Multi-intensity-layer particle-image velocimetry for spraymeasurement

In this research, unsteady particle image velocimetry (PIV) data are reconstructed / analyzed by a proper orthogonal decomposition (POD) with a cross validation approach. At unsteady PIV flow measurements we have interests, it is sometimes difficult to gather the PIV velocity vector in its entire flow field, mainly due to the lack of particles and/or low (even) contrasting density in the flow field. A POD-based gappy data reconstruction approach is applied to recover the gappy PIV data. The appropriate numbers of iterations as well as considering POD modes are determined by the cross validation approach. The PIV measurements are performed for the wake of a rectangular cylinder as well as for a self-developed flapping wing object. Especially in the case of the flapping wing object, the PIV measurement is performed at a hovering condition in which it is difficult to scatter the particles appropriately in its entire flow field. It results in the failures of PIV measurement at some regions of flow field, and then those incomplete PIV data are reconstructed by the developed approach. The flow/vortex structures can be discussed reasonably well by using the recovered PIV data.

Proper Orthogonal Decomposition (POD) is performed on Large Eddy Simulation (LES) data and Particle Image Velocimetry (PIV) data from an underexpanded axisymmetric jet. PIV measures the streamwise and radial velocity components along a streamwise plane while LES provides data for the full domain and includes all flow variables (three velocity components, density, pressure, and temperature). A qualitative comparison between the POD modes from the LES and PIV data shows features similar in wavelength and shape particularly just downstream of the nozzle exit. An energy analysis shows that the biggest difference between the two cases was near the nozzle exit and believed to be the low turbulence level near the nozzle exit in LES data. At downstream positions, both cases have a similar distribution of POD modal energy. The analysis also shows that the large-scale flow features captured by the POD modes are strongly dependent on the domain size. Small domains are good for extracting the smaller-scale and low energy flow features near the nozzle exit. Further downstream the smaller domain spatially filters the largest flow features from the POD modes. The LES data is further post-processed to see if the streamwise and/or radial velocity can be used to estimate the tangential velocity, pressure, or density. It is found that the streamwise velocity can predict the pressure field, the radial velocity can predict the tangential velocity, and neither component does a good job at predicting the density. The prediction of flow variables is most likely due to the under-expanded nature of the jet and shows how PIV data can be used to further estimate other flow variables.

We present a new approach to simulating unsteady fluid flows, with only very few degrees of freedom, by employing directly eigenmodes extracted from digital particle image velocimetry experimental data. In particular, we formulate standard Galerkin and nonlinear Galerkin approximations of the incompressible Navier–Stokes equations using hierarchical empirical eigenfunctions extracted from an ensemble of flow snapshots. We demonstrate that standard Galerkin approaches produce simulations capable of capturing the short–term dynamics of the flow, but nonlinear Galerkin projections are more effective in capturing both the short– and long–term dynamics, leading to bounded solutions. These findings are documented by applying these approaches to flow past a stationary circular cylinder at Reynolds number 610.

In this study, we introduce light-actuated electrothermal (ET) flow as a means for fluid mixing in lab-on-a-chip (LOC) systems. The study begins with physically understanding the light-induced heating of an electrode surface involved in the generation of the ET flow. Four electrodes are chosen arbitrarily, and the change in the light-actuated ET flow velocity over each of the electrodes is measured by particle image velocimetry (PIV). The selected electrodes contain the following: silver (Ag), gold (Au), aluminum (Al), and indium tin oxide (ITO). The PIV data are analyzed based on the Beer–Lambert law and the thermal properties of the electrodes. The analysis confirms that the light-induced heating occurs effectively on electrode materials with a high optical absorption rate and low thermal conductivity. The use of the PIV is extended to investigate the dependence of light-actuated ET flows on electrical and thermal parameters. The increase in the electric potential and light intensity causes a parabolic and linear change in the ET velocity, respectively. At AC frequencies below the charge relaxation frequency of the fluid used, the ET velocity exhibits a plateau except for the frequency region where electroosmosis occurs. Based on the understanding of the flow characteristics, the mixing of two aqueous solutions by the light-actuated ET flow is attempted. This demonstrates that the ET flow generated in the form of a toroidal vortex can significantly improve the mixing length and mixing efficiency of two different fluids in a microchannel.

