CyberSec Research

We introduce CFDagent, a zero-shot, multi-agent system that enables fully autonomous computational fluid dynamics (CFD) simulations from natural language prompts. CFDagent integrates three specialized LLM-driven agents: (i) the Preprocessing Agent that generates 3D geometries from textual or visual inputs using a hybrid text-to-3D diffusion model (Point-E) and automatically meshes the geometries; (ii) the Solver Agent that configures and executes an immersed boundary flow solver; and (iii) the Postprocessing Agent that analyzes and visualizes the results, including multimodal renderings. These agents are interactively guided by GPT-4o via conversational prompts, enabling intuitive and user-friendly interaction. We validate CFDagent by reproducing canonical sphere flows at Reynolds numbers of 100 and 300 using three distinct inputs: a simple text prompt (i.e., "sphere"), an image-based input, and a standard sphere model. The computed drag and lift coefficients from meshes produced by each input approach closely match available data. The proposed system enables synthesization of flow simulations and photorealistic visualizations for complex geometries. Through extensive tests on canonical and realistic scenarios, we demonstrate the robustness, versatility, and practical applicability of CFDagent. By bridging generative AI with high-fidelity simulations, CFDagent significantly lowers barriers to expert-level CFD, unlocking broad opportunities in education, scientific research, and practical engineering applications.

This study investigates the effect of surface dimples on the unsteady aerodynamics of a National Advisory Committee for Aeronautics airfoil (NACA0012) at a chord-based Reynolds numbers of $Re_c = 5300$ and $10{,}000$ using direct numerical simulations. Dimples were placed on the suction side at non-dimensional chordwise locations of $l_D/c = 0.035$ and $0.35$, and the flow response was studied at a fixed angle of attack $\alpha = 5^\circ$. At $Re_c = 5300$, dimples placed at $l_D/c = 0.35$ reduced lift and drag fluctuations by $26.5\%$ and $33.3\%$, respectively, with minimal change in mean forces. At $Re_c = 10{,}000$, the same configuration led to a seven-fold increase in force fluctuations, while the mean remained unchanged. The smooth airfoil exhibited irregular, aperiodic force signals at this $Re_c$, whereas the dimpled case showed highly periodic behavior, indicating wake stabilization. Flow visualizations revealed that dimples generate streamwise vortices within the boundary layer. These vortices are found to have a stabilizing effect on wake dynamics at $Re_c = 5300$, reducing vortex breakdown and enhancing the coherence of wake structures. Spectral Proper Orthogonal Decomposition (SPOD) showed that dimples redistribute modal energy depending on Reynolds number: at low $Re_c$, they reduce broadband content and suppress unsteadiness, while at high $Re_c$, they amplify dominant shedding modes and broaden the spectral energy distribution. These results demonstrate that dimples can passively modulate unsteady forces and wake dynamics for a flow over a streamlined body, either suppressing or enhancing flow instabilities depending on the regime.

A wide range of natural and engineered fluid flows exhibit spatial or temporal viscosity variations, spanning scales from microbial locomotion to planetary mantle convection. These variations introduce qualitatively new physical mechanisms absent in constant-viscosity flows. This review surveys such phenomena across scales. In low Reynolds number (Stokes) flows, viscosity gradients couple translation and rotation, enabling novel particle responses to uniform forcing-- mechanisms that microorganisms may exploit. In shear flows, viscosity variation alters base flow profiles and breaks symmetries, modifying stability and transition dynamics. At high Reynolds numbers, stratification fundamentally changes the singular perturbation structure governing energy production, enhancing or suppressing canonical instabilities and introducing new ones. Viscosity variation also affects nonnormal growth and nonlinear interactions that drive transition to turbulence. While laminar and fully developed turbulence have been extensively studied, transitional processes remain poorly understood in variable-viscosity flows. In turbulent regimes, viscosity variation impacts jets, wall-bounded flows, and mixing layers. At geophysical scales, incorporating eddy viscosity stratification in climate models may improve predictions, while in Earth's mantle, viscosity contrasts drive large-scale convection and geological evolution. Particle-laden flows, common across contexts, can generate effective viscosity stratification through inhomogeneous loading. Throughout, we highlight cases where viscosity variation alters flow behavior qualitatively, and point to open questions. This review aims to guide graduate students and researchers toward tractable, cross-disciplinary problems.

