4D characterization of the swirling gas jet flow field in a coaxial two-fluid atomizer at high pressure. Computational imaging techniques reveal the complex three-dimensional structure of the gas jet as it interacts with the liquid sheet, providing insight into the atomization mechanism. In collaboration with Prof. Alberto Aliseda's group at the University of Washington.

Volumetric flow velocity measurement using tomographic imaging techniques applied to gas-liquid coaxial atomization with swirl in high-pressure environments. Multi-camera setups enable reconstruction of the full 3D velocity field, revealing the complex vortex structures that drive atomization.

Fragmentation of an immiscible jet can be found in many industrial, environmental, and biological applications. However, quantitative data are hard to obtain experimentally in optically opaque flow, especially in the nearfield where dense oil ligaments and droplets exist. Liquid-liquid refractive index matching and simultaneous Particle Image Velocimetry (PIV) and Laser Induced-Fluorescence (PLIF) are developed to measure interface distributions and turbulent statistics across 2,500 instantaneous fields at three downstream locations. A comprehensive turbulence analysis toolkit was developed in MATLAB to compute phase-conditioned statistics: vorticity and swirl strength, TKE production and dissipation rates in cylindrical coordinates, Reynolds stress budgets, two-point velocity correlations, Kolmogorov scale estimation, and spreading rate analysis. The results show the dispersed phase significantly increases the turbulence level compared to a single-phase jet, with production mechanisms in the oil and water phases being fundamentally different.

Compound droplets containing multiple water droplets, some with smaller oil droplets, regularly form at jet Reynolds numbers Re>1358. The origin of some of the encapsulated water droplets can be traced back to the entrained water ligaments during the initial roll-up of Kelvin-Helmholtz vortices. Random forest-based machine learning (Weka) was trained and applied via automated ImageJ macros to distinguish compound from regular droplets in PLIF images. Blob analysis quantified the morphology of over 3,000 droplets: equivalent diameter, deformation index (axis ratio), interior pocket count, and oil fraction. A pendant drop interfacial tension measurement code was developed using Young-Laplace equation fitting with Newton-Raphson optimization to characterize the oil-water-dispersant systems. On average, the interior pockets raise the oil-water interfacial area by 15%, and while the oil droplets are deformed by the jet’s shear field, the interior interfaces remain nearly spherical, indicating a quiescent micro-environment in the local high shear zone.

A custom-built, submersible inline holography system designed for in situ measurement of droplet size distributions within opaque oil plumes. The instrument houses a 635 nm diode laser with spatial filter and collimating optics in one sealed aluminum tube, and a Schneider imaging lens with a 5-megapixel CMOS camera (3.3 μm/pixel resolution) in another, mounted on an optical rail for adjustable sample volume depth. Deployed in a 2.5 m towing tank, the Mini-Holocam was suspended within crude oil jet-in-crossflow plumes to measure sub-millimeter droplet distributions at three dispersant-to-oil ratios. Increasing dispersant concentration dramatically decreased droplet sizes (volume median diameter from 734 μm to 234 μm) and reduced plume rise rates. A force-balance trapping model correctly predicted which droplet sizes are captured by the counter-rotating vortex pair (CVP) characteristic of jets in crossflow. This work was funded by the Gulf of Mexico Research Initiative (GoMRI) and published in the Journal of Geophysical Research: Oceans (2016).

An end-to-end computational pipeline for extracting 3D droplet positions and sizes from inline holograms. The pipeline spans eight stages: preprocessing, GPU-CUDA holographic reconstruction into volumetric image stacks, high-pass filtering with minimum intensity projection, 2D segmentation (separate paths for large and small droplets using local adaptive thresholding and morphological operations), 3D circular Hough transform for droplet center and radius detection in the reconstructed volume, minimum intensity-based sizing for sub-resolution particles, and finally droplet count concentration (DCC) and mass concentration (DMC) spectrum computation. For high-magnification near-nozzle spray images where traditional methods fail, the pipeline is augmented with wavelet-FFT stripe removal, POD-based background subtraction, and Weka machine learning classifiers trained to distinguish droplets from background. MATLAB scripts auto-generate ImageJ macros for batch processing, enabling quantification of droplet sizes in previously inaccessible dense spray regions.

This study investigates the effects of premixing oil with chemical dispersant at varying concentrations on the flow structure and droplet dynamics within a crude oil jet transitioning into a plume in a crossflow. It is motivated by the need to determine the fate of subsurface oil after a blowout disaster. The flow within the plume consists primarily of a pair of counterrotating quasi-streamwise vortices (CVP) that characterize jets in crossflows. The size of droplets trapped by the CVP is predicted correctly using a trapping function based on a balance of forces on a droplet located within a horizontal eddy. This study is in collaboration with Prof. David Murphy and Dr. Kaushik Sampath.

Cavitation in the flapper–nozzle pilot stage is an important source for the noise, performance deterioration, and even failure of electrohydraulic servo-valves. In this study, experimental and numerical investigations of the cavitation phenomenon appearing in the flow field between the flapper and nozzle of an electrohydraulic servo-valve are carried out. As a result, cavitation source locations are confirmed as the nozzle tip and flapper leading edge. This study is supervised by Prof. Songjing Li, and in collaboration with Dr. Nay ZarAung, Dr. Shengzhuo Zhang, and Dr. Junzhang Cao.