Particle Image Velocimetry (PIV) is considered nowadays the state-of-the-art for non-intrusive and quantitative 3D velocity measurements. Its ability to measure the velocity field around complex geometries is a valuable tool that engineers can exploit for aerodynamic design optimization in various domains, such as aerospace, wind turbines and automotive, among others. Despite recent advancements, performing a PIV measurement in the industrial environment remains challenging due to several reasons: achieving large-scale measurements, complex geometries and high Reynolds numbers. The introduction of helium-filled soap bubbles, new Lagrangian Particle Tracking (LPT) algorithms and Robotic Volumetric PIV has allowed for the measurement of large-scale volumes around complex geometries. However, despite the described advancements, large-scale PIV and LPT measurements for industrial aerodynamics require further development to accelerate their applications. The first bottleneck considered is the maximum measurable velocity. For aerodynamic flows in the transport sector, the velocity is often larger than 50 m/s when considering aircraft and race cars. To apply the mentioned techniques, acquisition frequencies higher than the one commonly available are needed. The double-frame timing strategy, characterized by image pairs with a small time separation, is detrimental to the measurement accuracy, especially when low aperture systems, such as Robotic Volumetric PIV, are considered. This research has led to the development of novel acquisition strategies (chapters 3 and 4) that improve the accuracy of double-frame velocity measurements suited for high speed applications (U∞ > 50 m/s). Another current topic of research concerns the detection of data outliers in PIV measurements, which affect their reliability and trustfulness. In this thesis (chapter 5) a novel approach to outliers detection from time-averaged three dimensional PIV data is introduced. The principle invokes the physical mechanism of turbulence transport and is based on the agreement of the measured data to the turbulent kinetic energy (TKE) transport equation. The application of this new criterium to several experimental databases shows that spurious data can be detected more easily and unambiguously as an outlier along with a low fraction of false positives. This research also attempts to  decrease the gap between Computational Fluid Dynamics’ (CFD) and experiments’ aerodynamic data. In chapter 6, the application of PIV data for data assimilation is discussed. Data assimilation is a discipline in which observation and numerical or theoretical models are combined. This can be performed with two possible aims: improving the observation with physics-based models or increasing the capability of the model to represent reality. In this thesis, the latter is considered. A novel state observer technique is investigated for the assimilation of three-dimensional velocity measurements into computational fluid dynamics simulations based on Reynolds-averaged Navier–Stokes (RANS) equations. The state observer approach locally forces the solution to comply with the reference value, with increasing benefits when the density of forced points, or forcing density, is increased.@en

The occurrence of data outliers in PIV measurements remains nowadays a problematic issue; their effective detection is relevant to the reliability of PIV experiments. This study proposes a novel approach to outliers detection from time-averaged three-dimensional PIV data. The principle is based on the agreement of the measured data to the turbulent kinetic energy (TKE) transport equation. The ratio between the local advection and production terms of the TKE along the streamline determines the admissibility of the inquired datapoint. Planar and 3D PIV experimental datasets are used to demonstrate that in the presence of outliers, the turbulent transport (TT) criterion yields a large separation between correct and erroneous vectors. The comparison between the TT criterion and the state-of-the-art universal outlier detection from Westerweel and Scarano (Exp Fluids 39:1096–1100, 2005) shows that the proposed criterion yields a larger percentage of detected outliers along with a lower fraction of false positives for a wider range of possible values chosen for the threshold. Graphical abstract: [Figure not available: see fulltext.]

@en

State observer techniques are investigated for the assimilation of three-dimensional velocity measurements into computational fluid dynamics simulations based on Reynolds-averaged Navier–Stokes (RANS) equations. The method relies on a forcing term, or observer, in the momentum equation, stemming from the difference between the computed velocity and the reference value, obtained by measurements or high-fidelity simulations. Two different approaches for the forcing term are considered: proportional and integral-proportional. This technique is demonstrated considering an experimental database that describes the time-average three-dimensional flow behind a generic car-mirror model. The velocity field is obtained by means of Robotic Volumetric PIV measurements. The effects of the different forcing terms and the spatial density of the measurement input to the numerical simulation are studied. The state observer approach forces locally the solution to comply with the reference value and the extent of the region modified by the forcing input is discussed. The velocity distribution and flow topology obtained with data assimilation are compared with attention to the object wake and the reattachment point where the largest discrepancy is observed between the different approaches. The results show that the integral term is more effective than the proportional one in reducing the mismatch between simulation and the reference data, with increasing benefits when the density of forced points, or forcing density, is increased.@en

A novel approach is investigated to extend the range of measurable velocities by 3D-PTV systems. The method is specifically conceived for robotic volumetric PTV measurements, but it has applications for other similar techniques. The multi-Δt method relies upon combining the information from two or more sets of double-frame images with pulse separation of different time duration. Measurements with a short time separation yield a robust particle velocity field estimation with a higher percentage of valid vectors, yet a low measurement precision. Conversely, measurements with longer time separation potentially offer a higher measurement precision but suffer from an increased probability of spurious particle pairing. Reynolds decomposition is used to combine the two (dual-Δt) sets where a predictor for the mean particle displacement and its statistical dispersion is used to pair particle recordings from a longer time separation. For this reason, this method is aimed at the analysis of turbulent flows where the Reynolds decomposition is meaningful (e.g. turbulent flows with steady/quasi-steady boundary conditions). The extent of the search region is selected dynamically, based on the estimate of the velocity fluctuations from the short time separation evaluation. A more advanced variant of the algorithm contemplates the progressive increase of the pulse separation (multi-Δt) until the expected dispersion of data due to turbulent fluctuations eventually exceeds the distance between neighbouring particles. Flow measurements in the near wake of a truncated cylinder obstacle and of an Ahmed body are carried out to examine the performance of the proposed method. Reference data is taken from time-resolved multi-frame analysis based on the Shake-The-Box (STB) algorithm. The two experiments differ for the measurement principle used: The first one is conducted with a tomographic-like system (large aperture), whereas the latter uses coaxial volumetric velocimetry. The rate of correct pairing as well as the velocity dynamic range dependence upon the choice of the time separation are monitored and discussed. The results compare favourably with the STB analysis, indicating that the measurement of the time-average velocity field can be based on dual-Δt 3D-PTV measurements removing the constraint of time-resolved particle motion recording.

@en

A novel approach is introduced to enlarge the range of measurable velocities by PTV systems. The approach relies upon the acquisition of two or more sets of double-frame images with increasing pulse separation time Δt. The underlying principle is that measurements with a short Δt yield a velocity field with high percentage of valid vectors, but low measurement precision. Conversely, the measurements with longer Δt potentially offer a higher measurement precision but suffer from an increased probability of spurious particle pairing. Their combination is shown possible making use of Reynolds decomposition to form a predictor for the mean displacement and its statistical dispersion. The time-averaged velocity field produced with a short Δt is used as predictor to set the expected average displacement. Moreover, the extent of the search region is based on the estimate of the velocity fluctuations from the evaluation at short Δt. The algorithm can be applied progressively, increasing the pulse separation till truncation errors are found to limit the accuracy of the measurement. An experiment on the near wake of the Ahmed body performed with the Robotic Volumetric PIV system is used to assess the performance of the proposed method, which is compared with reference data from multi-frame measurements based on the Shake-the-Box (STB) algorithm. Results are firstly evaluated in terms of velocity pdf along the in-plane and coaxial directions. Furthermore, the vorticity field obtained by the different methods is compared. @en