2026 Archive

Abstract: Nonlinear aeroelastic flutter remains an active research topic due to the complexity associated with the coupling between structural dynamic responses and unsteady aerodynamic responses. Limit-cycle oscillation (LCO) is one such aeroelastic phenomenon in which a structure oscillates at a bounded amplitude. At different freestream speeds, the initial structural displacement of an aeroelastic system may either decay without LCO, or the system may reach an LCO state characterized by sustained oscillations at a nearly constant amplitude. The minimum freestream speed at which sustained LCO occurs is often referred to as the flutter speed. Previous research has shown that both structural properties and aerodynamic nonlinearities affect LCO kinematics. Therefore, modeling efforts for LCO prediction must incorporate accurate structural properties into the equations of motion, while simultaneously capturing aerodynamic nonlinearities during the unsteady process.

In this ongoing research effort, we explore the use of the discrete vortex method (DVM) with leading-edge vortex (LEV) modeling capabilities for aeroelastic simulations. The aeroelastic system is modeled as a rigid wing with pitch and heave degrees of freedom, represented as a coupled mass-spring-damper system. The DVM is modified to incorporate a simple finite-wing-effect correction and is coupled with the structural equations of motion. This framework dynamically connects the structural and aerodynamic models, enabling comprehensive simulations of the aeroelastic system. Data from a previous experimental study are used to identify the structural parameters of the target system and to validate the simulation results. The aeroelastic simulations using DVM show promising results, capturing both the flutter speed and the LCO behavior observed in the experiments. These initial findings demonstrate the strong potential of low-order DVM for rapid aeroelastic simulations.

Abstract: Delta wings are fundamental to high-performance aircraft like the Eurofighter Typhoon, Saab Gripen, and supersonic transports like Concorde. A defining feature of the flow over a delta wing is the formation of counter-rotating vortices along its leading edges. While Reynolds number and sweep angle influence flow dynamics, the angle of attack (AoA) is a primary driver. Previous hot-wire anemometry research identified three wake regimes: (a) attached vortices below 45°, (b) simultaneous shedding near 60°, and (c) alternate shedding near 75°. However, hot-wire measurements are limited by their single-point nature, failing to resolve three-dimensional, instantaneous structures or the specific frequencies driving these modes. This study addresses these gaps by investigating the angle of attack effects on a  non-slender delta wing to identify the coherent flow structures associated with each distinct shedding regime.

To achieve these objectives, 3D Lagrangian particle tracking measurements were conducted in the near wake at AoA of 45°, 60°, and 75°. At 45°, time-averaged flow reveals two well-defined, counter-rotating vortices, confirming an attached vortex regime. At 60° and 75°, the wake expands significantly and time-averaged coherent structures fade, suggesting intermittent shedding from the wing edges. Spectral Proper Orthogonal Decomposition (SPOD) shows that while most energy resides in the first spatial mode for both high angles, the dominant frequency is notably higher at 60° than at 75°. Structurally, shedding occurs at both the leading and trailing edges at 60° but is primarily confined to the leading-edge at 75°. Flow reconstruction at the dominant frequency reveals in-phase velocity signals at 60° and out-of-phase signals at 75°, confirming simultaneous and alternate shedding modes, respectively.

These findings provide new insights into the time-averaged and instantaneous three-dimensional flow fields in the delta wing wake, as well as the specific frequencies and spatial structures that characterize them. The upcoming talk will further detail these aspects, alongside the nonlinear interactions and energy transfer between different frequencies within the flow field.

Abstract: Bluff bodies with fixed, sharp trailing edges that pierce the interface are ubiquitous in naval architecture and offshore structures, where the wake governs loading, surface signature, and sediment transport. At high Reynolds numbers, the turbulent flow near the unsteady free surface drives complex transport processes. The attenuation of Karman vortex shedding in the near-surface wake of surface-piercing bodies has been studied extensively for circular cylinders. Large eddy simulation and planar PIV attribute the attenuation to a free surface induced lateral gradient in the normal stresses, which enhances a counter-rotating streamwise vortex pair, widening the wake and inhibiting interaction between the shear layers (Suh, Yang, and Stern, 2011). A fundamental understanding of the physics for fixed separation point bodies is critical for the design and operation of a broad range of engineering applications.

