Atmospheric Turbulence poses a challenge to land-based observatories operated by the United States Air Force (USAF) tasked with space situational awareness. By developing new methods for quantifying Turbulence, we intend to provide increased USAF capability in this domain. Current models for quantifying atmospheric Turbulence include Kolmogorov and Non-Kolmogorov methods. Through the nature of Fourier Transform, sinusoidal function, it is possible to determine the frequency at which velocities occur in a specified vertical distance and eventually determine eddy size in a control volume. First, an ANSYS Computational Fluid Dynamics (CFD) model will be created to simulate atmospheric Turbulence in a defined control volume. The simulation will include a one-dimensional flow over a flat plate. The data we acquired from the simulation were used to derive an equation relating the velocity to the vertical distance (velocity profile). We will perform a regression analysis to fit data from Large-Eddy Simulations (LES) and apply Fourier transformation from a time domain to a frequency domain. The objective is to use Fourier transform analysis to determine eddy size distribution and turbulent cascade dissipation in a control volume by analyzing the frequency of velocities. By calculating such eddy size distribution, we may quantify Turbulence in said control volume and compare results with the traditional Kolmogorov method.

Understanding the transition from laminar to turbulent flow – Boundary-Layer Transition (BLT), we can design better state-of-the-art vehicles for defense and space applications, which can mitigate the limitations in current high-speed temperature conditions. BLT is a subject of fluid flow disturbances created by geometric parameters and flow conditions, such as surface roughness, increased velocity, and high-pressure fluctuations, to name a few. These disturbances lead to the development of turbulent spots and differential heating. Historically, the Reynolds number has been used to predict whether a system will develop turbulent flow. However, it has been known for decades that it is not always reliable and cannot indicate where the BLT will occur: some experiments present scenarios where the flow is laminar at a high Reynolds number and vice versa. We can predict the BLT from performing physical experiments, but they are expensive and physical configurations are limited. Despite many community efforts and successes, no general computational solution to simulate different flows and vehicle types that fully incorporate BLT exists. Many are a considerable number of parameters that affect BLT. Therefore, we use Causal Inference to predict BLT by cause-and-effect analysis on multivariate data obtained from BLT studies. Data generated using high-fidelity Computational Fluid Dynamics (CFD) with resolved Large-Eddy Simulations (LES) scales, will be analyzed for turbulence intensity by decomposing velocity in mean and fluctuations. In this paper, we will be discussing approaches on how we predict BLT scenarios using cause and effect relationships driven by causal inference analysis.

It has been suggested that stasis (stagnant zones over a period of time, dependent on other factors such as age, or underlying medical conditions, such as cancer or covid19) in the valve pockets may increase the risk of clots due to stasis in combination with other factors increases the risk of Deep Venous Thrombosis (DVT) formation, blood stasis may also result in a decrease in the anticoagulants factors that prevent clots from forming, and if the vein wall is damaged this further increases the risk of clot formation. We propose a proactive framework to predict DVT vulnerability, track progression and provide patient care checkpoints is of clear benefit. The framework is based on leading-edge cloud computing technologies and promises to offer user-friendly Software- & Platform-as-a-Service (SaaS/PaaS) solutions via novel machine learning (ML) algorithm and high fidelity blood flow modeling through the venous network under various valve configurations. In this work, we will present the progress made towards the leaflet morphology extraction from in-vitro images using ML assisted stereological analysis for obtaining a sufficiently accurate representation of morphology. Ultimately, the workflow can be tailored to specific patients. The extracted valve is used to identify red-flag stagnant zones by a detailed, physics-based computational study of the blood flow through the leaflet models.

