Wind Tunnel Data Research Articles

Aerodynamic Performance Investigation of a Small Horizontal Axis Wind Turbine with Multi-Airfoil Blade Profiles of SD2030 and E231 Using Wind Tunnel Experiments and BEM Theory Method

The goal of this study is to investigate the performance of a small horizontal axis wind turbine blade at wind speeds of lower than 5 m/sec numerically and experimentally. The rotor blade has a diameter of 0.7m and consists of two airfoils profile: the thin airfoil SD2030 selected at the outboard region, and the thick airfoil E231 at the inboard region. The former airfoil, SD2030, is highly recognized for aerodynamic performance as it is suitable for low Reynolds number (Re), while the latter airfoil E231 is likewise appropriate for low Re flow but has a higher thickness-to-chord ratio which acts as structural support and good strength at the hub. The selected airfoils were analyzed using QBlade software based on XFOIL and Blade Element Theory (BEM) for different Reynolds number. Using a low-speed wind tunnel and Reynolds number ranging from 1x105 to 5x105, experiments at pitch angles of 5ο were conducted to determine the variation in torque and predict blade performance. The results reveal that the turbine with two hybrid different airfoil profiles have achieved a better start-up response at low wind speed of 1.5 m/s and a high-power coefficient of 0.34 with 3.7 tip speed ratio. The turbine can be used to generate a power of 0.5W at 1.5 m/s and a maximum power harvest of up to 20W at a wind speed of 7m/sec. The turbine's performance compared to other small wind turbines from the literatures demonstrate good performance and the results obtained from wind tunnel data were found in satisfactory agreement with the results provided by the BEM code.

Experimental assessment of Theodorsen's function for uncoupled pitch–plunge motion

The accuracy of Theodorsen's lift model for pure-pitch, pure-plunge and combined pitch–plunge oscillations of a two-dimensional model is compared with wind tunnel results. The reduced frequency of the oscillation was in the range of 0.01<k<0.15, and the freestream Reynolds number was in the range of 130 000<Re<400 000. The lift response to an uncoupled combined pitch–plunge motion (for which the frequency of pitch and plunge motions were not the same) is discussed using experimental results. The Theodorsen’s lift model is rewritten for the general uncoupled pitch–plunge motions by a linear superposition of all components of the airfoil bound circulation. Both amplitude and phase from the Theodorsen's function are compared with those of the wind tunnel data, and the results are discussed. The Theodorsen’s function is found to be a good estimator for both pure-pitch and pure-plunge motions. It further appropriately estimates the lift amplitude for the case of coupled pitch–plunge motion; however, the prediction is not accurate for the uncoupled pitch–plunge motion. A motion amplitude ratio is defined, which shows the level of aperiodicity of the motion. Discrepancy between experimental and analytical results increase with the reduction of the lift amplitude ratio and with the deviation of frequency ratio from unity.

Characterization of Non-Neutral Urban Canopy Wind Profile Using CFD Simulations—A Data-Driven Approach

Horizontally averaged wind profiles inside the urban canopy are used in many studies and numerical models. The existing analytical models are only applicable to a small range of aspect ratios and mostly neutral atmospheric conditions due to their underlying assumptions. In this study, a surrogate model for predicting horizontally averaged wind profiles in the street canyons of an idealized urban canopy for a wide range of urban morphologies and thermal forcing scenarios is developed with the help of machine learning techniques and computational fluid dynamics (CFD) simulation data. The influence of urban morphological parameters, atmospheric stability and wind conditions on the urban canopy wind flow is modeled using machine learning algorithms applied on CFD simulation results. The numerical model is validated using wind-tunnel data. Steady-state Reynolds averaged Navier–Stokes (RANS) simulations with a standard k-ϵ turbulence model for 252 different simulation conditions are performed on an idealized building geometry that consists of a regular array of 4 × 4 cubes. The simulation results are averaged horizontally to obtain the mean velocity and temperature profiles. Surrogate models are developed using the simulation outputs as training examples and the best model is chosen by comparing the performance of different machine learning models. The surrogate artificial neural network (ANN) model of this study outperforms the current state-of-the-art models in the prediction of horizontally averaged mean wind profile inside the urban canopy. The mean error (ME) and root-mean-square error (RMSE) of the discrete point prediction ANN model of this study are 0.016 m/s and 0.060 m/s, respectively, which is significantly lower compared to the best of the legacy models for which the errors are 0.048 m/s and 0.387 m/s, respectively.