Here we present wake flow measurements of two surface-piercing geometries, a triangular wedge and an extended slender wedge with a tripped turbulent boundary layer, in the subcritical Reynolds and Froude number regimes. The flow was characterized using fluorescent dye visualization, stereoscopic PIV, and color-encoded dual-layer planar PIV. The canonical vortex shedding was consistently delayed and attenuated near the free surface, and the affected region extended significantly deeper than the 1D cutoff. A streamwise counter-rotating vortex pair was again identified in the near wake, and analysis of the time-averaged vorticity transport equation indicates a tilting and stretching mechanism. The findings lay the foundation for high-fidelity anisotropic modeling of turbulent wakes. 

Abstract: The hydrodynamic impact of school size in dense fish schools is investigated using canonical configurations, including long (inline), wide (lateral), and diamond arrangements. High-fidelity simulations are performed using a two-dimensional immersed boundary method-based flow solver on a Cartesian grid, coupled with tree-topological local mesh refinement (TLMR) to efficiently resolve multiscale vortex interactions around multiple swimmers. Swimming performance is evaluated in terms of thrust generation and propulsive efficiency as a function of school size and spatial arrangement. The study systematically examines how increasing the number of swimmers alters flow structures, wake interactions, and performance distribution within the fish school. Rather than exhibiting uniform gains, the effect of school size is found to depend strongly on configuration.

In the long configuration, increasing school size leads to diminishing returns in thrust generation, indicating a saturation of beneficial wake interactions in extended inline formations. In the wide configuration, performance varies across the school: upstream individuals experience reduced benefits, while downstream individuals gain from favorable flow conditions, resulting in limited overall improvement. In contrast, the diamond configuration supports more uniform hydrodynamic advantages, maintaining balanced performance across individuals as school size increases.

These results highlight that the hydrodynamic benefits of schooling are governed not only by group size but, more critically, by spatial organization and inter-individual flow coupling, providing insight into the role of configuration in collective swimming dynamics and bio-inspired design.

Abstract: Convolutional neural networks (CNNs) are increasingly used to model fluid flows, yet studies on how to design these networks for unsteady flows with moving boundaries—such as flapping wings—remain scarce. In practice, architecture choices have been guided largely by intuition and empirical heuristics rather than by a clear understanding of how these networks process flow data. This limitation becomes especially important in moving-boundary problems, where time-varying discontinuities and complex vortex interactions make effective network design far from straightforward. To address this, we use interpretability techniques to uncover how CNNs learn from flow data, and use those insights to guide principled model design and optimization. We introduce a new saliency-mapping method based on the Convolutional Neural Feature Matrix (CNFM), which provides a layer-by-layer view of what a CNN learns from flow fields. The method not only shows which architectural choices perform well, but explains why they work, enabling targeted redesign instead of blind trial and error. We demonstrate this approach on two tasks: aerodynamic load estimation from instantaneous flow fields and flow reconstruction with an autoencoder, for a pitching-plunging airfoil at Re=300, across three phase offsets spanning distinct wake topologies in both periodic and quasi-periodic regimes. The results make the case for interpretability-guided design. With CNFM-guided redesign, drag coefficient prediction errors fall from 28% to around 2% on unseen quasi-periodic cases; autoencoder reconstruction errors drop from 60% to approximately 5%. Overall, this work demonstrates that interpretability can provide principled guidance for designing efficient, explainable neural networks, with broader applicability beyond CNNs to architectures such as transformers.

Abstract: High-aspect-ratio flexible wings offer clear aerodynamic benefits but can enter post-critical aeroelastic regimes where structural deformation and unsteady separated flow become strongly coupled. In these states, structural responses can transition from mono-frequency limit cycle oscillations (LCO) to multi-frequency dynamics involving coupled bending, edgewise, and torsional modes, complicating state-aware control. This work combines comprehensive experiments with a cluster-based reduced-order modelling (CROM) framework to quantify how operating conditions govern these responses and to extract objective, low-dimensional flow states from time-resolved measurements. A flexible wing was tested across varying flow velocities and angles of attack. Digital image correlation (DIC) demonstrates that these conditions shift LCO onset and alter dominant frequencies and mode shapes. Concurrently, particle tracking and image velocimetry (PTV/PIV) reveal that leading- and trailing-edge vortical structures dynamically adapt to the active structural mode. To translate this high-dimensional data into interpretable dynamics, CROM partitions the flow into representative states to build probabilistic transition networks. Analyzing a mono-frequency LCO (second bending) and a multi-frequency response (coupled bending/edgewise) reveals that structural complexity dictates the necessary ROM dimensionality: five clusters resolve the single-mode dynamics, while six fully capture the multi-frequency cycle. By linking operating conditions to modal content and objective flow pathways, this unified framework delivers a robust low-dimensional basis for monitoring and controlling highly flexible wings in post-flutter regimes.

