Flow-Noise: Mechanisms and Modeling

• Flow-noise is the acoustic and vibrational fluctuation generated by fluid flows interacting with boundaries, characterized by turbulence and coherent structures.
• Advanced PDE and stochastic models quantify its broadband spectrum, linking turbulent dynamics to measurable noise characteristics.
• Applications span aerodynamic optimization, signal processing, and quantum noise control, driving both engineering improvements and fundamental research.

Flow-noise refers to the broad class of stochastic or deterministic acoustic, macroscopic, or microscopic fluctuations generated by fluid (liquid or gas) flows and their interactions with boundaries, inhomogeneities, or embedded structures. The term encompasses both the physical mechanisms by which fluctuating flows generate observable noise (e.g., mechanical vibrations, sound, signal corruption) and, in theoretical and computational contexts, the explicit mathematical models that encode or exploit the presence of noise within flow-driven systems. Flow-noise is foundational in fields ranging from turbulence physics and hydrodynamics to signal processing and inverse problem solving, and is approached with a diversity of analytic, computational, and experimental techniques.

1. Physical Mechanisms: Flow-Generated Noise in Fluids and Structures

Flow-noise arises through several well-characterized mechanisms in fluid and boundary layer dynamics. In high-Reynolds-number flows, turbulence-induced velocity and pressure fluctuations in moving fluids near solid boundaries generate broadband noise, with key sources such as:

• Turbulent Boundary Layer–Induced Noise (TBLN): Wall pressure fluctuations underneath turbulent layers induce structural vibrations and thus radiated noise. Large-eddy simulation (LES) and power-spectral analysis reveal characteristic features, including a convective ridge and a high-frequency spectral decay, typically between $\omega^{-1}$ and $\omega^{-5}$ (Henke, 2016).
• Coherent Structure Interactions: Formation, evolution, and random switching of coherent structures (vortices, shear layers, circulation cells) create non-Gaussian, heavy-tailed statistics in flow-velocity time series. Random reversals or jumps between large-scale modes are directly responsible for $1/f^\alpha$–type noise, with exponents determined by the statistics (symmetry, bursting, waiting-time distributions) of the underlying flow transitions (Pereira et al., 2018, Herault et al., 2016).
• Aerodynamic and Aeroacoustic Coupling: Oscillating foils, cylinders, and turbulent wakes radiate flow-noise via dipole and quadrupole acoustic sources, quantified through rigorous acoustic-analogy frameworks (Lighthill, Ffowcs-Williams–Hawkings) and observed as frequency-dependent directivity patterns (e.g., dipole axes defined by excitation frequency and modulated by reduced frequency) (Khalid et al., 2018, Wu et al., 2023).
• Noise in Engineered Systems: Flow-noise impacts the design and operation of sonar arrays (hydrodynamic coupling to acoustic waves via flexible windows (Henke, 2016)), wind farms (interaction of turbulence-induced noise and trailing-edge noise, with downstream focusing determined by farm layout and wake superposition (Colas et al., 18 Aug 2025)), and aircraft engines (indirect noise from compositional–entropy fluctuations accelerated through turbine nozzles, with strong dependence on reaction rates and inhomogeneity properties (Jain et al., 2023)).

2. Mathematical and Stochastic Models of Flow-Noise

Central to the analysis of flow-noise is the explicit modeling of its statistical or stochastic properties.

• Renewal Processes and $1/f^\alpha$ Spectra: For turbulent flows with random switching between discrete large-scale states, the low-frequency power-spectral density follows $S(f)\sim 1/f^\alpha$, where the exponent $\alpha$ is linked to the tail exponent $\beta$ of the waiting-time between switching events. The precise relation—$\alpha + \beta = 3$ (symmetric), $\beta - \alpha = 1$ (asymmetric)—depends on flow symmetry and has been confirmed in both experiment and theory (Pereira et al., 2018, Herault et al., 2016). This connects long-term flow memory and emergent flicker noise directly to the dynamics of coherent structures.
• PDE and Semi-Analytic Frameworks: Flow-noise transfer is captured in coupled partial differential equation (PDE) systems, e.g., a turbulent wall–bending plate–acoustic coupling system, which allows prediction of signal power-spectral densities at sensor positions and accounts for both deterministic (mean-field) and stochastic (LES-derived) forcing (Henke, 2016).
• Stochastic Differential Equations and Internal Noise: In passive particle transport, Langevin equations with explicit internal noise regularize singularities (divergent transit times) near stagnation points and enable ensemble-averaged macroscopic transport, with transport statistics strongly modulated by noise intensity and distribution (Gaussian vs. Lévy) (Maryshev et al., 2023).
• Quantum and Nonclassical Systems: In quantum devices and optomechanical platforms, the engineered flow of thermal noise is described using cascaded-system master equations and input–output theory. Unidirectional noise flow and its rectification are achievable via controlled reservoir engineering, with mathematical analysis provided by Lindblad formalisms and Lyapunov equations (Barzanjeh et al., 2017).
• Statistical Field Theory and Exotic Noise: In the context of the stochastic heat equation, "black noise"—where no nontrivial linear functional detects the increment of the flow—is rigorously characterized and linked to unpredictability and absence of an Itô-type calculus (Gu et al., 19 Jun 2025).

3. Flow-Noise in Signal Processing and Machine Learning

In data-driven and inverse-problem domains, flow-noise is exploited both as a nuisance to suppress (denoising) and a structure to explicitly model for robustness.