Propeller Rotation Effects on Pitching Moment for Transport Aircraft with Conventional Tail

The effects of the propeller rotation direction on the pitching moment characteristics for a typical twin-engine turboprop aircraft with a conventional tail at low and moderate angles of attack under high-power conditions are investigated through unsteady Reynolds-averaged Navier–Stokes simulations. The numerical method is validated using wind-tunnel data for a twin-engine T-tail turboprop aircraft. The computational results of the conventional tail configurations indicate that the corotating propellers significantly degrade the longitudinal stability at low angles of attack in the takeoff configuration. By comparison with corotating propellers, counter-rotating inboard-up propellers not only increase the lift but also lead to a notable improvement in longitudinal stability, especially at low angles of attack. In contrast, the counter-rotating outboard-up propellers further reduce the negative pitching moment curve slope of the airframe. Analysis of the flowfield demonstrates that the effects of the propeller rotation direction significantly alter the downwash gradient at the horizontal tailplane by affecting the vertical location of the high-downwash region. The effectiveness of the horizontal tailplane and the longitudinal stability of the aircraft are therefore changed.

Numerical study on transient aerodynamic characteristics of high-speed trains during the opening of braking plates based on dynamic-overset-grid technology

The safety of high-speed trains is considerably compromised by increasing speed trends. Thus, braking plate technology has been applied to high-speed trains. The purpose of this study is to clarify the evolution law of the flow field of the vehicle during the opening of the braking plate and analyze the influence of the plate movement on the aerodynamic performance of the train. In this study, the flow field was computed using incompressible Navier–Stokes equations and the shear-stress transport (SST) k–ω turbulence model, and the unsteady flow over the opening brake plates was simulated using moving overlapping grids and dual time-stepping. The numerical method was verified through comparison with wind tunnel data (error <8%). The results reveal that the upstream braking plate significantly decreases the aerodynamic forces of the downstream plate during opening of the plates and causes the aerodynamic drag of the downstream braking plate to fluctuate significantly when it increases. The operation of the braking plate produces a small increase in the drag force of the train body (2.6%), but it significantly decreases the lift force of the train body (by up to 94%), especially during the opening of the braking plates. The flow field in the upper part of the train is significantly changed by the opening of the braking plate. In particular, the pressure in the cavity of the braking device changes sharply, and the surface is subjected to a large pulse pressure.

Numerical Simulation of a Y-Shaped Diverterless Supersonic Inlet

Three-dimensional (3-D) numerical simulations of a Y-shaped body-integrated diverterless supersonic inlet (DSI) have been investigated. A diverterless supersonic inlet is an external supersonic flow compression that uses a 3-D geometry of a bump for both supersonic flow compression and boundary-layer diversion. The simulations are performed in the presence of a typical body comprising an elliptical nose in order to model a more realistic approaching flow to the inlet entrance. All simulations were performed at a freestream Mach number of , the design Mach number, and at a 0 deg angle of attack (AOA) and 0 deg angle of sideslip (AOS). The Navier–Stokes relations along with shear stress transport equations are used to simulate turbulent flow on a multiblock structural grid. The present numerical simulation results show a good agreement with the wind-tunnel test data for the same DSI model. Moreover, the results show that the present numerical method has an acceptable accuracy in modeling a fully 3-D flow over the nose, along the supersonic compression region, and inside the inlet diffusers. The streamlines passed over the DSI supersonic compression clearly showed that the bump was successfully efficient in diverting the approaching boundary layer.

Multivariate Extreme Wind Loads: Copula-Based Analysis

A probabilistic approach is developed to estimate multivariate extreme wind loads by introducing copula theory. Three major concerns are addressed: (1) the characterization of pairwise dependence among extreme load coefficients, (2) the construction of a probability model of multivariate extreme load coefficients, and (3) the probabilistic estimation of multivariate extreme wind loads with randomness of the mean wind speed (i.e., wind climate change). Theoretical and numerical analyses are carried out with the aid of wind tunnel data. The results show that using rank dependence (Kendall’s tau and Spearman’s rho) is more appropriate than using Pearson correlation coefficient in defining dependence for extreme load coefficients. The Gaussian copula is convenient for deriving the joint distribution of multivariate extreme load coefficients but is not applicable for high-dimensional problems. In contrast, the vine copula is flexible and can provide a better estimate of the joint distribution function without dimension limitations. Multivariate annual maximum wind loads can be estimated via either first- or full-order methods. Dependence of the extreme load coefficients and randomness of the wind speed are both found having effects on the dependence of extreme wind loads. Moreover, the procedure of simulating multivariate annual maximum wind loads is presented to facilitate the use in practical problems.