Abstract: A parallel optimization algorithm is developed for various bio-inspired flow problems, including schooling formation, phase, and shape optimization. This surrogate-based approach maintains a training cost comparable to that of a standard Gaussian process (GP), while exhibiting non-stationary behavior to effectively handle the complex solution space of bio-inspired flows. The algorithm demonstrates excellent parallel scalability, ensuring that even with large-scale parallel sampling, the sampling strategy remains highly efficient and the number of required Computational Fluid Dynamics (CFD) simulations is kept relatively low. Although bio-inspired flow problems typically require high-fidelity CFD simulations to resolve detailed flow physics, the low training cost of the surrogate model, combined with efficient parallel sampling, enables the overall optimization to be completed within a short time frame.

Abstract: This study presents high-fidelity computational fluid dynamics (CFD) simulations of wingtip vortices and the development of a new model for the associated turbulence intensity. Three simulations are performed using Large Eddy Simulation (LES), fully turbulent Improved Delayed Detached Eddy Simulation (SST-IDDES), and transitional GT-IDDES. Applied to a finite wing with a NACA0012 airfoil, the transitional IDDES matches LES in predicting wingtip vortex trajectory, diffusion, axial deficit, swirl velocity, turbulence intensity, and Reynolds shear stress, whereas the fully turbulent SST-IDDES exhibits over-diffusive and Reynolds-Averaged Navier-Stokes (RANS)-dominated behavior, which results in overpredictions of the core radius and vortex positions. An asymmetric distribution of turbulence components is observed in the vortex at a small wake age, while the distribution becomes axisymmetric at approximately five chord lengths downstream of the trailing edge. Finally, based on the CFD results, a new semi-empirical vortex turbulence-intensity model is developed, which accurately captures the turbulence-intensity distribution at the core radius and beyond, as well as the downstream evolution of vortex turbulence. 

Abstract: Flow-induced vibrations (FIV) occur in a range of engineering applications, including mooring lines, pylons, and chimneys, where the suppression of these vibrations is paramount to structural fatigue mitigation. In contrast, renewable energy harvesting technologies seek to enhance these vibrations. Thus, a fundamental understanding of FIV is essential for the design of critical infrastructure and enabling emerging energy-harvesting technologies. 

Many studies have focused on the FIV of 2D bluff bodies such as circular cylinders. A singular elastically mounted cylinder is not subject to persistent vibrations beyond reduced velocities of U^*=U/f_{nw}D~12, where U is freestream velocity, f_{nw} is the natural frequency in water and D is body diameter. However, when placed in tandem behind a static cylinder, it can exhibit persistent monotonically increasing vibration response with increasing reduced velocities (Assi et al., JFM, 2010, 2013). An isolated sphere, on the other hand, has been reported to undergo persistent vibrations until U^*~300 (Jauvtis & Williamson, JFS, 2001). However, there is no published study yet investigating the FIV response of tandem spheres.

This talk will discuss the behavior of a spanwise elastically mounted sphere placed in tandem with an upstream static sphere of equal diameter. A series of systematic experiments are performed employing simultaneous displacement and wake measurements using a linear encoder and stereo article image velocimetry, respectively. The FIV response is characterized for a wide range of reduced velocities 2.5<U^* <22.5 and spacing ratios 1.5<L^* = L/D<10, where L is the separation distance between the two spheres . 