<div class="htmlview paragraph">Particle Image Velocimetry (PIV) was used to make gas velocity and turbulence measurements in a motored diesel engine. The experiments were conducted using a single-cylinder version of the Caterpillar 3406 production engine. One of the exhaust valves and the fuel injector port were used to provide optical access to the combustion chamber so that modifications to the engine geometry were minimal, and the results are representative of the actual engine. Measurements of gas velocity were made in a plane in the piston bowl using TiO<sub>2</sub> seed particles. The light sheet necessary for PIV was formed by passing the beam from a Nd:YAG laser through the injector port and reflecting the beam off a conical mirror at the center of the piston. PIV data was difficult to obtain due to significant out-of-plane velocities. However, data was acquired at 25° and 15° before top dead center of compression at 750 rev/min. Experiments were also conducted in a turbulent jet for comparison with previous jet measurements made with other techniques. It was found that high and low estimates of the turbulence intensity can be obtained from limited PIV data, and that length scales can be found by computing the turbulent dissipation and using kε turbulence model relations. In addition, the size of the eddies visible in high-pass filtered PIV data, and the size of vorticity concentrations were also found to be representative of the turbulence integral length scale. The engine PIV data was compared to flow and turbulence predictions from the multidimensional KIVA-3 code. Reasonably good agreement was found between measured and predicted turbulence intensity and length scales in the engine.</div>

Particle Image Velocimetry (PIV) has been used to map the complex flow field generated by simulated battle damage to a two-dimensional wing. Previous studies have relied on surface flow visualisation techniques to study the flow but here PIV data has enabled the flow field away from the surface to be analysed for the first time. Damage was simulated by a single hole with a diameter equal to 20% of the chord, located at mid-chord. Wind tunnel tests were conducted at a Reynolds number of 500,000 over a range of incidences from 0-10 with two-component PIV measurements made on three span-wise planes; on the damage centre line and o set by 0.5 and 1.0 hole radii. The PIV data was seen to be in good agreement with existing surface flow visualisation showing strong evidence of the formation of a horse shoe vortex, a counter-rotating vortex pair and reverse flow regions. Large variations in the flow structure were observed over the range of incidences tested as the jet transitioned from weak at lower angles to strong at higher angles. The data also revealed some significant differences in the flow compared to classic Jets In Cross-Flow (JICF) behaviour. Notably in the case of battle damage the jet never fully occupies the hole and jet velocity pro le is highly skewed towards the rear of the hole. Additionally, the measured velocity ratios are much less than would be expected for typical JICF. For example, strong jet behaviour is observed at a velocity ratio as low as 0.22 whereas JICF studies would suggest a much higher ratio (> 2) is required. Increasing velocity ratio has been related to a reduction in lift and an increase in drag. At the highest incidence tested (10 ) the velocity ratio of 0.32 resulted in a reduction of the lift coe fficient by 0.18 and an increase in the drag coeffi cient of 0.035.

The present work was aimed at the quantitative particle image velocimetry (PIV) measurement of a velocity field near the front of a propagating shock wave and the study of the dynamics of liquid tracers crossing the shock front. For this goal, a shock tube with a rectangular cross-section (48 × 24 mm) was used. The flat shock wave with Mach numbers M = 1.4–2.0 propagating inside the tube channel was studied as well as an expanding shock wave propagating outside the channel with M = 1.2–1.8 at its main axis. The PIV imaging of the shock fronts was carried out with an aerosol of dioctyl sebacate (DEHS) as tracer particles. The pressures of the gas in front of the shock waves studied ranged from 0.013 Mpa to 0.1 MPa in the series of experiments. The processed PIV data, compared to the 1D normal shock theory, yielded consistent values of wake velocity immediately behind the plain shock wave. Special attention was paid to the blurring of the velocity jump on the shock front due to the inertial particle lag and peculiarities of the PIV technique. A numerical algorithm was developed for analysis and correction of the PIV data on the shock fronts, based on equations of particle-flow interaction. By application of this algorithm, the effective particle diameter of the DEHS aerosol tracers was estimated as 1.03 ± 0.12 μm. A number of different formulations for particle drag were tested with this algorithm, with varying success. The results show consistency with previously reported experimental data obtained for cases of stationary shock waves.

