Session T01: Multiphase Flows: Atomization and Sprays II

The dynamics of the liquid-gas interface in two-fluid coaxial atomization in a high-pressure environment

We study the dynamics of the liquid-gas interface in the near field of a two-fluid atomizer. Specifically, we analyze how the ambient pressure influences the liquid destabilization and break-up process, as well as the early stages of spray formation and dispersion. The gas density increase associated with high environmental pressure, in the range of 1 - 5 atmospheres, changes the dynamics by which the high momentum coaxial gas jet destabilizes the liquid inner jet. We present new experimental results for a coaxial liquid-air atomizer operating in a high-pressure environment, with gas-to-liquid momentum ratio in the range of M = 5 - 50, and swirl ratio of SR = 0 - 1, in agreement with previously published results for atomization under atmospheric pressure conditions. High-speed shadowgraphy images are used to quantify the spatially and temporally varying liquid-gas interface field. Liquid core length, spreading angle, and other spray metrics are presented, and the influence of gas density/pressure identified from the comparison with atomization at atmospheric conditions. Results show a reduction of the spreading angle with the increase in gas density/pressure at injection.   

A numerical study of an atomizing jet in a resonant acoustic field

Resonant acoustic fields are influential to the performance of rocket engines, where instabilities caused by the coupling of heat release, atomization, and acoustics can lead to catastrophic failure. Furthermore, resonant acoustic fields show potential for manipulating and controlling fuel sprays dynamically. By applying recently advanced computational techniques, we investigate the relationship between the acoustics and atomization processes involved in such scenarios. After first demonstrating the accuracy of our flow solver in predicting the levitation of liquid droplets, we simulate a turbulent liquid jet in a periodic domain, exposed to resonant acoustic fields of differing magnitude. Described by the acoustic radiation Bond number, which involves the sound properties, surface tension, and initial jet diameter, these cases range in Bond number from 0 to 1.5. In this canonical setting, the resulting liquid instabilities, break-up behavior, and evolution of hydrodynamic variables are studied.   

Atomization of the optimally disturbed liquid jet

   Atomization is understood to be initiated by modal instability. In recent work by Hwang et al. (2021), strong amplification of the interface perturbation by a multi-phase Orr-mechanism is shown as a possible pathway to distort the liquid jet in the absence of exponential instability. Our study investigates the atomization of the liquid jet initiated by the optimal initial condition, which maximizes the perturbation energy transfer from the mean flow to the surface tension energy. Numerical simulations of the optimized jets are conducted with a weakly compressible solver, charLES.

   We demonstrate the multi-phase Orr-mechanism as a realizable process for atomization with various flow conditions. The comparison study between the randomly disturbed jet and the optimized jet reveals the formation of small-scale structures, and ligaments, occurring within a relatively short downstream distance in the latter case. It is shown that the emergence of small-scale structures does not necessarily coincide with the maximum interface distortion. Atomization characteristic is reported for several representative dimensionless parameters.   

Extraction of Atomization Process from High-Fidelity Simulations

Recent advances in computational efficiency and numerical methods allow researchers to simulate atomizing systems robustly and accurately. These simulations have proven crucial in furthering atomization research, which is vital in numerous industrial and environmental applications. Significant limitations still exist in this field, however. Numerically simulating multiple phases is computationally expensive and requires high-performance computing. Most researchers and engineers do not have access to supercomputers and require low-fidelity atomization models. The accuracy of these simplified models relies largely on the quality of data extracted from atomization simulations. However, resultant data sets from high-fidelity simulations are often tens to hundreds of terabytes, severely limiting researchers’ ability to parse them for relevant data. To address this problem, we developed a methodology that can be applied to numerical simulations of atomizing systems to identify and extract data from breakup and coalescence events vital to the atomization process. The data is orders of magnitude smaller but can provide statistics on the local conditions which lead to liquid breakup. This information can be used to better inform low-fidelity atomization models and elucidate the underlying physics of atomization.   

Transonic Core-Disrupting Atomization of Banana Puree

We characterized the disintegration of non-Newtonian banana puree, revealing atomization mechanisms, instabilities, and droplet size distribution using a novel twin-fluid atomizer via CFD. We approximated banana puree viscosity, which drives droplet size, using the Herschel-Bulkley model. Simulations reveal that interfacial unsteadiness leads to a pulsing flow, which amplifies the growth of instabilities to produce smaller droplets. The pulsing nature of the system produces radial bursts, resulting in periodic fluctuations in droplet size moving axially through the system in a wave pattern. High strain rates at the nozzle exit and heat transfer from the steam contribute to reduction in puree viscosity by three orders of magnitude from the puree annulus to the primary atomization region. Furthermore, to investigate the effects of temporally viscosity variability, we assess and characterize the transition during a 10-fold step increase in banana puree viscosity.   

Experimentally validated high-fidelity simulations of a liquid jet in supersonic crossflow

A liquid jet in supersonic crossflow is representative of a "cold-start" scenario in a scramjet engine, when the fuel is cool enough to remain liquid at the point of injection. Through the use of high-fidelity simulations, the liquid atomization process can be better understood, and other predictive models can be developed to inform design. We apply recently advanced computational methods to perform large-eddy simulation of a liquid jet in supersonic crossflow. In addition to its stability at any convective or acoustic CFL, the flow solver we use conserves volume in the low-Mach limit and conserves kinetic energy in the single-phase, low-Mach, inviscid limit. Simulating four different ratios of momentum flux, we validate the mean liquid distribution using recent experimental measurements of equivalent path length. Finally, we measure the statistics of liquid structures in the spray and investigate liquid surface instabilities.   