Zonal jets manifest themselves as bands with sharp interfaces in the vorticity configuration. We develop an algorithm to track these fluctuating vorticity interfaces and systematically investigate their characteristic spatio-temporal behavior. While the interfacial height fluctuations are typically sub-Gaussian, the corresponding $\textit{fluctuation speeds}$ exhibit wider, heavy-tailed distributions reflecting the influence of lateral dispersion induced by the zonal velocity profile along the interfacial contours. The temporal evolution of these fluctuations is further characterized through their power spectrum displaying scale invariance in the frequency domain. The sharp, dense, shock-like features present in the time series of the $\textit{height}$ field suggest a possible lacking of differentiability. We confirm this by calculating the moments of the time-increments of the interfacial height fluctuations. Finally, the fractal nature of these boundaries is investigated systematically through a multifractal approach, revealing the non-trivial, complex statistics of interfaces in such geophysical, turbulent flows.

Time-varying flow-induced forces on bodies immersed in fluid flows play a key role across a range of natural and engineered systems, from biological locomotion to propulsion and energy-harvesting devices. These transient forces often arise from complex, dynamic vortex interactions and can either enhance or degrade system performance. However, establishing a clear causal link between vortex structures and force transients remains challenging, especially in high-Reynolds number nominally three-dimensional flows. In this study, we investigate the unsteady lift generation on a rotor blade that is impulsively started with a span-based Reynolds number of 25,500. The lift history from this direct-numerical simulation reveals distinct early-time extrema associated with rapidly evolving flow structures, including the formation, evolution, and breakdown of leading-edge and tip vortices. To quantify the influence of these vortical structures on the lift transients, we apply the force partitioning method (FPM) that quantifies the surface pressure forces induced by vortex-associated effects. Two metrics - $Q$-strength and vortex proximity - are derived from FPM to provide a quantitative assessment of the influence of vortices on the lift force. This analysis confirms and extends qualitative insights from prior studies, and offers a simple-to-apply data-enabled framework for attributing unsteady forces to specific flow features, with potential applications in the design and control of systems where unsteady aerodynamic forces play a central role.

The extensional rheology of dilute suspensions of spheres in viscoelastic or polymeric liquids is studied computationally. At low polymer concentration (c) and Deborah number (De), a wake of highly stretched polymers forms downstream of the particles due to larger local velocity gradients than the imposed flow, indicated by a positive deviation in local De. This increases the suspension's extensional viscosity with time and De for De less than 0.5. When De exceeds 0.5 (the coil-stretch transition), the fully stretched polymers from the far field collapse in regions with lower local velocity gradients around the particle's stagnation points, reducing suspension viscosity relative to the polymer-only liquid. The interaction between local flow and polymers intensifies with increasing c. Highly stretched polymers impede local flow, reducing local De, while it increases in regions with collapsed polymers. Initially, increasing c aligns local De and polymer stretch with far-field values, diminishing particle-polymer interaction effects. However, beyond a certain c, a new mechanism emerges. At low c, fluid three particle radii upstream exhibits increased local De, stretching polymers beyond their undisturbed state. As c increases, this deviation becomes negative, collapsing polymers and resulting in increasingly negative stress from particle-polymer interactions at large De and time. At high c, this negative interaction stress scales as c squared, surpassing the linear increase in polymer stress, making dilute sphere suspensions more effective at reducing the viscosity of viscoelastic liquids at larger De and c.

Wall-pressure fluctuations beneath turbulent boundary layers drive noise and structural fatigue through interactions between fluid and structural modes. Conventional predictive models for the spectrum--such as the widely accepted Goody model--fail to capture the energetic growth in the subconvective regime that occurs at high Reynolds number, while at the same time over-predicting the variance. To address these shortcomings, two semi-empirical models are proposed for the wall-pressure spectrum in canonical turbulent boundary layers, pipes and channels for friction Reynolds numbers $\delta^+$ ranging from 180 to 47 000. The models are based on consideration of two eddy populations that broadly represent the contributions to the wall pressure fluctuations from inner-scale motions and outer-scale motions. The first model expresses the premultiplied spectrum as the sum of two overlapping log-normal populations: an inner-scaled term that is $\delta^+$-invariant and an outer-scaled term whose amplitude broadens smoothly with $\delta^+$. Calibrated against large-eddy simulations, direct numerical simulations, and recent high-$\delta^+$ pipe data, it reproduces the convective ridge and the emergence of a sub-convective ridge at large $\delta^+$. The second model, developed around newly-available pipe data, uses theoretical arguments to prescribe the spectral shapes of the inner and outer populations. By embedding the $\delta^+$ dependence in smooth asymptotic functions, it yields a formulation that varies continuously with $\delta^+$. Both models capture the full spectrum and the logarithmic growth of its variance, laying the groundwork for more accurate engineering predictions of wall-pressure fluctuations.