Gust effect factors for regions of separated flow around rigid low-, mid-, and high-rise buildings

This paper examines the gust effect factor model for regions of separated flow on rigid buildings. Boundary layer wind tunnel data for a total of 58 buildings, with heights ranging from low-rise to high-rise were analyzed. The results show that the characteristics of separated flow determine the statistical distribution of surface pressures. It was found that the reattachment of separated flows onto building surfaces reduces the skewness and kurtosis of area-averaged surface pressures, causing them to follow the Gaussian distribution more closely. However, the pressures tend to have non-Gaussian distributions for surfaces without flow reattachment. Strong Karman-like vortex shedding increases aerodynamic admittance functions around the Strouhal number, along with the peak factors and gust effect factors of area-averaged side wall pressures. For surfaces with flow reattachment, Solari's (1993a) model predicts the gust effect factors with reasonable accuracy, particularly for uplift on low-rise buildings, mid-rise buildings with height-to-width ratios, H/W = 1 and high-rise buildings with height-to-width ratios, H/W > 4, and length-to-breadth ratios, L/B > 2. For separated flows without reattachment on the building, particularly for H/W > 4 and L/B < 2, periodic Karman-like vortex shedding tends to cause inaccuracies in the gust effect factor model, as might be expected.

Aerodynamic Force and Aeroelastic Response Characteristics Analyses for the Galloping of Ice-Covered Four-Split Transmission Lines in Oblique Flows

In order to study the galloping mechanism of ice-covered four-split transmission lines in oblique flows, the aerodynamic forces and aero-elastic response characteristics of the crescent-shaped four-split ice-covered transmission lines are investigated through wind tunnel tests on rigid and aero-elastic models. According to Den Hartog and Nigel’s galloping theories, the damping coefficients are calculated based on the experimental data. The results show that the crescent-shaped ice-covered four-split transmission lines usually suffer from torsional galloping. Furthermore, based on the aero-elastic wind tunnel data, the galloping is characterized by an elliptical trajectory, negative damping ratio, and a negative strain at hanging position. In addition, the galloping appears to be more prone to occur under oblique flows, with a larger galloping amplitude and a lower critical wind speed. This might be because an out-of-plane vibration of the third-order mode is excited at a lower wind speed, leading to a coupled resonance between in-plane and out-of-plane vibrations at the third-order mode with a frequency ratio of 1:1. The experimental results in this paper can also be used to verify the fluid-structure interaction simulation method of ice-covered transmission lines.

Aerodynamics of a Modified High-Lift Low Reynolds Number Airfoil: Preliminary Analysis

Airfoil selection is a crucial phase in the design of a small unmanned fixed-wing aircraft to allow a minimum size and weight of the lifting surfaces. The present study analyzesairfoils to be used in low operating Reynolds number unmanned aerial vehicles (UAV) using XFLR5 software, which was validated against wind tunnel data. Three baseline airfoils were investigated, namely NACA4412, Miley M06-13- 128 and Selig S1223. The best performing airfoils are then modified using the inverse airfoil design method to improve their performance in cruising flight. The modified airfoil resulted in larger leading edge radius and slightly thicker chord of the airfoil compared to the baseline airfoil. In general, the modified airfoil showed an improvement in stall characteristic, lift and drag coefficients in the post-stall regime, and a lower magnitude of moment coefficient for almost all investigated angles of attack compared to the baseline airfoil.

