Particle image velocimetry (PIV) is a powerful diagnostic tool for analyzing the properties and structures of laminar and turbulent flows. Modern PIV systems provide high spatial and temporal resolution measurements, often generating hundreds of gigabytes of raw data. Before these data can be used for flow analysis, they must be post-processed to identify and remove erroneous velocity vectors. Post-processing techniques typically require substantial manual intervention by highly trained analysts, making the process tedious, time-consuming, and inherently subjective. To address these challenges, we developed a physics-aware, data-driven framework for automated PIV post-processing. Our machine learning-based tool, PIVPostProcess, employs an ensemble of gradient-boosted decision trees trained on real experimental PIV datasets and is validated using direct numerical simulation data with synthetic noise. The fully trained model achieves excellent performance, with an F1 score—a performance metric that combines precision and recall—of 0.991 on the testing dataset. To evaluate generalization beyond the training distribution, we apply the model to six additional experimental datasets featuring previously unseen flow fields. The model achieves a mean F1 score of 0.990 across tested experimental datasets, demonstrating robust predictive capability. Importantly, this automated technique processes each dataset significantly faster compared to manual post-processing. By substantially reducing processing time while maintaining high accuracy, this tool accelerates and standardizes PIV vector validation approaches.
Liquid foams have demonstrated a strong capability to reduce the impact of pressure disturbances, including acoustic, sustained shock, and blast waves. However, the wide range of experimental methodologies reported in the literature often prevents direct comparison of results, largely due to incomplete characterisation of foam properties, notably bubble size. In this study, experiments were conducted using a dedicated set-up enabling a precise control of critical parameters, including bubble size (160–450 m), liquid fraction (2–8
This paper presents an extensive study of the air-blast atomization cascade from liquid core to ligaments and droplets. Microscopic backlit images of the sprays were captured along the jet centerline at various axial locations using two orthogonal high-speed cameras. A comprehensive imaging and image processing methodology was performed to enable direct measurements of the wavelength and amplitude of ligament instabilities forming on individual ligaments ejected from the primary liquid jet. This has enabled a complete reconstruction of the link between the probability distribution of the wavelength of ligament instabilities and the subsequent formation of droplets. Statistical analysis revealed the dependence of ligament instabilities on inlet conditions while enabling quantification of the full cascade of processes from the formation of liquid jet as a function of inlet conditions to the subsequent formation of ligaments and then, finally, to droplets.
Early-stage condensation droplets set initial conditions for subsequent growth, coalescence, and surface re-exposure, yet their time-resolved three-dimensional morphology at the microscale remains difficult to measure using conventional two-dimensional imaging techniques. Here, digital holographic microscopy (DHM) was used to reconstruct time-resolved height fields of an individual condensing microdroplet on a silane-coated glass surface, with lateral and temporal resolutions of 0.83 m and 66 ms, respectively. The reconstructed phase fields were converted into droplet height maps, from which maximum height, footprint-equivalent diameter, integrated volume, and projected-boundary descriptors were quantified. For the representative droplet analyzed in this study, early growth departed from uniform spherical-cap-like enlargement. Vertical growth was more pronounced than lateral footprint expansion during an early growth interval, while projected-boundary analysis showed that the initially less regular footprint became more circular as growth proceeded. At three selected growth stages, threshold-aligned spherical-cap cross-estimates underestimated the DHM-measured maximum height by 10.47–17.29
Reconstruction of large-scale atmospheric boundary layer (ABL) dynamics from sparse measurements is essential for interpreting wind-tunnel experiments and for data-driven modelling of complex flows. Using a time-resolved experimental campaign providing three-component velocity fields on a transverse-vertical (YZ) plane from stereoscopic particle image velocimetry (SPIV), together with sparse hot-wire anemometry (HWA) probe signals, we assess an autoencoder (AE)-defined large-scale representation and benchmark it against a spectral proper orthogonal decomposition (SPOD)-based representation. Building on linear stochastic estimation (LSE) and convolutional neural networks (CNNs), including a temporal CNN (TCN) variant, four approaches (SPOD-LSE, SPOD-CNN, AE-CNN, and AE-TCN) are evaluated to identify the most accurate prediction of large scales from YZ-plane centreline measurements and probes. AE-TCN achieves the best overall composite performance and provides the strongest centroid tracking. Physical-consistency checks are then used to assess whether the reconstructed fields retain coherent-structure organisation and nonlinear interactions with the small scales.