In this study we have compared two modalities for flow quantification from measurement data; ultrasound (US) and shadow particle image velocimetry (PIV), and a flow simulation model using computational fluid dynamics (CFD). For the comparison we have used an idealized Quasi-2D phantom of the human left ventricular outflow tract (LVOT). The PIV data will serve as a reference for the true flow field in our setup. Furthermore, the US vector flow imaging (VFI) data has been post processed with model-based regularization developed to both smooth noise and sharpen physical flow features. The US VFI flow reconstruction results in an underestimation of the flow velocity magnitude compared to PIV and CFD. The CFD results coincide very well with the PIV flow field maximum velocities and curl intensity, as well as with the detailed vortex structure, however, this correspondence is subject to exact boundary conditions.

The capabilities of 3-D-Particle Tracking Velocimetry (PTV) to measure flow fields during the blending of Newtonian and non-Newtonian fluids in a standard baffled cylindrical vessel are assessed. The results are benchmarked against conventional 2-D Particle Image Velocimetry (PIV) data. The vessel, of diameter T=0.19m, is equipped with a 6-blade down-pumping PBT impeller of diameter, D=0.5T. Experiments in the low transitional (Re∼70), and transitional (Re∼1000) regimes have been conducted, using a range of Newtonian and non-Newtonian fluids. Turbulent flow measurements (Re>20,000) are made using Newtonian fluids. Data from both techniques are compared in terms of average flow field and, where appropriate, turbulent fluctuating velocity components. Particular emphasis is given on how comparisons can be made between the Eulerian PIV data and the Euler-Lagrangian PTV data. The overall results demonstrate the validity of the PTV technique in this application to acquire average flow fields which are in good agreement with PIV. Turbulent flow properties are less well resolved by PTV due in part to the large size of the tracer particle used. Other advantages and limitations of PTV versus PIV are also discussed.

High-resolution planar laser-induced fluorescence and particle image velocimetry (PIV) measurements were acquired during the intake stroke in a motored engine to investigate the mixing behavior of in-cylinder flows. The data were analyzed to determine the scalar energy and kinetic energy spectra, which were used to find the corresponding dissipation spectra. The results were compared with a model turbulent spectrum. The scalar energy and scalar dissipation spectra were shown to be resolved through the full dissipation range, enabling the determination of the Batchelor/Kolmogorov length scale and agreed well with the model turbulent spectrum at all but the highest wavenumbers where the effects of random noise were present. The 2% point in the scalar dissipation spectra was used to estimate the Batchelor scale, which was found to be approximately 32 μm. The PIV data, which had a 675 μm interrogation region, were used to calculate a one-dimensional kinetic energy spectrum. The kinetic energy spectrum agreed well with the scalar energy spectrum and the model spectrum up to wavenumbers corresponding to approximately two times the PIV interrogation region size. For the present measurements, this meant that the PIV data were not able to resolve the peak in the dissipation spectrum, i.e., the full high-wavenumber part of the inertial subrange.

Particle image velocimetry (PIV) data are a valuable asset in fluid mechanics. It is capable of visualizing flow structures even in complex physics scenarios, such as the flow at the exit of the rotor of a centrifugal fan. Machine learning is also a successful companion to PIV in order to increase data resolution or impute experimental gaps. While classical algorithms focus solely on replicating data using statistical metrics, the application of physics-informed neural networks (PINN) contributes to both data reconstruction and adherence to governing equations. The present study utilizes a convolutional physics-informed auto-encoder to reproduce planar PIV fields in the gappy regions while also satisfying the mass conservation equation. It proposes a novel approach that compromises experimental data reconstruction for compliance with physical restrictions. Simultaneously, it is aimed to ensure that the reconstruction error does not considerably deviate from the uncertainty band of the test data. A turbulence scale approximation is employed to set the relative weighting of the physical and data-driven terms in the loss function to ensure that both objectives are achieved. All steps are initially evaluated on a set of direct numerical simulation data to demonstrate the general capability of the network. Finally, examination of the PIV data indicates that the proposed PINN auto-encoder can enhance reconstruction accuracy by about 28% and 29% in terms of mass conservation residual and velocity statistics, respectively, at the expense of up to a 5% increase in the number of vectors with reconstruction error higher than the uncertainty band of the PIV test data.

