14:20   Experimental and numerical data assimilation II Low-order modelling of the wake of an experimental actuated fluidic pinball Alicia Rodríguez-Asensio, Luigi Marra, Juan Alfaro-Moreno, Andrea Ianiro, Stefano Discetti Abstract: Introduction Turbulent flows are inherently chaotic and complex, yet they often display repeating patterns known as coherent structures. These structures hint that, despite their high dimensionality, the essential behavior of a flow is governed by a relatively simple underlying structure and can be captured in a low-dimensional space. Finding these low-dimensional representations helps us to understand and control flows to tackle engineering and physical challenges like reducing drag or smoothing out unsteady loads. Several data-driven techniques exist to find these low-dimensional representations: Proper Orthogonal Decomposition (POD) decomposes the flow into modes ranked by their energy content; Multidimensional Scaling (MDS) embeds a set of snapshots into a low-dimensional space while preserving their pairwise distances; and more recently, manifold-learning methods like isometric mapping (ISOMAP) have revealed even sharper compression by finding a curved surface (manifold) where the flow data lie on. The fluidic pinball, a simple arrangement of three rotating cylinders in a uniform flow, offers a versatile test case for these ideas. Simulations have revealed a so-called actuation manifold \cite{marra2024actuation} that links cylinder rotation commands to the resulting wake states over a wide range of operating conditions. This five-dimensional manifold captures both the global flow patterns and the key force coefficients, making it a powerful tool for model-based control and flow estimation. Additionally, Farzamnik et al. demonstrated that manifold-learning techniques like ISOMAP can compress flow phenomena into a few coordinates that are directly related to physical quantities like force coefficients and vortex-shedding phases. In order to test this approach in a turbulent regime, in this work we present an experimental demonstration. A water-tunnel setup has been designed to measure forces and velocity fields simultaneously of the fluidic pinball. These synchronized measurements are used to build an actuation manifold, whose capabilities to describe the flow structures and force coefficients in a lab-scale experiment are then tested. Flow configuration The fluidic pinball (Figure 1) consists of three identical cylinders of radius R arranged as the vertices of an equilateral triangle of side 3R in a uniform incompressible flow at speed U_\infty. The upstream vertex points into the flow, while the base of the triangle sits orthogonal to it. Each cylinder can spin independently, providing three continuous actuation inputs. The non-dimensional rotation speeds of the front, top, and bottom cylinders are denoted by b_1, b_2, and b_3, respectively. Each b_i is defined as the cylinder’s surface tangential speed normalized with U_\infty. This simple geometric layout produces a rich set of flow regimes. At low Reynolds number, the baseline (unforced) flow undergoes a sequence of bifurcations, from a steady solution, through periodic shedding, and on to chaotic wake dynamics. By varying the three rotation rates, one can reproduce key actuation mechanisms like boat-tailing, the Magnus lift effect, and modified stagnation-point behavior. The fluidic pinball’s combination of geometric simplicity, multiple independent controls, and a wide range of dynamic responses has led to explore various control strategies in an experimental scale. Recently, different control strategies involving the experimental fluidic pinball have been tested. Experimental test campaign All tests are carried out in a water tunnel whose section is 500\ \mathrm{mm} \times 550\ \mathrm{mm} at a free‐stream velocity of U_\infty \approx 0.15\ \mathrm{m/s}. The Reynolds number based on the cylinder diameter D = 2R = 30\ \mathrm{mm} is \[ \mathrm{Re}_D = \frac{U_\infty D}{\nu} \approx 4500\,, \] where \nu is the kinematic viscosity of water. The blockage ratio produced by the interference of the walls with the experimental model is approximately 14\%. During the experiments, both the flow field and the forces on the cylinders are measured simultaneously for reconstructing and validating the actuation manifold under real conditions. Velocity snapshots are the input of an ISOMAP embedding, yielding low‑dimensional coordinates directly from the laboratory data. At the same time, synchronized force readings are paired with each velocity field. This dual dataset enables the reconstruction of the five‑dimensional manifold from experiment and the demonstration that its coordinates predict both wake patterns and force outputs, just as in the original simulations, even in the presence of measurement noise and at higher Reynolds numbers. To cover the full actuation space of the manifold described by Marra \textit{et al.