• Normalizing Flow–Based Noise Models: Image denoising and inverse problems benefit from modeling noise with invertible flows. Approaches such as FINO perform bijective mappings of image–noise pairs, disentangling structural (clean) components from noise in latent space, and enabling robust removal of both Gaussian and heavy-tailed or spatially structured artifacts (Guo et al., 2021, Whang et al., 2020).
• Flow-Based Generative Modeling: Generative models based on rectified flow and related frameworks integrate learnable noise distributions (quantile-based, heavy-tailed, data-adapted), enhancing data-fidelity and realism in tasks like denoising and signal recovery (Chemseddine et al., 14 Oct 2025, Dai et al., 14 Jul 2025, Osman et al., 2024). Optimization-based coupling between noise and data is shown to produce "straight" inference trajectories and improved perceptual and quantitative performance.
• Rectified and Score-Based Flow Approaches: In tasks such as LoRa signal demodulation, rectified-flow SDEs—with tailored noise-scheduling and score-matching objectives—enable recovery of high-fidelity signals at extreme low SNR, leveraging tailored augmentation and hybrid frequency–time neural architectures (Osman et al., 2024).

4. Quantitative Characterization: Spectral, Statistical, and Performance Metrics

Empirical and theoretical quantification of flow-noise utilizes statistical and spectral metrics:

Domain	Metric/Relation	Representative Reference
Turbulent flows	$S(f)\sim 1/f^\alpha$; $\omega^{-5}$0 or $\omega^{-5}$1	(Pereira et al., 2018, Herault et al., 2016)
Aeroacoustics	OASPL; directivity patterns; scaling with Re, St, k	(Khalid et al., 2018, Wu et al., 2023)
Sonar applications	Power-spectral density $\omega^{-5}$2	(Henke, 2016)
Wind farm noise	OASPL, amplitude modulation, focusing zones	(Colas et al., 18 Aug 2025)
Generative modeling	FID, KID, IS, BER, NRMSE, NFSS	(Dai et al., 14 Jul 2025, Osman et al., 2024)
Quantum noise channels	Rectification coefficient $\omega^{-5}$3	(Barzanjeh et al., 2017)
Special flow transport	Mean cell-to-cell transit time $\omega^{-5}$4	(Maryshev et al., 2023)

Such metrics provide both a basis for comparing theoretical predictions with empirical observations and for guiding design and control of physical systems susceptible to or leveraging flow-noise.

5. Control, Design, and Engineering Implications

Understanding flow-noise phenomena leads to design guidelines and mitigation strategies:

• Structural and Acoustic Mitigation: In sonar and underwater vehicles, plate geometry, damping (orthotropic viscoelastic layers), and sensor placement modulate transmitted noise. In bioinspired oscillating fins, kinematic parameters (frequency, amplitude, phase) and geometry (reduced frequency, aspect ratio) optimize for minimal or redirected acoustic output (Henke, 2016, Khalid et al., 2018).
• Aerodynamic Optimization: Wind farm layout—aligned vs. staggered—significantly alters both the source and propagation characteristics of noise, with staggered arrays producing higher downstream noise levels due to complex wake superposition effects not seen in isolated-turbine models (Colas et al., 18 Aug 2025).
• Information Processing and AI: Flow-noise-aware models enable robust denoising, inverse problem recovery with highly structured/heteroskedastic data, and data-adaptive generative sampling beyond standard Gaussian assumptions (Guo et al., 2021, Whang et al., 2020, Chemseddine et al., 14 Oct 2025).
• Quantum Noise Control: Engineered quantum circuits with deliberate introduction of non-reciprocal thermal noise flow offer new primitives for thermal logic, amplifiers, and routers at the nanoscale (Barzanjeh et al., 2017).

6. Advanced Theoretical Connections and Broader Implications

Research into flow-noise phenomena reveals deep connections between statistical physics, dynamical systems, information theory, and quantum many-body systems:

• The classification of stochastic heat flow as "black noise" situates it in the hierarchy of mathematical objects with vanishing linear predictability, aligning with objects such as the Brownian Web and critical percolation, and indicating the intrinsic necessity of higher-order statistical approaches (Gu et al., 19 Jun 2025).
• In turbulent flows and associated systems, flow-noise is both a diagnostic for underlying long-term memory and symmetry-breaking phenomena, and an active variable in the design and control for technological and geophysical applications (Pereira et al., 2018, Herault et al., 2016).
• A plausible implication is that continued unification of physical, computational, and statistical perspectives on flow-noise will drive both improved physical understanding (e.g., turbulence closure, noise-induced transitions) and technological progress (e.g., stealth hydrodynamics, robust AI-aided communication).

7. Open Questions and Research Directions

Ongoing challenges include:

• Extension of renewal-process-based $\omega^{-5}$5 noise theory to multistate or continuous-amplitude dynamics, accounting for multi-scale turbulent features (Herault et al., 2016).
• Incorporation of complex, possibly non-Gaussian or non-Markovian noise structures in both physical flow modeling and AI-based signal processing (Chemseddine et al., 14 Oct 2025, Guo et al., 2021).
• Further experimental and numerical validation of theoretical predictions—such as the precise $\omega^{-5}$6 relations in complex turbulence, or the impact of engineered noise in quantum devices (Pereira et al., 2018, Barzanjeh et al., 2017).
• Systematic coupling of realistic flow fields and full-wave acoustic models to produce predictive, end-to-end environmental and technological noise assessments (Colas et al., 18 Aug 2025).
• Detailed theoretical and experimental assessment of black-noise classification for stochastic flows in higher dimensions and their practical consequences for predictability and information transfer (Gu et al., 19 Jun 2025).

Research into flow-noise continues to provide both fundamental insights into the mathematics and physics of randomness in extended systems and concrete strategies for modeling, control, and exploitation in real-world engineered and natural systems.