Hemocompatibility is highly affected by the flow field in Left Ventricular Assistant Devices (LVAD). An asymmetric inflow and outflow channel arrangement with a 45° intersection angle with respect to the blood chamber is proposed to approximate the vascular structure of the aorta and left atrium on the left ventricle. The structure is expected to develop uninterruptible vortex flow state which is similar to the flow state in human left ventricle. The Computational Fluid Dynamics (CFD) asymmetric model is simulated using ANSYS workbench. To validate the velocity field calculated by CFD, a Particle Image Velocimetry (PIV) experiment is conducted. The CFD results show that the proposed blood chamber could generate a shifting vortex flow that would be redirected to the aorta during ejection to form a persistent recirculating flow state, which is similar to the echocardiographic flow state in left ventricle. Both the PIV and the CFD results show the development of a persistent vortex during the pulsatile period. Comparison of the qualitative flow pattern and quantitative probed velocity histories in a pulsatile period shows a good agreement between the CFD and PIV data. The goal of developing persistent quasi intra-ventricle vortex flow state in LVAD is realized.

Toensuregoodparticleimagevelocimetrycorrelations,itisnecessarytocharacterizetheimportantvariablesthat control image acquisition and image analysis. Among these, image spot size as it changes across the entire imaging surfacebearsinvestigation.Therefore, wedescribeanerroranalysisforsubpixel resolutionofcorrelation peaksus- ingWhittaker' smethod.Theseresultsseta toleranceonoptimumimagespotsizes.Typicallensesarethenanalyzed usingray-tracesoftware,toascertainwhatopticalparametersandcameramisalignmentsareallowablewithinthese optimum tolerances. We then describe an alignment procedure that allows us to set a camera to those tolerances. O acquire high-quality particle image velocimetry (PIV) data, it is necessary to use a cross-correlation technique, for exam- ple, two-color or image-shifted PIV, to align carefully the optical system and use an optimized data analysis package. In recent work performed at the Colorado School of Mines (CSM), two PIV inves- tigations resultedin1 )an assessmentofthe effects ofparticleimage spot size on correlation accuracy, 2 ) determination of the minimum spot size for least error, and 3 ) the development of an experimental system that can reliably generate this minimum spot size. One of these projects involved the development of PIV analy- sis software that generates vector data from PIV images (Ref. 1; T. Drouillard and M. Linne, Experiments in Fluids, manuscript in preparation ). In that work, several techniques for resolving cor- relation peaks were evaluated. These included the basic centroid method, Whittaker' s reconstruction, an analytical three-pointGaus- sian e t, and a numerical e ve-point Gaussian e t using the Marquardt method. An assessment of reconstruction errors was included. The least error was produced by the e ve-point Gaussian e t using Mar- quardt nonlinear e tting, whereas Whittaker' s reconstruction was a close second. Furthermore, it was found that the size of the cor- relation peak (directly related to particle image spot size ) has the greatest effect on the accuracy of the reconstruction algorithm. This sets a tolerance on the image spot size. That work is discussed in detail by Drouillard and Linne (manuscript in preparation ). Asecond projectinvolvedthe development of a method by which a PIV experimental apparatus may beconsistently and reproducibly aligned to meet this tolerance on spot size. 2 This topic is of primary importance to chemical e lm-based techniques, where turnaround time is very long. It also has implications for digital systems, espe- cially when it is necessary to align a camera across a large system, from within a wind tunnel, for example. A portion of this inves- tigation included modeling photographic lens performance under ideal and nonideal conditions using computer-based geometric ray tracing.These resultsprovide insightinto howvariousexperimental cone gurations can lead to spot size variation, which in turn affects the accuracy of PIV analysis. Based on such results, a system for alignment of the experiment was developed. That work is the pri- mary focus of this paper.