It is observed that at low reduced velocities (U^* < 10.5), an attenuated vibration response is observed for spacing ratios L^* <4, which presumably lies within the wake-deficit of the static sphere, thereby reducing the mean flow experienced by the sphere. Interestingly, at higher reduced velocities of U^* >10.5, vibrations monotonically increase with increasing reduced velocities for all spacing ratios. The vibrations were regularized and enhanced by up to 153% compared to an isolated sphere. These enhanced vibrations do not appear to be driven by streamwise vorticity known to drive vortex-induced vibrations of an isolated sphere. These vibrations are termed wake-enhanced vibrations and are characterized by significant energy transfer from fluid to the structure. This study advances our understanding of fluid-structure interaction mechanisms of tandem 3D bodies and motivates further work to understand the origins of fluid energy transfer to the system.

Abstract: A novel hierarchical decomposition for wake flows is introduced, enabling a nonlinear modeling framework for dissipation. The approach combines proper orthogonal decomposition with the maximal overlap discrete wavelet transform, yielding a method we denote as wPOD. The decomposition serves two purposes: (i) it provides a coherent reconstruction in space and time that isolates the shedding frequency and its harmonics, yielding modes whose temporal evolutions are narrowband and centered at the shedding frequency, its higher harmonics, and low-frequency components; and (ii) it extracts a representation of the dissipative small scales that remains nonlinearly correlated with the coherent motion. This nonlinear correlation reflects the well-known phenomenon of small scales riding on coherent structures and therefore sharing their convective velocity. Accordingly, both motions can be represented as convective waves of the form 

R(t)cos(2πft−k⋅x). For the large scales, the phase, cos(2πft−k⋅x), sustains the convective velocity Uc=f/kx, while the envelope R(t) varies slowly in time. For the small scales, the phase is irregular, whereas the envelope enforces the same convective velocity. By distinguishing these mechanisms, we characterize the dynamical coupling between coherent and dissipative scales and establish a physically grounded basis for nonlinear modeling of dissipation in wake vortex flows. 

Abstract: Fluid-structure interactions between elastically mounted pitching plates and uniform flow produce a diverse set of aeroelastic responses, including static and dynamic instabilities, limit-cycle oscillations, and chaotic dynamics. In these systems, large-amplitude oscillations often occur at low flow speeds and coexist with vortex-dominated wake dynamics.

This talk presents cyber-physical investigations of an inverted flag aeroelastic system consisting of a rigid plate mounted on a trailing-edge torsional spring and aligned parallel to a uniform freestream flow. Aerodynamic loads are imposed physically in a subsonic wind tunnel (Re ~10^5), while structural loads are imposed artificially using an electromechanical control loop. This framework allows for exploration of complex nonlinear structures with variable damping levels, ultimately leading to the design and optimization of a fluid-oscillator energy-harvesting platform.

Results are presented through bifurcation diagrams, reconstructed phase-plane portraits, and flow visualizations. Both linear and nonlinear spring designs are examined. Linear spring experiments detail distinct straight, flapping, and deflected flow regimes and serve as a baseline case for validating the cyber-physical framework. Hardening spring designs extend dimensional excitation regimes, providing guidance for future energy-harvesting efforts. Lastly, cubic spring designs facilitate a targeted investigation into underlying excitation mechanisms by removing the straight flow regime observed in the baseline experiments. 

Abstract: How can you ensure a balance between accuracy and performance when doing adaptive direct-numerical-simulations of complex moving geometries? While classical adaptive codes usually focus on heuristic measures as an error indicator, this presentation explains the benefit of using the wavelet decomposition for multi-resolution analysis as well as error control and how we use it to resolve the turbulent structures shed by flapping insect flight. Our solver WABBIT solves the incompressible Navier-Stokes equations by continuously ensuring a sufficient representation of all necessary scales at each time-step, the domain is discretized on a block-based octree grid and the solution evolved with a higher order projection method. The ansatz and its robustness will be validated for a challenging inviscid, incompressible test-case of crashing vortex tubes forming a near-singularity, portraying how the adaptation can be adjusted to accuracy requirements. The results on the finest grid with an effective resolution of 8192^3 match well with the pseudo-spectral one from Hou&Li, while maintaining a grid compression lower than 0.2%. Together with the volume penalization method, the adaptive framework allows for a flexible approach to easily deploy efficient simulations for different insects and flight scenarios.