Hybrid Neural Network-Based Approach for the Estimation of Sound Levels under Variable Meteorological Conditions
DOI:
https://doi.org/10.70567/mc.v42.ocsid8370Keywords:
Hybrid model, sound propagation, meteorological effectAbstract
A hybrid methodology is presented that combines artificial neural networks with the semiempirical acoustic propagation model of ISO 9613-2, with the aim of estimating the sound level at a receiver while accounting for the influence of local meteorological variables. Two scenarios are distinguished: calm conditions (no wind), where the ISO model is used as a reference, and windy conditions, where real measurements are employed. The network is trained using a mixed loss function that combines the fit to measurements under wind conditions with a penalty term for deviations from the structure of the reference model. In addition, functional constraints are incorporated on the wind dependence for extreme velocity values, adopting a sigmoid-like form as reported in the literature. This approach balances the predictive power of data-driven models with a reference structure guided by physical principles, yielding a regularized surrogate model with reduced dependence on large data volumes.
References
Di Giorgio, L. E. y Sequeira, M. E. Estimación de los efectos del viento en la medición de ruido urbano utilizando lógica difusa. Mecánica Computacional, 41:23-31, 2024.
Goodfellow, I., Bengio, Y. y Courville, A. Deep Learning; MIT Press: Cambridge, MA, USA, 2016.
Hohenwarter, D., Mursch-Radlgruber, E. y Kirisits, C. S-shaped dependence of the sound pressure level in outdoor propagation on the effective sound speed gradient. Acta Acustica, 6, 13, 2022. https://doi.org/10.1051/aacus/2022005
ISO 9613-2, Acoustics – Attenuation of Sound During Propagation Outdoors – Part 2: General Method of Calculation, 1996.
Jul-Rasmussen, P., Kumar, M., Pinto, J., Oliveira, R., Liang, X. y Huusom, J. K. Incorporating first-principles information into hybrid modeling structures: Comparing hybrid semiparametric models with Physics-Informed Recurrent Neural Networks. Computers & Chemical Engineering, 199:109-119, 2025.https://doi.org/10.1016/j.compchemeng.2025.109119
Salomons, E. M. y Bakri, T. (2018). Fluctuating traffic noise levels calculated from time-dependent traffic data: an engineering approach. Noise Control Engineering Journal, 66, 5:432–445, 2018. https://doi.org/10.3397/1/376637
Sequeira, M. E., Di Giorgio, L. E., Azzurro, A. P. y Cortínez, V. H. Modelo Computacional para Caracterización Acústica de Zona Residencial Aledaña al Polo Industrial de la Ciudad de Bahía Blanca. Mecánica Computacional, 40(3):59-66, 2023a.
Sequeira M. E., Di Giorgio L. E., Azzurro A. P. y Cortínez V. H. Análisis del ruido ambiental en una zona residencial lindera a un parque industrial en la ciudad de Bahía Blanca. XVIII Congreso Argentino de Acústica, 6 pp., 6 y 7 de diciembre, Universidad Nacional de Quilmes, Bernal, Buenos Aires, 2023b.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Argentine Association for Computational Mechanics
This work is licensed under a Creative Commons Attribution 4.0 International License.
This publication is open access diamond, with no cost to authors or readers.
Only those papers that have been accepted for publication and have been presented at the AMCA congress will be published.