Specializing in high-speed testing, Holloman High-Speed Test Track (HHSTT) uses a process called ‘water braking’ as a method to bring vehicles at the test track to a stop. This method takes advantage of the higher density of water, compared to air, to increase braking capability through momentum exchange. By studying water braking using Computational Fluid Dynamics (CFD), forces acting on track vehicles can be approximated and prepared for prior to actual test. In this study, focus will be made on the brake component of the track sled that is responsible for interacting with the water for braking. By discretizing a volume space around our brake, we accelerate water and air to relatively simulate the brake engaging. The model is a multi-phase flow that uses the governing equations of gas and liquid phases with the finite volume method, to perform 3D simulations. By adjusting the inflow velocity of air and water, it is possible to simulate HHSTT sled tests at various operational speeds. In the development of the 3D predictive model, convergence issues associated with the numerical mesh, initial/boundary conditions, and compressibility of the fluids were encountered. Once resolved, the effect of inflow velocities of water and air on the braking of the sled are studied.

Atmospheric turbulence studies indicate the presence of self-similar scaling structures over a range of scales from the inertial outer scale to the dissipative inner scale. A measure of this self-similar structure has been obtained by computing the fractal dimension of images visualizing the turbulence using the widely used box-counting method. If applied blindly, the box-counting method can lead to misleading results in which the edges of the scaling range, corresponding to the upper and lower length scales referred to above are incorporated in an incorrect way. Furthermore, certain structures arising in turbulent flows that are not self-similar can deliver spurious contributions to the box-counting dimension. An appropriately trained Convolutional Neural Network can take account of both the above features in an appropriate way, using as inputs more detailed information than just the number of boxes covering the putative fractal set. To give a particular example, how the shape of clusters of covering boxes covering the object changes with box size could be analyzed. We will create a data set of decaying isotropic turbulence scenarios for atmospheric turbulence using Large-Eddy Simulations (LES) and analyze characteristic structures arising from these. These could include contours of velocity magnitude, as well as of levels of a passive scalar introduced into the simulated flows. We will then identify features of the structures that can be used to train the networks to obtain the most appropriate fractal dimension describing the scaling range, even when this range is of limited extent, down to a minimum of one order of magnitude.

The CO 2 Dragster competition is performed at a national level as part of the Technology Student Association (TSA), and it seeks to increase and implement STEM skills for high school students. Students are expected to create CO 2 -powered Dragsters, which are placed on a 20-meter track to compete against each other. The CO 2 Dragster consists of a pressure vessel containing CO 2 , which is placed in a pre-drilled cartridge located at the back of the vehicle. The CO 2 propellant is used to provide thrust to the Dragster during the race, and sand smoothed balsawood material is used for the solid body of the Dragster. The CO 2 Dragster competition is introduced at the high school level to provide students with a hands-on activity that allows them to learn and apply the fundamentals of fluid dynamics and solid mechanics. The students are placed in charge of designing a 3D model of their Dragster and are required to submit the 3D model with their respective drawings. In this paper, a mathematical, computational, and experimental analysis of a CO 2 Dragster is provided. This research consisted of creating a shell model on NX 10, then inserted into ANSYS Fluent to simulate the flow around the model to identify and track the stagnation points, pressure loadings, and flow separation effects caused by the fluid interaction with the Dragster. Mathematical formulas were implemented to find the fracture of slim body CO 2 Dragsters, using a coupling between fluid mechanics and solid mechanics, Fluid-Structure Interactions (FSIs). In conclusion, by creating a computational model, it is possible to drive dragster design by simulating different dragster designs’ computationally to improve its performance. Creating a computational functional model to simulate the Dragster, implementing mathematical formulas to understand the behavior, and experimentally evaluating the parameters based on performance will lead to more innovative practices of creating better Dragsters for this competition.

The Large Eddy Simulations (LES) modeling of turbulence effects is computationally expensive even when not all scales are resolved, especially in the presence of deep turbulence effects in the atmosphere. Machine learning techniques provide a novel way to propagate the effects from inner- to outer-scale in atmospheric turbulence spectrum and to accelerate its characterization on long-distance laser propagation. We simulated the turbulent flow of atmospheric air in an idealized box with a temperature difference between the lower and upper surfaces of about 27 degrees Celsius with the LES method. The volume was voxelized, and several quantities, such as the velocity, temperature, and the pressure were obtained at regularly spaced grid points. These values were binned and converted into symbols that were concatenated along the length of the box to create a ‘text’ that was used to train a long short-term memory (LSTM) neural network and propose a way to use a naive Bayes model. LSTMs are used in speech recognition, and handwriting recognition tasks and naïve Bayes is used extensively in text categorization. The trained LSTM and the naïve Bayes models were used to generate instances of turbulent-like flows. Errors are quantified, and portrait as a difference that enables our studies to track error quantities passed through stochastic generative machine learning models — considering that our LES studies have a high state of the art high-fidelity approximation solutions of the Navier-Stokes. In the present work, LES solutions are imitated and compare against generative machine learning models.