Investigating the rocket base flow characteristics and thermal environment of different nozzle configurations considering afterburning

During launch vehicle ascent, the base flow regimes of multi-nozzle rockets become increasingly complex owing to plume interactions. We developed a numerical method using the hybrid RANS/LES approach and the discrete ordinate method to accurately estimate the base thermal environment. Moreover, the finite-rate chemical kinetics are employed to calculate afterburning reactions. The validity of the numerical method is established by achieving a good agreement between the numerical results and wind tunnel data. Subsequently, the reacting and non-reacting flows of the three types of nozzle configurations were simulated at six different altitudes. The numerical results reveal that the plume interactions become stronger with the increase in flight height and total number of nozzles, and the maximum Cp of the four-nozzle rocket was 34.75% and 56.52% higher than that for three- and two-nozzle rocket at 40 km, respectively. After considering the afterburning effect, the maximum temperatures of the two-, three-, and four-nozzle rockets increased by 10.16%, 7.69%, and 5.43%, respectively, owing to the difference in plume collisions. With increasing altitude, the rocket base heating rate distribution changed significantly. The peak heat flux of the rocket base initially increases before decreasing with the maximum value of 147.32 kW/m2, 872.75 kW/m2, and 1395.80 kW/m2 for two-, three-, and four-nozzle rockets, respectively.

Fast-response hot-wire flow sensors for wind and gust estimation on UAVs

Due to limitations in available sensor technology, unmanned aerial vehicles (UAVs) lack an active sensing capability to measure turbulence, gusts, or other unsteady aerodynamic phenomena. Conventional in situ anemometry techniques fail to deliver in the harsh and dynamic multirotor environment due to form factor, resolution, or robustness requirements. To address this capability gap, a novel, fast-response sensor system to measure a wind vector in two dimensions is introduced and evaluated. This system, known as ‘MAST’ (for MEMS Anemometry Sensing Tower), leverages advances in microelectromechanical (MEMS) hot-wire devices to produce a solid-state, lightweight, and robust flow sensor suitable for real-time wind estimation onboard an UAV. The MAST uses five pentagonally-arranged microscale hot-wires to determine the wind vector’s direction and magnitude. The MAST’s performance was evaluated in a wind tunnel at speeds up to 5 m s−1 and orientations of 0∘–360∘. A neural network sensor model was trained from the wind tunnel data to estimate the wind vector from sensor signals. The average error of the sensor is 0.14 m s−1 for speed and 1.6∘ for direction. Furthermore, 95% of measurements are within 0.36 m s−1 for speed and 5.0∘ for direction. With a bandwidth of 570 Hz determined from square-wave testing, the MAST stands to greatly enhance UAV wind estimation capabilities and enable capturing relevant high-frequency phenomena in flow conditions.

Data Reduction Analysis on UTM-LST External Balance

Abstract: The effective use of wind-tunnel testing in determining aerodynamic properties of a body is very much dependent upon the reliability and speed with which wind-tunnel data can be reduced. The operating efficiency of the wind tunnels is substantially improved by the capability of providing lower aerodynamic coefficients in real time, or online, which decreases the expensive wind-tunnel time necessary for each test. This paper describes a system for presenting reduced wind-tunnel data in real time for UTM-LST. The requirements for data-handling equipment and data reduction procedures for wind tunnels are quite diverse, and depend upon the wind tunnel design and the type of tests for which they are used. The supersonic wind tunnels mentioned in this description have a variety of force-balance systems and are used for force tests, pressure tests, and other research projects. Consequently, the problems associated with in order to solve this diversity we build a computerized program where we can find the transformation of axis and aerodynamic characteristics at ease. This program can find the values of different aerodynamic coefficients with certain angle of attacks.

Wind Resource Assessment in Building Environment: Benchmarking of Numerical Approaches and Validation with Wind Tunnel Data

In the framework of the wind energy network for distributed generation in urban environments for most South American countries, wind resource assessment studies have been carried out in activities to establish a suitable methodology to assess the wind potential in urban environments. Some methodologies are already published in research studies, and the wind tunnel is the most accurate solution to obtain insight into the wind resource when measurements are unavailable, which is the most frequent case. Nevertheless, besides its validity, one cannot disregard the high cost of producing a scaled urban model and access to a wind tunnel. In this sense, this paper compares results from a wind tunnel experiment and different numerical modeling approaches. Two commercial models (WindSim and Wasp Engineering) and one open-source CFD code (OpenFOAM) are used. The results from the modeling simulation concluded that CFD models could achieve lower deviations for the mean wind speed and turbulence intensity when compared with non-CFD models. With such results, CFD modeling is a promising tool for reliably evaluating wind potential in urban environments.