Identifying the location and characteristics of pollution sources in turbulent flows is challenging, especially for environmental monitoring and emergency response, due to sparse, stochastic, and infrequent cue detection. Even in idealized settings, accurately modeling these phenomena remains highly complex, with realistic representations typically achievable only through experimental or simulation-based data. We introduce TURB-Smoke, a cutting-edge numerical dataset designed for investigating odor and contaminant dispersion in turbulent environments with and without mean wind. Generated via direct numerical simulations of the fully resolved three-dimensional Navier-Stokes equations, TURB-Smoke tracks hundreds of millions of Lagrangian particles released from five distinct point sources in fully developed turbulence, thus providing a reliable ground-truth framework for developing and evaluating source-tracking strategies using stationary sensors or mobile agents in realistic flows. Each particle's trajectory is continuously tracked on many characteristic turbulence timescales, recording both the position and the local flow velocity. Additionally, we provide coarse-grained concentration fields in 3D and in quasi-2D slabs containing the source, ideal for quickly testing and optimizing search algorithms under varying flow conditions.

The interaction between acoustic waves and turbulent grazing flow over an acoustic liner is investigated using Lattice-Boltzmann Very-Large-Eddy simulations. A single-degree-of-freedom liner with 11 streamwise-aligned cavities is studied in a grazing flow impedance tube. The conditions replicate reference experiments from the Federal University of Santa Catarina. The influence of grazing flow (with a centerline Mach of 0.32), acoustic wave amplitude, frequency, and propagation direction relative to the mean flow is analysed. Impedance is computed using both the in-situ and the mode-matching methods. The in-situ method reveals strong spatial variations; however, averaged values throughout the sample show minimal differences between upstream and downstream propagating waves, in contrast to the mode-matching method. Flow analyses reveal that the orifices displace the flow away from the face sheet, with this effect amplified by acoustic waves and dependent on the wave propagation direction. Consequently, the boundary layer displacement thickness (${\delta}$*) increases along the streamwise direction compared to a smooth wall and exhibits localised humps downstream of each orifice. The growth of ${\delta}$* alters the flow dynamics within the orifices by weakening the shear layer at downstream positions. This influences the acoustic-induced mass flow rate through the orifices, suggesting that acoustic energy is dissipated differently along the liner. The role of near-wall flow features highlights the need to consider a spatially evolving turbulent flow when studying the acoustic-flow interaction and measuring impedance. The spatial development of the turbulent flow may also partly explain the upstream-downstream impedance differences, as current eduction methods do not account for it.

The results of direct numerical simulation of plane-symmetric turbulence of water waves for potential flows within the framework of conformal variables taking into account low-frequency pumping and high-frequency viscous dissipation are presented. In this model, for a wide range of pumping amplitudes, the weak turbulence regime was not detected. It is shown that for typical turbulence parameters, the main effects are the processes of wave breaking, the formation of cusps on wave crests, which make the main contribution to the turbulence spectra with a dependence on frequency and wavenumber with the same exponent equal to $-4$. In this strongly nonlinear regime, the probability density of wave steepness at large deviations has power-law tails responsible for the intermittency of turbulence.

Droplets coalescing on a superhydrophobic surface exhibit coalescence-induced droplet jumping. However, water vapor condensing on a superhydrophobic surface can result in simultaneous formation of condensate droplets with two distinct wetting states, cassie state (CS) and partially wetting (PW) state. Droplets in PW state exhibit high contact angle but are connected to the substrate though a thin liquid condensate column. Coalescence between CS and PW droplets has been recently identified as a possible mechanism for generating droplets exhibiting in-plane roaming motion during dropwise condensation on nanotextured superhydrophobic surfaces. Here, we systematically investigate this phenomenon through experiments on coalescence between sessile droplets in CS and PW state on a nanostructured superhydrophobic surface endowed with a micro-scale hydrophilic spot. Here, a sessile droplet carefully placed on the hydrophilic spot simulates the PW state. Overall, our investigations demonstrate that when a CS droplet coalesces with a PW droplet pinned to a hydrophilic defect, the interaction generates substantial in-plane momentum. We find that when the coalescing CS and PW droplets are nearly of the same size and about ~3 to ~3.5 times the size of the hydrophilic spot pinning the PW droplet, the vertical momentum generation is nearly completely suppressed, and the resulting maxima in in-plane momentum results in detachment of merged droplet from hydrophilic spot and its subsequent in-plane motion.