Particle image velocimetry (PIV) is a well-established and powerful technique for performing whole-field velocity measurements under a broad range of flow conditions, and is widely used in fluid mechanics laboratories.The aim of this feature issue of Measurement Science and Technology is to provide the reader with an overview of some recent progress in PIV applications developed in the EUROMECH 411 colloquium, which was held at CORIA, Rouen University, France from 29 to 31 May 2000.The topics of the meeting organized by M Trinité and B Lecordier were focused on three main aspects: Advanced PIV techniques (2D2C, 2D3C or 3D) used to study turbulent flows. Accuracy and limitations of PIV treatments for measurement of turbulence properties. Relevant post-processing tools for PIV data to extract information on turbulent flows.In total, 24 papers were presented on these topics and four keynote lectures were given by recognized specialists. Seven papers have been selected by the EUROMECH Scientific Committee for publication in this feature issue.If 3C, 3D (4D with time) is the ultimate goal of PIV developments, then holography is the only practical way to extend sheet-oriented PIV to measuring a volume of considerable depth in 3D. However, one major problem is the pollution from particles located at different depths: image quality is spoiled by noise from out-of-focus particles. Digital holography (Coëtmellec et al) using a 2D wavelet algorithm is a very simple promising new method in this area despite poor resolution.Stereoscopic PIV, using two cameras viewing the flow field from different angles, is one technique for measuring all three components in a fluid plane. To reconstruct the three-dimensional velocity field, a calibration stage is necessary since the imaging geometry causes significant aberrations. A 3D vector is calculated at a physical point in the object plane. The first step consists in computing two 2D vectors from both cameras at the physical point. Two methods are discussed (warping and mapping) in the paper by Coudert and Schon.Cross-correlation based PIV algorithms have a limited spatial resolution, which is determined by the size of the measurement window. The study and development of an integrated PIV-PTV method is described by Stitou and Riethmuller. This combines the robustness of Particle Image Velocimetry with the spatial resolution of Particle Tracking Velocimetry for super resolution analysis.Different experimental studies using known flow fields such as grid turbulence or pipe flows have been used to try to show the capability of the PIV technique to extract the information from turbulent vector maps. Nevertheless, from these studies, it always seems difficult to evaluate which treatment is the most accurate to measure a given flow. This is mainly due to the difficulty in knowing the exact characteristics of the flows. A solution can be to numerically impose the properties of the flow and to generate synthetic images (Lecordier et al). This approach is now possible by using Direct Numerical Simulation (DNS).If PIV has now evolved from its embryonic state to reach a certain level of confidence, we are now facing a problem directlylinked to the high spatial resolution and frequency ofacquisition that can nowadays be achieved by manufactured CCDcameras: the overabundance of information, which has to beappropriately reduced if one desires to draw conclusions fromthe data set. Two recent tools can now be used to resolve theproblem of unsteady turbulent flows and to try to separate thefluctuation intensity due to the unsteadiness and the turbulenceitself: Proper Orthogonal Decomposition (POD) and wavelettransform analysis. Promising results are obtained using POD inrotating flows (Graftieaux et al, Patte-Rouland et al). With regard to spatial correlation and patternrecognition this tool is also promising in evaluating thecontribution of coherent structures to the generation andself-sustaining of wall turbulence in boundary layers. Wavelettransform analysis can be applied to analyse temporal evolutionof coherent structures and to characterize the leading vortexgenerated in a starting flow (Schram and Riethmuller).We believe that this special feature provides a valuable overview of the proceedings at the colloquium and we would like to express our thanks to all the participants and to the contributing authors and reviewers who have made this special feature possible.

Agricultural crop residues (e.g., straw and stover) are a current focus for bioenergy feedstocks, with new technologies being developed to improve the economics of bioenergy production. Among the emerging technologies focused on feedstock engineering is the selective harvest concept. Due to the complexity of the biomass separations required for addressing the challenges and requirements of selective harvest, high fidelity models and advanced experimental methods that allow observation and measurement of the physical system are needed. These models and methods were developed and include computational fluid dynamics (CFD) modeling to simulate the cleaning shoe of a grain combine and a particle image velocimetry (PIV) technique to quantitatively and qualitatively characterize the cleaning shoe performance. While these techniques alone can be sufficient engineering and analysis tools for developing selective harvest technologies, this paper presents a new methodology, Virtual Engineering (VE), that integrates the CFD and PIV data into a virtual environment, where the data is coupled with the geometric model of a grain combine to provide a virtual representation of the cleaning shoe performance. Using VE visualization capabilities, the CFD and PIV data can be viewed in the context of the physical system for an interactive evaluation of characteristics and performance. This paper also discusses the concepts of additional VE tools that are being developed to provide necessary visualization, simulation and integration functionality.