Kolmogorov-Arnold networks modeling of wall pressure wavenumber-frequency spectra under turbulent boundary layers

• Published in: Theoretical and applied mechanics letters Vol. 15; no. 2; p. 100573
• Main Authors: Zhou, Zhiteng, Liu, Yi, Wang, Shizhao, He, Guowei
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2025; Elsevier
• Subjects: Kolmogorov-Arnold networks modeling; Turbulent boundary layers; Wavenumber-frequency spectra; Kolmogorov-Arnold networks modeling; Wavenumber-frequency spectra; Turbulent boundary layers
• ISSN: 2095-0349
• DOI: 10.1016/j.taml.2025.100573

•Develop wavenumber-frequency spectrum model with Kolmogorov-Arnold Networks.•Obtain a smooth wavenumber-frequency spectra model with sparse samples in turbulent channel flows.•Capture the wavenumber-frequency spectra near the convective peak in the turbulent flow over an axisymmetric body of revolution. [Display omitted] The empirical models for wavenumber-frequency spectra of wall pressure are broadly used in the fast prediction of aerodynamic and hydrodynamic noise. However, it needs to fit the parameter using massive data and is only used for limited cases. In this letter, we propose Kolmogorov-Arnold networks (KAN) base models for wavenumber-frequency spectra of pressure fluctuations under turbulent boundary layers. The results are compared with DNS results. In turbulent channel flows, it is found that the KAN base model leads to a smooth wavenumber-frequency spectrum with sparse samples. In the turbulent flow over an axisymmetric body of revolution, the KAN base model captures the wavenumber-frequency spectra near the convective peak.