The case of a conventional flexible sheet in a uniform flow has been of interest in understanding the underlying physics of passive coupled dynamics between a flexible structure and a flow field. Gravity is known to influence the flapping instability and post-critical dynamics. Interestingly, the flapping instability and dynamics of a thin flexible structure have been investigated either by neglecting the effects of gravity or by considering gravity along the length/span of the sheet. This study experimentally investigates the self-induced and sustained flapping dynamics of a thin flexible sheet positioned horizontally, with gravity acting along its bending direction. To explore the coupled interplay between gravitational, aerodynamic, and structural effects on the onset of instability and post-critical flapping dynamics, wind tunnel experiments are conducted across a range of physical parameters of the flexible sheet, such as its length (L), and aspect-ratio ($\AR$) for different wind speeds. To further understand the effects of gravity, the flapping behaviour of a vertically mounted flexible sheet with gravity acting along its span has been investigated, and comparisons have been drawn with its horizontal counterpart. It has been observed that gravity along the bending does not influence the onset of flapping instability. The horizontally mounted flexible structures exhibit higher flapping amplitudes and frequencies when compared to their vertical counterparts. The observations in this study have direct relevance in the field of smart propulsion and energy harvesting devices.

Albeit the hemodynamics of artificial heart valves has been investigated for several decades, the local shear-induced activation potential and subsequent transport phenomena of activated platelets in different valve designs, which mediate thrombosis, remains poorly understood. Here, platelet activation due to local shear stresses and the associated transport phenomena are investigated in two designs of mechanical heart valves (MHVs), namely a trileaflet MHV (TMHV) and a bileaflet MHV (BMHV) and compared against a surgical bioprosthetic heart valve (BHV) as a control. It is observed that the local activation and transport of platelets in any aortic region reach a cyclic state, with MHVs showing higher levels of both activation and transport than BHV. When integrated over the volume of the aortic sinuses and central lumen, the local activation is, respectively, 5.90 and 2.26 times higher in BMHV whereas 2.97 and 1.39 times higher in TMHV than in BHV. The washout of activated platelets from the sinuses and central lumen is, respectively, 10.40 and 2.39 times higher in BMHV while 4.90 and 1.40 times higher in TMHV compared to BHV. The low washout of sinuses in BHV is also demonstrated by higher residence time in sinuses compared to MHVs. These findings indicate that the risk of clinical thrombosis in MHVs is likely due to higher levels of local shear-induced activation than BHV despite the lower residence time (i.e. a better washout). Conversely, the subclinical thrombosis in BHVs is probably due to prolonged platelet residence time relative to MHVs.

Based on mesoscale lattice Boltzmann numerical simulations, we characterize the Rayleigh-B\'enard (RB) convective dynamics of dispersions of liquid droplets in another liquid phase. Our numerical methodology allows us to modify the droplets' interfacial properties to mimic the presence of an emulsifier (e.g., a surfactant), resulting in a positive disjoining pressure that stabilizes the droplets against coalescence. To appreciate the effects of this interfacial stabilization on the RB convective dynamics, we carry out a comparative study between a proper emulsion, i.e., a system where the stabilization mechanism is present (stabilized liquid-liquid dispersion), and a system where the stabilization mechanism is absent (non-stabilized liquid-liquid dispersion). The study is conducted by systematically changing both the volume fraction, $\phi$, and the Rayleigh number, Ra. We find that the morphology of the two systems is dramatically different due to the different interfacial properties. However, the two systems exhibit similar global heat transfer properties, expressed via the Nusselt number Nu. Significant differences in heat transfer emerge at smaller scales, which we analyze via the Nusselt number defined at mesoscales, Nu$_{\mathrm{mes}}$. In particular, stabilized systems exhibit more intense mesoscale heat flux fluctuations due to the persistence of fluid velocity fluctuations down to small scales, which are instead dissipated in the interfacial dynamics of non-stabilized dispersions. For fixed Ra, the difference in mesoscale heat flux fluctuations depends non-trivially on $\phi$, featuring a maximum in the range $0.1 < \phi < 0.2$. Taken all together, our results highlight the role of interfacial physics in mesoscale convective heat transfer of complex fluids.