We have simulated atmospheric turbulence at several instances of possible laser propagation paths and turbulent flow regime regions within the area of interest. At a control volume that is in between two Hawaii mountains. We have applied statistical correlations between Large-Eddy Simulations results using CFD modeling and parametrizing optical variables of interest, such as refractive index structure function. By comparing our Large-Eddy Simulations with specified parameters against other Large-Eddy Simulations with almost all same parameters except for one allows us to perform a sensitivity study. To study the changes on how a parameter can affect other scenarios of Large-Eddy Simulations parametric studies. With the end goal of validating the capacity of a sensitivity analysis study using Large-Eddy Simulations versus other Large-Eddy Simulations by a way of simulating and parametrizing turbulent flow studies found in the field of CFD modeling. Allowing us to achieve our stochastic analysis by applying sensitivity studies to see how our distributions change as a function of different parameters, but one at a time. After a comparison between CFD modelling simulations we have found that after a complete parametric study, a correlation was formed between turbulent flow parameters and optical parameters of interest.

Gas-solid fluidized beds are widely used in petroleum, chemical, mineral, pharmaceutical, and power plant applications. The performance of fluidized bed reactors strongly depends on the flow dynamics. Characterization of a particle-laden flow has been one of the challenging issues in fluidization research. The simulation of flow in such processes is challenging as the complex dynamic systems comprised of numerous particles and fluidizing gas confined in specific devices. Nonlinear particle-particle/wall and particle-gas interactions lead to complex flow behavior of the gas-solid flows. We used MFiX to simulate a gas-solid flow in fluidized beds. A data-driven framework is trained with the data from MFiX-PIC simulations. The trained and tested machine learning model is used to characterize the flow regimes in fluidized beds. In the present study, the void fraction is used to characterize the flow regimes. However, others in the literature have used pressure, temperature, heat transfer coefficient, acoustic emission, vibration, and electrostatic charge for the characterization of flow regimes.

MFiX is a multiphase open-source suite that is developed at the National Energy Technology Laboratories. It is widely used by fossil fuel reactor communities to simulate flow in a fluidized bed reactor. It does not have advanced linear iterative solvers even though it spends 70% of the run time in solving the linear system. Trilinos contains algorithms and enabling technologies for the solution of large-scale, sophisticated multi-physics engineering and scientific problems. The library developed at Sandia National Laboratories has more than 60 packages. It consists of state-of-the-art preconditioners, nonlinear solvers, direct solvers, and iterative solvers. The packages are performant and portable on various hybrid computing architectures. To improve the capabilities of MFiX, we developed a framework, MFiX-Trilinos, to integrate the advanced linear solvers in Trilinos with the FORTRAN based multiphase flow solver, MFiX. The framework changes the semantics of the array in FORTRAN and C++ and solve the linear system with packages in Trilinos and returns the solution to MFiX. The preconditioned iterative solvers considered for the analysis are BiCGStab and GMRES. The framework is verified on various fluidized bed problems. The performance of the framework is tested on the Stampede supercomputer. The wall time for multiple sizes of fluidized beds is compared.

