ADComp

Acoustic Dispersion Compensation in Reduced-Order Models

ADComp is a research project for studying numerical dispersion and its interaction with model-order reduction in acoustic wave propagation.

Overview

Reduced-order acoustic models can exhibit lower error than their underlying spatial discretizations when modal frequency shifts partially compensate numerical dispersion. This effect depends on the excitation, receiver configuration, and observation interval.

ADComp provides a numerical framework to examine these dependencies using heterogeneous acoustic cavities and continuous or refined numerical references.

Methods

• Craig–Bampton and projection-based model-order reduction
• Modal frequency and residue analysis
• Source–receiver sensitivity and finite-time error decomposition
• Reference-based validation and reduced-dimension selection

Intended Use

ADComp is intended for research, numerical experimentation, and reproducibility studies in computational acoustics. It is not a pretrained neural network or a general-purpose acoustic prediction service.

Limitations

The analysis focuses on prescribed linear, lossless acoustic systems. Results should not be assumed to transfer to damped, radiating, nonlinear, or otherwise different configurations without further validation. Frequency substitution is a diagnostic procedure, not a validated correction solver.