Physics-informed Machine Learning for Acoustic Simulations

Project outline

Numerical methods for acoustic simulations are well established, but scale poorly to high frequencies and large domains. Important applications include simulation of outdoor noise propagation and room acoustics. Physics-informed machine learning is a new and rapidly developing field that has shown promising early results in other physical sciences, and offers significant potential for accelerating and improving accuracy for acoustic simulations.

Key research questions that must be answered in order for this to become a practical methodology for acoustics include:

• Guaranteeing physically accurate results
• More efficient training
• Scaling and generalisation to unseen problem configurations

Project Partners

This project is hosted at the University of Salford.

Student

Supervisors

• Trevor Cox 
• Jonathan Hargreaves