In the present study, particles in cell method, a Eulerian-Lagrangian approach is used to simulate the flow in an inhaler. The number of uncertain parameters, including properties of particles, fluidizing agents’ properties, initial/boundary conditions, and numerical parameters related to PIC simulations, is fourteen. The residence time of 280 PIC simulations for different values of the uncertain parameters is used to test/train a data-driven framework. The values of the uncertain parameters are generated via the Latin Hypercube Sampling method and a normal distribution. The trained algorithm is used to predict the residence time for various unknown parameters. This framework is used to carry out the sensitivity analysis to find the most influential settings on the residence time of the particles in the inhaler. The optimum parameters of the influential parameters for a given residence time is calculated via the data-driven framework.

Fluidized beds are used in a wide range of applications in gasification, combustion, and process engineering. Multiphase flow in such applications involves numerous uncertain parameters. Uncertainty quantification provides uncertainty in syngas yield and efficiency of coal/biomass gasification in a power plant. Techniques such as sensitivity analysis are useful in identifying parameters that have the most influence on the quantities of interest. Also, it helps to decrease the computational cost of the uncertainty quantification and optimize the reactor. We carried out a nondeterministic analysis of flow in a biomass reactor. The flow in the reactor is simulated with National Energy Technology Laboratory’s open source multiphase fluid dynamics suite MFiX. It does not possess tools for uncertainty quantification. Therefore, we developed a C++ wrapper to integrate an uncertainty quantification toolkit developed at Sandia National Laboratory with MFiX. The wrapper exchanges uncertain input parameters and critical output parameters among Dakota and MFiX. We quantify uncertainty in key output parameters via a sampling method. In addition, sensitivity analysis is carried out for all eight uncertain input parameters namely particle-particle restitution coefficient, angle of internal friction, coefficient of friction between two-phases, velocity of the fluidizing agent at the inlet, velocity of the biomass particles at the inlet, diameter of the biomass particles, viscosity of the fluidizing agent, and the percentage of nitrogen/oxygen in the fluidizing agent.

In the manufacturing of metal matrix composites (MMC), liquid-metal reactive infusion with a solid mesh or particles composed of ceramic or metal may be used. The objective of this study is to determine the uncertainty quantification of the modeling of liquid hafnium infusion to expedite the processing and improve properties of MMCs ultimately. Uncertainty quantification (UQ) characterized the uncertainty scientifically especially for high-performance computing with observed physics and/or chemistry of the phenomena and predicted from estimated parameters. In this work, molten hafnium infusing through a boron carbide packed bed is modeled to optimize the manufacturing of components used for a hypersonic vehicle. The creation of molten matrix composites by the infiltration of molten metal represents a formidable challenge to be accurately modeled. First, the structural randomness associated with porous mediums complicates the prediction of the flow passing through it. Secondly, the properties of the molten metal could vary inside our control volume, since the temperature inside the control volume is not constant. Also, there are several chemical reactions and solidification rates occurring in during the impregnation. Given the recent advances in high-performance computing, an in-house pore network simulator are implemented along with Dakota, an open-source, exascale software, to determine the optimal parameters (e.g., porosity and temperature) and uncertainty quantification for the modeling.

At the Holloman High-Speed Test Track (HHSTT), momentum exchange with water is used to decelerate sleds from very high speeds (> Mach 1). This process, at the HHSTT, is called ‘water braking’. Improving the prediction capabilities of water braking phenomena has the potential to result in radical changes in the designs of sleds, improve rocket sled velocity-time test profile predictions, provide greater confidence of braking mechanisms, and decrease risk in the recovery of critical infrastructures. Understanding the water?s behavior with the sled is critical to predicting how the water could damage the sled, which affects the recoverability of the sled and can determine the success of a mission. Traditionally, sled design for the test missions for water braking has been guided by empirical/hand calculations to estimate the forces on various components. The calculations involved various approximations in arriving at the force balance law and predicting the acceleration/deceleration profile. In partnership with the HHSTT, we performed preliminary simulations to develop a predictive model for the HHSTT sled tests at various velocity regimes. The (preliminary) CFD results from different geometry configurations for the sled and modeling parameters will be presented.