}, each control parameter b_1, b_2, and b_3 is varied independently over the range [-3, 3]. The sampling density is chosen so that it ensures that the experimental dataset provides the breadth needed to reconstruct the actuation manifold. For the velocity field, planar Particle Image Velocimetry (PIV) is employed in a plane just downstream of the cylinders and perpendicular to them (see Figure 1). A thin laser sheet illuminates microscopic particles seeded in the water, and a high-speed camera captures their motion. By comparing pairs of images taken at 70 Hz, the local velocity vectors are computed. The PIV configuration (window size, resolution, and frame rate) is chosen to resolve both the small vortices close to the cylinders and the larger wake structures downstream. The forces on each cylinder are measured using three one-dimensional load cells mounted flush with the cylinder supports. By employing three sensors, any coupling between force measurements is eliminated and also the out-of-plane (z-direction) momentum is captured, even though that value is not actively used. Each load cell measures force along a single axis, and their combined readings provide an uncoupled, high-fidelity record of lift and drag in real time as the cylinders rotate under different actuation inputs. The load-cell signals are sampled at a frequency of 70 Hz to capture rapid fluctuations in the forces associated with vortex shedding and other unsteady effects. To ensure that each velocity snapshot corresponds exactly to a force measurement, the PIV camera, the laser and the force data acquisition system are triggered from a single master clock. This precise synchronization allows to pair each velocity field with its instantaneous force coefficients. By aligning the datasets in time, the reduced-order actuation manifold can be validated by showing that the model’s predicted force responses match those derived from the synchronized experimental measurements. Figure 2 shows the three-dimensional model of the test rig, comprising a rigid base plate that holds three independently driven cylinders via bearing blocks and stepper-motor assemblies. Support bars and alignment plates ensure geometric accuracy. Load-cell mounts are integrated directly into the cylinder support plate to minimize mechanical coupling. The experimental setup is now fully assembled and operational. While the system is undergoing final adjustments, such as refining its placement in the water tunnel to ensure symmetry in the flow, initial tests confirm its robustness and readiness for detailed measurements. By resolving minor issues, we aim to soon begin producing experimental data under a range of cylinder rotation settings for manifold modelling. Once aligned, we will compare time-resolved manifold coordinates from experiment against simulation. Although the actuation manifold was developed and tested in clean, low-Reynolds-number simulations, it is expected to perform equally well in the experiment, even at higher Reynolds numbers and with real measurement noise. Manifold-learning techniques are designed to capture the dominant flow patterns and their link to force outputs, making them robust against disturbances. Validating this in the laboratory will confirm that data-driven reduced-order models could be applied for practical flow-control applications. Wind Tunnel Measurements on Fixed-Wing UAV Wings for the Validation of an Automated CFD Framework Konstantinos Kellaris, Marinos Manolesos, Giorgos Efrem, Pavlos Kaparos - Tsafos, Pericles Panagiotou Abstract: This study presents a validation of an automated CFD framework for the aerodynamic analysis of fixed-wing UAVs, through a comparative investigation with wind tunnel measurements. Experiments were conducted in a low-speed closed-circuit WT using three representative wing geometries used in both conventional and BWB UAV configurations. Force measurements and subsequent flow visualization techniques were employed. The CFD simulations demonstrated good agreement with experimental results. However, discrepancies in lift slopes were observed and discussed. The results affirm the CFD framework's robustness and underscore the importance of enhanced turbulence and transition modeling for accurate aerodynamics assessment in UAV design. On the flow behind the F1 Imperial Front Wing, comparison between volumetric PTV and numerical simulations. Isabella Fumarola, Alexandra I. Liosi, Parv Khurana, Isaac Balbolia, Adam Meziane, Spencer J. Sherwin, Jonathan Morrison Abstract: Abstract The recent development of Particle Tracking Velocimetry (PTV) for large-volume investigation in air has made this technique particularly well-suited to validate numerical simulations of complex aerodynamic configurations. At the same time, Large Eddy Simulations (LES) using high-order spectral h/p element methods have proven to be a powerful tool for investigating flows around complex geometries at high Reynolds (Re) numbers. In a recent study Liosi et al. [1] have successfully implemented this numerical methodology to analyse the flow behind a Formula 1 wing. However, the validation of their simulation was based on limited experimental data, mostly pointwise velocity measure, pressure surveys, or a limited number of PIV planes. In this work, the advantages of using volumetric PTV to validate numerical simulations is explored by presenting an experiment on the Imperial Front Wing (IFW) of the same geometry used by Liosi et al. [1]. The test was performed in the 10x5 wind tunnel over a moving ground (rolling road). Results show excellent agreement with the numerical prediction revealing new insights into the development of the vortical structures shed from the different wing elements. The work offers a unique dataset of significant value for the Formula 1 community. Extended Abstract attached.

end %-->