Modeling secondary breakup in atomization processes via machine learning

A highly sought goal of simulations of liquid jet atomization is to accurately predict the size distribution and number density of the atomized drops. This data is crucial for prediction of the surface-to-volume ratio, influencing evaporation rate, combustion efficiency and other important quantities. For realistic jets though, there is extreme separation of scales between the dynamics of the jet bulk and the breakup processes that produce the smallest drops, making it impractical to resolve all drop breakups. Thus, models must be employed for secondary breakup. However, existing breakup models are not universal and do not account for the local and instantaneous flow field and drop geometry. We introduce a machine-learning based model to predict the outcome of under-resolved drop breakups. Training the model on outcomes from the corresponding high-resolution simulations, we predict breakup outcomes from low-resolution, under-resolved input fields. Compared to results generated by low-resolution simulations, our ML-based approach achieves higher accuracy at predicting drop breakup and predicts a resulting droplet size distribution that is closer to the ground-truth distribution. Furthermore, our method learns the importance of physically relevant features such as the Weber number.   

Why we need to care about supercritical and non-ideal injection

Injection and primary break-up in engines are typically understood as multiphase phenomena, classified using Weber and Ohneseorge numbers, and simulated using methods that preserve the liquid volume and account for surface tension forces. However, in Diesel engines, in jet engines during take-off, and in rocket engines, injection occurs at conditions that exceed the respective fuel thermodynamic critical pressures. Injection processes under supercritical conditions behave fundamentally different from injection at ambient conditions - fluids become more compressible than ideal gases, break-up and combustion are tightly coupled, fluid properties vary wildly even for small changes in temperature or pressure, and surface tension may vanish altogether. This is becoming relevant for more and more engines. While jet engines have mostly transitioned to subcritical conditions between take-off and cruise since the 1980s, the first engines with fully supercritical conditions between take-off and cruise (10km, M 0.8) are already operational.  New automotive engine concepts such as spark controlled compression ignition (SCCI) and reaction controlled reaction ignition (RCCI) push gasoline engines into the realm of supercritical injection as well. Our models and analysis need to reflect this.   

Fragmentation of emulsion-based liquid sheets: influence of the oil viscosity

Agricultural spraying uses oil-in-water emulsions as additives to reduce drift and mitigate environmental contamination by increasing the mean size of spray droplets. Here, we investigate the effect of the oil phase viscosity on the dynamics of a liquid sheet of oil-in-water emulsion. We use T-junction microfluidic devices to produce emulsions of controlled composition and drop size. We then release a drop of emulsion on top of a small target. Upon impact, a transient free-liquid sheet expands radially. We use high-speed imaging to capture the expansion of the sheet and its fragmentation into droplets. Our study suggests that increasing the viscosity of the oil phase in oil-in-water emulsions results in suspension-like dynamics of the liquid sheet.   

Bags mediated film atomization in a cough machine

The fragmentation of a liquid mass into smaller droplets is an important physical process for epidemiology as it dictates the generation and transport of pathogen carrying aerosol. In the present study, we combine experiments and numerical computations to examine fluid mechanical processes associated with the generation of minute respiratory droplets (bioaerosol) during violent respiratory manoeuvres, such as coughing or sneezing. Analogous experiments performed in a cough machine allow us to illustrate the changes in liquid topology due to the shearing action of the air stream on a thin film. We identify that aerosol generation due to the shearing of liquid films is mediated by the formation of bag-like structures. It is found that the occurrence of weak spots triggers the eventual demise of these inflated structures by hole expansion. Crucially, increasing viscosity makes the bags puncture when thinner, thus generating smaller droplets. Finally, we show that the breakup of these bags gives rise to a droplet size distribution that is well captured by a log-normal distribution.   

Variational Data Assimilation for Stationary Euler-Euler Spray Simulation

Simulations of flow and transport in sprays are computationally expensive and rely on models to incorporate the interaction between the different phases. Here, we employ the Euler-Euler multi-fluid approach with a finite number of liquid phases to simulate the evolution of sprays downstream of primary and secondary break-up.

To compensate for model uncertainties and to reduce discrepancies between simulation results and experimental reference data we perform variational data assimilation in a two-step process. In particular, after identifying optimal boundary conditions, the eddy viscosity of the continuous phase and the drag coefficient fields are adapted by the discrete adjoint method. This method provides a cost function gradient at computational cost independent of the number of parameters. A gradient based optimizer is then applied to update the parameter values.

An OpenFOAM Euler-Euler multi-fluid solver is extended to solve the corresponding adjoint equations for the gradients of the parameter fields and the inlet boundary conditions. This solver is based on an analogous solver for the stationary RANS equations, presented earlier.

We present results for different spray configurations. An emphasis is put on the analysis of the reference data distribution and regularization.   

Breakup of swirling liquid sheet in the presence of acoustically perturbed gas jet

We aim to investigate the effects of acoustically perturbed gas jet on the near orifice breakup of swirling, annular liquid sheet. The gas jet has a tendency to respond to the combustor acoustic field generated from the unsteady heat release, which in turn creates unsteady atomization responsible for heat release rate, thereby forming two-way coupling. Present experiments focus on these unsteady effects with the help of upstream acoustic excitation up to 2000 Hz. The interaction between the swirling liquid sheet and the gas jet perturbations was studied using time-resolved shadowgraphy followed by the POD analysis. The steady spray starts pulsating periodically in the presence of forcing, and this triggered pulsation picks up the forcing frequency as the dominant pulsation frequency at lower frequencies (<1000 Hz). The self-pulsation regime, common in the coaxial atomizers, however, has a different response to the acoustics. Apart from revealing various large scale modes, the unsteady dynamics show the existence of two dominant frequencies viz. self-pulsation frequency and the forcing frequency. When the forcing frequency is close to the pulsation frequency, the lock-in occurs resulting from the extraction of forcing energy by the self-pulsation phenomenon.