Dry powder inhalers (DPIs), used as a means for pulmonary drug delivery, typically contain a combination of active pharmaceutical ingredient (API) and significantly larger carrier particles. The micro-sized drug particles — which have a strong propensity to aggregate and poor aerosolization performance — mixed with significantly large carrier particles that are unable to penetrate the mouth-throat region to deagglomerate and entrain the smaller API particles in the inhaled airflow. The performance of a DPI, therefore, depends on entrainment the carrier-API combination particles and the time and thoroughness of the deagglomeration of the individual API particles from the carrier particles. Since DPI particle transport is significantly affected by particle-particle interactions, very different particles sizes and shapes, various forces including electrostatic and van der Waals forces, they present significant challenges to Computational Fluid Dynamics (CFD) modelers to model regional lung deposition from a DPI. In the current work, we present a novel high fidelity CFD discrete element modeling (CFD-DEM) and sensitivity analysis framework for predicting the transport of DPI carrier and API particles. The work integrates exascale capable CFD-DEM and sensitivity analysis capabilities by leveraging the Department of Energy (DOE) laboratories libraries: Multiphase Flow Interface Flow Exchange (MFiX) for CFD-DEM, and Trilinos for leading-edge portable/scalable linear algebra. We carried out a sensitivity analysis of various formulation properties and their effects on particle size distribution with Dakota, an open source software designed to exploit High-Performance Computing (HPC) capabilities of a massively parallel supercomputer. We developed wrappers to exchange information among these state-of-the-art tools for DPI.

Due to global demand for energy, there is a need to maximize oil extraction from wet reservoir sedimentary formations, which implies the efficient extraction of oil at the pore scale. The approach involves pressurizing water into the wetting oil pore of the rock for displacing and extracting the oil. The two-phase flow is complicated because of the behavior of the fluid flow at the pore scale, and capillary quantities such as surface tension, viscosities, pressure drop, radius of the medium, and contact angle become important. In the present work, we use machine learning algorithms in TensorFlow to predict the volumetric flow rate for a given pressure drop, surface tension, viscosity and geometry of the pores. The TensorFlow software library was developed by the Google Brain team and is one of the most powerful tools for developing machine learning workflows. Machine learning models can be trained on data and then these models are used to make predictions. In this paper, the predicted values for a two-phase flow of various pore sizes and liquids are validated against the numerical and experimental results in the literature.

Liquid metal infiltration consists of infusing liquid metal into a porous media or a packed bed of boron carbide powder to react and create ultimately a metal or ceramic matrix embedded with boride-carbide precipitates. The purpose of the study is to model the liquid flow into the capillaries of the packed bed by using machine learning algorithms from an open source available as TensorFlow library created by Google Brain. The library has a variety of algorithms including training and inference algorithms forming deep neural network models to predict the wetting dynamics, flow resistance, and the depth/rate of penetration into the capillaries of the packed bed. In the present work, the results from the machine-learning python code based on the TensorFlow library is compared against the experimental data obtained for molten Hf-Ti-Y-Zr alloys infiltrating into a packed bed of boron carbide at temperatures up to 2300°C. A summary of the techniques used to tweak the machine learning algorithms to predict the infusion behavior will be presented.

Fragmentation of molten metal droplets is an important process in steam explosions caused by melt-coolant interactions. Ciccarelli and Frost (1994) found the formation of melt jets (or spikes) in hot melt drops immersed in water. In order to gain insight into this mechanism, they carried out experiments where melt jets were formed in a stratified water/liquid metal system with local generation of high-pressure vapor at the interface. This paper is dedicated to investigating how melt jets are formed in this mechanism when a stratified water/liquid metal system is analyzed. Also, a study of the most significant parameters in this process is performed. A 2D computational fluid dynamics (CFD) simulation is carried out using ANSYS Fluent software to study these phenomena by having water above hot liquid metal, a vapor film in between and a pressure pulse in the vapor film. The results show that the larger the pressure or density, the greater the melt jet length. In order to confirm this, deep neural network algorithm created by TensorFlow library was implemented to facilitate the understanding of the studied phenomena. The formation of melt jets observed in Ciccarelli and Frost’s experiments is also observed in current simulation.