Parametrization of irregularity of urban morphologies for designing better pedestrian wind environment in high-density cities – A wind tunnel study

Breezeways are crucial for horizontal urban ventilation. However, their morphologies, which are highly irregular, still lack investigation. This study aims to gain a categorical understanding of how irregular breezeway morphologies affect pedestrian-level wind performance in high-density cities. Firstly, we extracted breezeway areas and centerlines by adopting different definitions of breezeways: Method 1 defining all near-ground open-spaces (i.e., roads, other non-building areas, and low-rise-building areas) as breezeways; Method 2 defining roads as breezeways. Secondly, multiple morphological parameters were developed to describe irregularity (i.e., fragmentation, angularity, and sinuosity) and permeability of 19 typical urban sites in Hong Kong. Thirdly, linear regression analysis was conducted to correlate the parameters with wind tunnel data. The results show that in a dense built environment, urban ventilation relies on morphologies of all near-ground open-spaces rather than only roads. The regression analysis reveals a negative correlation between pedestrian-level wind performance and urban irregularity, where fragmentation has the largest impact due to the increase of building wall surfaces and viscous friction. The regression analysis reveals a positive correlation between pedestrian-level wind performance and urban permeability, where the newly-defined parameter (i.e., open-space width), which incorporates fragmentation into calculation, achieves higher coefficient of determination than the widely-used parameter (i.e., open-space coverage ratio). The newly-developed regression model, which considers both urban permeability and irregularity, predicts wind speed with a simple mathematical manipulation and a comparable accuracy to models in previous studies. More importantly, it provides applicable spatial information of urban irregularity for design of open-spaces and improvement of urban ventilation.

Machine-learning prediction of aerodynamic damping for buildings and structures undergoing flow-induced vibrations

Aerodynamic damping, as the principal determinant of wind-induced vibrations, bears vast significance in building engineering. This study presents a new modular framework, called the Deep Neural Network-Genetic Algorithm (DNN-GA) architecture, to predict aerodynamic damping directly from surface pressure measurement while serving as one of the earliest references for GA applications in aerodynamic damping predictions. The accurate prediction can give building engineers better insights into potential designs' feasibility at substantially reduced costs, benefiting building design and engineering implementations. The DNN-GA is demonstrated on an aeroelastic tapered prism—a nonlinear bi-directional fluid-structure interaction (FSI) system with solid connections to building design—with synchronous, high-fidelity wind tunnel data for training and prediction. With pressure input, the DNN module predicts tip response as the intermediate product, based on which the GA module optimizes for aerodynamic damping in a fully automated workflow. Results showed that the DNN-GA outperformed six benchmark machine learning algorithms by at least 400%. A comparison between the GA module and the traditional Random Decrement Technique (RDT) showed an accuracy improvement of at least 700%. Finally, the DNN-GA predicted the aerodynamic damping for the Vortex-Induced Vibration (VIV), Galloping, and VIV-Galloping regimes with the maximum root-mean-squared and mean-absolute errors of only 3.874 × 10−3 and 3.053 × 10−3, attesting to the method's excellent accuracy and suitability to complex, nonlinear, and many types FSI vibrations. Given its data-driven nature, the DNN-GA is also applicable to experimental, numerical, and even field data, making it an attractive tool for building engineering applications.

Generalized lift force model under vortex shedding

This paper proposes a generalized model for vortex shedding around a static cylinder. This model is a first step to bridge the gap between two existing families of Vortex Induced Vibration (VIV) models , namely (1) stochastic spectral models and (2) wake-oscillator models with coupled fluid and structural equations. The first family is experimentally-based while the models of the second family typically rely on theoretical and phenomenological concepts. These models use a Van der Pol, Rayleigh or a combination of both oscillators for the lift equation. The proposed model generalizes these models by combining all third degree terms in the case of a static model, i.e. by considering the equation of the fluid oscillator only. It is possible to enrich existing models with a more complex parametric model because its parameters are identified from experimental data. A second specificity of the proposed model stems from the observation that, for a static cylinder in low turbulence flow, the envelope of the measured lift force is not perfectly mono-harmonic, nor deterministic. Turbulence in the wake is expected to create fluctuations in the lift envelope. In the proposed model, stochasticity is consequently introduced to reproduce these fluctuations. It is based as an additive exogenous noise as an input to the generalized shedding model, using a Von Karman spectrum. Its coefficients are adjusted so that the model reproduces the probability density function of the measured lift envelope (and its power spectral density). The methodology is applied to Wind Tunnel data of a static circular cylinder in subcritical and postcritical regimes. This set of data is used to identify the coefficients of non-linear terms and Von Karman spectrum. The lift fluctuation coefficient generated with the model matches results available in the literature for the considered regimes. This work focuses on a methodology to obtain a generalized vortex shedding model from experimental data measured on a static body. It can be applied to other cylinder cross-sections or even extended to cylinders arranged in tandem.

