Surrogate model-based multi-objective Bayesian optimisation of porous acoustic barriers.

Area of Science:

• Acoustics and Noise Control
• Computational Engineering
• Materials Science

Background:

• Optimizing multiple conflicting criteria in engineering, especially for acoustic wave propagation, is challenging for standard methods.
• Designing effective noise barriers requires balancing acoustic performance (sound pressure level - SPL) with economic and geometric constraints.

Purpose of the Study:

• To develop a noise prediction surrogate model for multi-objective optimization of acoustic barriers.
• To apply a multi-objective Bayesian optimization algorithm to optimize acoustic barrier design for reduced SPL.

Main Methods:

• A two-dimensional singular boundary method was used to generate a dataset for the surrogate model.
• Multi-objective Bayesian optimization was applied to acoustic line source diffraction with porous noise barriers (straight-walled and T-shaped).
• Surface impedance boundary conditions modeled material dissipation, integrating microstructural and macrostructural parameters.

Main Results:

• The proposed framework efficiently explores trade-offs between acoustic performance, material cost, and shape.
• Optimal barrier designs were achieved by balancing SPL reduction, barrier height, cap length, porosity, tortuosity, and airflow resistivity.
• The surrogate model facilitated computationally intensive optimization for complex acoustic barrier designs.

Conclusions:

• The developed noise prediction surrogate model and Bayesian optimization framework enable efficient multi-objective optimization of acoustic barriers.
• This approach effectively balances acoustic performance with economic and shape constraints for practical engineering solutions.
• The study demonstrates a viable method for designing optimized noise barriers considering various physical and economic factors.