Robust shape optimization of acoustic diffusers for spatial sound distribution using Co-Kriging metamodel

This study introduces a robust shape optimization approach using a Co-Kriging metamodel constructed from hierarchical fidelity simulation. We apply to design the acoustic diffuser panels, which play a crucial role in improving spatial sound distribution across listener areas in large auditorium enclosures. Monte Carlo simulation (MCS) uncertainty propagation is applied to identify output distributions, aiming to compute both performance and robustness. The optimal results are demonstrated on the Pareto front by implementing the Non-dominated Sorting Genetic Algorithm (NSGA-II). The proposed method achieves a reduction in the Sound Pressure Level (SPL) difference from the baseline, validating an improvement in sound uniformity while accounting for performance variance. A deterministic optimal solution was also investigated, achieving a lower SPL difference than the mean performance, but without guaranteed robustness. This study highlights the potential of multi-fidelity-assisted robust optimization as an efficient strategy to leverage multiple sources of data while maximizing the performance under uncertainty in acoustically desirable design. • A hierarchical multi-fidelity approach is applied to energy-based acoustic simulation for robust shape optimization. • Co-Kriging metamodeling combines multi-fidelity data with adaptive sampling to improve global prediction accuracy. • The method improves spatial sound uniformity under geometric uncertainty in auditorium design.