Effect of roof shape on wind vulnerability of roof sheathing panels

Roofs of low-rise residential structures often suffer great damage during strong wind events. Sheathing loss that occurs due to excessive wind uplift may lead to internal damage due to water ingress and may in worst cases even initiate a progressive collapse of the structure. Hence, sheathings are an important aspect of any engineering-based wind vulnerability model for residential buildings. Vulnerability model is the part of a catastrophe model that relates the hazard intensity to the expected damage. The focus of this study is the application of a Monte Carlo simulation (MCS) based wind vulnerability model for roof sheathings considering the shape irregularities of the roofs of typical low-rise residential structures. The damage curves are derived based on wind tunnel tests on regular and irregular shaped structures. A comparison is made between the roof sheathing vulnerability curves of l-shaped, T-shaped, C-shaped, and rectangular shaped roofs. Through this comparison the effect of roof shape on performance of roof sheathings is derived. Newly acquired wind-tunnel data for irregular shaped structures are used for the analysis. Peak wind loads on roof sheathing are estimated based on a Hermite polynomial based probability distribution model for the wind pressure coefficients. The probability of failure of individual roof sheathings and the expected damage at incremental wind speeds is estimated. Significant variation in roof sheathing vulnerability is observed between the non-rectangular and rectangular roof shapes. The most critical directions for different roof shapes are identified, and the most vulnerable locations for roof sheathing damage within typical irregular shaped roofs are shown.

Simulations of Hypersonic Boundary-Layer Transition over a Flat Plate with the Spalart-Allmaras One-Equation BCM Transitional Model

Transitional flow has a significant impact on vehicles operating at supersonic and hypersonic speeds. An economic way to simulate this problem is to use computational fluid dynamics (CFD) codes. However, not all CFD codes can solve transitional flows. This paper examines the ability of the Spalart–Allmaras one-equation BCM (SA-BCM) transitional model to solve hypersonic transitional flow, implemented in the open-source CFD code Eilmer. Its performance is validated via existing wind tunnel data. Eight different hypersonic flow conditions are applied. A flat plate model is built for the numerical tests. The results indicate that the existing SA-BCM model is sensitive to the freestream turbulence intensity and the grid size. It is not accurate in all the test cases, though the transitional length can be matched by tuning the freestream intensity. This is likely due to the intermittency term of the SA-BCM model not being appropriately calibrated for high-velocity flow, though if the model can be recalibrated it may be able to solve the general high-velocity flows. Although the current SA-BCM model is only accurate under certain flow conditions after one calibration process, it remains attractive to CFD applications. As a one-equation model, the SA-BCM model runs much faster than multiple-equation flow models.

Kinematic training of convolutional neural networks for particle image velocimetry

Convolutional neural networks (CNNs) offer an alternative to the image cross-correlation methods used in particle image velocimetry (PIV) to reconstruct the fluid velocity field from the experimental recording. Despite the flexibility of CNNs, the accuracy and robustness of the standard image processing remains unsurpassed for general PIV data. As CNNs are non-linear and typically entail up to millions of trainable parameters, they require large and carefully designed training datasets to avoid over-fitting and to obtain results that are accurate for a wide range of flow conditions and length scales. Most training datasets consist of PIV-like data that are generated from displacement fields resulting from numerical flow simulations, which, in addition of being computationally expensive, may be able to inform the network only about relatively few classes of flow problems. To overcome this issue and improve the accuracy of the velocity reconstructed by CNNs, we propose to train the networks with synthetic PIV-like data generated from random displacement fields. The underlying idea is that the training dataset simply needs to teach the network about the kinematic relationship between position and velocity. These kinematic training datasets are computationally inexpensive and may allow a much richer variability in terms of length scales by varying the generation parameters. By training a state-of-the-art CNN, we investigate the accuracy of the reconstructed displacement and velocity with synthetic and experimental test cases, such as a sinusoidal flow and wind-tunnel data from a turbulent-boundary-layer and a cylinder-wake experiment. We demonstrate that kinematic training can drastically improve the accuracy of the CNN and allows the network to outperform conventional cross-correlation methods, being more robust with respect to data noise and providing reconstructed velocity fields that have considerably higher spatial resolution (at pixel level).