Quantum computation offers potential exponential speedups for simulating certain physical systems, but its application to nonlinear dynamics is inherently constrained by the requirement of unitary evolution. We propose the quantum Koopman method (QKM), a data-driven framework that bridges this gap through transforming nonlinear dynamics into linear unitary evolution in higher-dimensional observable spaces. Leveraging the Koopman operator theory to achieve a global linearization, our approach maps system states into a hierarchy of Hilbert spaces using a deep autoencoder. Within the linearized embedding spaces, the state representation is decomposed into modulus and phase components, and the evolution is governed by a set of unitary Koopman operators that act exclusively on the phase. These operators are constructed from diagonal Hamiltonians with coefficients learned from data, a structure designed for efficient implementation on quantum hardware. This architecture enables direct multi-step prediction, and the operator's computational complexity scales logarithmically with the observable space dimension. The QKM is validated across diverse nonlinear systems. Its predictions maintain relative errors below 6% for reaction-diffusion systems and shear flows, and capture key statistics in 2D turbulence. This work establishes a practical pathway for quantum-accelerated simulation of nonlinear phenomena, exploring a framework built on the synergy between deep learning for global linearization and quantum algorithms for unitary dynamics evolution.

This study employs large-eddy simulations with a flamelet progress variable approach to systematically quantify the influence of nozzle geometry on combustion efficiency, mixing, and blowout resistance in non-assist methane flares. Five canonical nozzle shapes-circle, low aspect ratio ellipse, high aspect ratio ellipse, diamond, and square-were evaluated under relevant industrial flare conditions. Results demonstrate that cornered geometries enhance near-field recirculation, promote mixing, and sustain flame attachment, resulting in up to a 5% improvement in combustion efficiency compared with streamlined nozzles. The square nozzle performed best irrespective of the wind direction (orientation) and maintained a combustion efficiency greater than 96.5% even at the highest tested crosswind velocities, while other streamlined designs exhibited early flame lift-off, reduced recirculation, and efficiency losses. Analysis of mixing and vorticity reveals that sharp-edged nozzles accelerate scalar homogenization and buffer flames against crosswind-induced strain, directly translating to increased blowout resistance.

Accurate and efficient modeling of cardiac blood flow is crucial for advancing data-driven tools in cardiovascular research and clinical applications. Recently, the accuracy and availability of computational fluid dynamics (CFD) methodologies for simulating intraventricular flow have increased. However, these methods remain complex and computationally costly. This study presents a reduced order model (ROM) based on higher order dynamic mode decomposition (HODMD). The proposed approach enables accurate reconstruction and long term prediction of left ventricle flow fields. The method is tested on two idealized ventricular geometries exhibiting distinct flow regimes to assess its robustness under different hemodynamic conditions. By leveraging a small number of training snapshots and focusing on the dominant periodic components representing the physics of the system, the HODMD-based model accurately reconstructs the flow field over entire cardiac cycles and provides reliable long-term predictions beyond the training window. The reconstruction and prediction errors remain below 5\% for the first geometry and below 10\% for the second, even when using as few as the first 3 cycles of simulated data, representing the transitory regime. Additionally, the approach reduces computational costs with a speed-up factor of at least $10^{5}$ compared to full-order simulations, enabling fast surrogate modeling of complex cardiac flows. These results highlight the potential of spectrally-constrained HODMD as a robust and interpretable ROM for simulating intraventricular hemodynamics. This approach shows promise for integration in real-time analysis and patient specific models.

We present diffSPH, a novel open-source differentiable Smoothed Particle Hydrodynamics (SPH) framework developed entirely in PyTorch with GPU acceleration. diffSPH is designed centrally around differentiation to facilitate optimization and machine learning (ML) applications in Computational Fluid Dynamics~(CFD), including training neural networks and the development of hybrid models. Its differentiable SPH core, and schemes for compressible (with shock capturing and multi-phase flows), weakly compressible (with boundary handling and free-surface flows), and incompressible physics, enable a broad range of application areas. We demonstrate the framework's unique capabilities through several applications, including addressing particle shifting via a novel, target-oriented approach by minimizing physical and regularization loss terms, a task often intractable in traditional solvers. Further examples include optimizing initial conditions and physical parameters to match target trajectories, shape optimization, implementing a solver-in-the-loop setup to emulate higher-order integration, and demonstrating gradient propagation through hundreds of full simulation steps. Prioritizing readability, usability, and extensibility, this work offers a foundational platform for the CFD community to develop and deploy novel neural networks and adjoint optimization applications.