An Analytical-Prior Framework for Data-Efficient Prediction of Side-Branch Resonator Acoustics

Forum topic · 小凯 · 2026-08-19

Summary

High-fidelity finite-element simulations provide accurate predictions for side-branch resonators, but generating large simulation datasets is expensive, and purely data-driven surrogate models can become unreliable when simulation-labelled data are scarce. This study (arXiv:2608.16873, author Jiaming Li) introduces an analytical-prior learning framework that reuses a low-cost analytical model to improve data efficiency under limited high-fidelity simulation budgets. Two complementary routes are explored: when the analytical model remains available at inference, it serves as an explicit baseline and simulation data are used only to learn the analytical-to-simulation discrepancy; when a self-contained predictor is needed, the analytical mapping is first distilled from abundant low-cost evaluations into a learned prior and then calibrated with limited simulation data. The framework was evaluated on rectangular side-branch Helmholtz resonators using 86 simulation-labelled geometries and 8,998 non-overlapping analytical-only geometries. Results show that analytical-prior information substantially improves high-fidelity predictions when simulation data are scarce, with explicit correction and prior distillation serving complementary deployment needs.

Paper Overview

Research area: Machine Learning Author: Jiaming Li Published: 2026-08-17 arXiv: 2608.16873

Abstract

High-fidelity finite-element simulations can provide accurate numerical predictions for side-branch resonators, but large simulation datasets are expensive to generate and purely data-driven surrogates may become unreliable when simulation-labelled data are scarce. This study develops an analytical-prior learning framework that reuses a low-cost analytical model to improve data efficiency under limited high-fidelity simulation budgets.

Two complementary routes are considered:

Explicit discrepancy correction: when the analytical model remains available at inference, it is retained as an explicit baseline, and the simulation data are used to learn only the analytical-to-simulation discrepancy.
Prior distillation: when a self-contained predictor is required, the analytical mapping is first distilled from abundant low-cost evaluations into a learned prior, then calibrated with limited simulation data.

Evaluation

The framework was evaluated on rectangular side-branch Helmholtz resonators using:

86 simulation-labelled geometries
8,998 non-overlapping analytical-only geometries

Findings

Analytical-prior information significantly improves high-fidelity predictions when simulation data are scarce, and explicit correction and prior distillation serve complementary deployment scenarios.

--- *Auto-collected on 2026-08-19*