphone impedance tube method is standard for acoustic parameter measurement but limited by plane wave assumptions, restricting its effective frequency range.
• Existing methods struggle with multi-modal sound fields, necessitating advanced techniques for accurate acoustic analysis.

Purpose of the Study:

• To propose a data-driven impedance tube method for accurate normal sound absorption coefficient prediction in multi-modal fields.
• To overcome the frequency range limitations inherent in traditional plane wave-based impedance tube measurements.

Main Methods:

• Integration of a neural network model with the impedance tube's transfer relationship.
• Generation of extensive pre-training datasets by constraining physical model boundary conditions.
• Supervised learning strategy to train the neural network for mapping sound pressure to amplitude vectors.

Main Results:

• The proposed data-driven method accurately predicts the normal sound absorption coefficient.
• Simulations and experimental validations confirm the method's predictive capabilities.
• Effective operation within multi-modal sound fields, surpassing traditional limitations.

Conclusions:

• The data-driven impedance tube method offers a robust solution for accurate sound absorption coefficient measurement.
• This approach expands the applicability of impedance tube testing to complex acoustic environments.
• The integration of machine learning provides a powerful tool for advancing acoustic material characterization.