Research

One of the most significant engineering and societal challenges of today is global warming and space exploration. To reduce global warming, we need to reduce the pollution from vehicles & engines, produce more clean energy, and reduce the cost to generate clean energy. The desire to explore space and make a supersonic airliner needs vehicles that can efficiently operate at various speeds. There is no unique solution to achieve these goals, and a diverse portfolio of technology initiatives is needed.

My expertise and proposed work focus on air-breathing engines, renewable energy, and carbon capture and sequestration. While trying to solve problems in these domains, I always appreciate the vast and interdisciplinary nature of the problem, wherein I employ principles of Computational Science, Material Science, Chemistry, and Data Science in conjunction with those of Aerospace/Mechanical Engineering.

Professional Research Accomplishments

My research work during my academic appointments has focused on understanding coupled fluid flow, heat and mass transfer in multi-phase, multi-species systems to improve the fluidization. As a graduate student, I studied the coupled fluid flow, and heat transfer in components of air-breathing engines and improved their performance.

Components of air-breathing engines

Air-breathing propulsion systems use atmospheric air to oxidize the liquid fuel. The atmospheric air is brought to the engine face via an inlet. This air is burned with the fuel in the combustor and creates hot air. The nozzle is an exhaust duct of the engine which generates thrust by expanding the hot stream as efficiently as possible to ambient pressure.

Figure 1: Flow over a flat plate with 6-noded linear and 18-noded quadratic wedge elements.[J2]

I developed a compressible Navier-Stokes equations solver based on stabilized finite element method and carried out computations to compare the performance of linear and quadratic elements. In three dimensions, the relative performance is evaluated for 6-noded linear and 18-noded quadratic wedge elements. I compared the results for the solutions to Euler, laminar, and turbulent flows at different Mach numbers for several flow problems (see Fig. 1). The finite element meshes considered for comparison have the same location of nodes for the linear and quadratic interpolations. Perhaps for the first time, a stabilized finite element method with higher-order interpolation C0 functions has been utilized to solve a 2D turbulent flow via Reynolds-Averaged Navier-Stokes equations and 3D laminar flow. I found that the quadratic elements yield better performance than the linear elements. For the same location of nodes, the computations with the linear triangular and wedge elements are approximately 20% and 100% faster than those with quadratic triangular and wedge elements, respectively. However, if the same quadrature rule for numerical integration is used for both interpolations, the computations with quadratic elements are approximately 20% and 45% faster in 2D and 3D, respectively. It is well known that higher-order methods provide geometric flexibility and are more efficient than low order methods for problems that involve wave propagation, vortex-dominated flows, including high-lift configuration. They also present a richer space for higher accuracy and might be especially attractive for conducting large-eddy simulations and direct numerical simulations of turbulence.

Further, I studied the flow in a convergent-divergent (CD) nozzle numerically for various exit to throat area ratios and pressure ratios (NPR). I developed a novel method and it allows the generation of different nozzle geometries with no discontinuity in slope in the diverging section. I identified various flow regimes that are possible in the entire parameter space. I observed asymmetric flow at moderate NPR (see Fig. 2) following earlier experimental and computational studies. The NPR range for which asymmetric flow is noted increases with an increase in area ratio. The side loads due to the flow asymmetry increase with the increase in area ratio and NPR. I studied the effect of the boundary layer bleed on the performance of the nozzle. I observed that bleed delays the flow separation and makes the flow steady and symmetric for all operating conditions. The bleed suppresses the side loads which may damage the nozzle.

Figure 2: Asymmeric flow in a planar CD nozzle[J7].
Figure 3: Unsteady flow (Mach number) in a Y-intake at various time instants during buzz cycle [J5]

In air intakes, buzz instability may appear at low mass flow rates. It involves periodic filling and discharge of the plenum chamber, shock-boundary layer interaction, transient shock movement, and flow separation. Asymmetric flow in the two limbs of the Y-intake produces an imbalance of pressure forces on the airplane, which may lead to severe problems with the stability and control of the aircraft. I simulated flow in a Y-intake, at freestream Mach number 1.5 for sideslip angles of 0 to 5 and backpressure ratio of 2.1 to 3.44. I identified various flow regimes that are possible in the entire parameter space. The flow regimes are classified based on shock structure and their locations and the unsteadiness in the flow. I observed buzz instability in the intake at low system mass flow rates for non-zero sideslip angles (see Fig. 3). I found hysteresis in the flow, as well as in the onset of buzz. I studied the effect of bleed on the performane of the intake. The amount of bleed and the regions on which it is applied is varied to understand their impact on the performance of the intake.

Further, I studied, flow in a two-dimensional mixed compression intake at a freestream Mach number of 3.0. I varied the amount of bleed and its location to understand its effect on the performance of the intake. I observed two kinds of unsteady oscillations: ’little’ and ’big’ buzz (see Fig. 4). The frequency of both types of buzz oscillations is found to be super-harmonic of the fundamental acoustic frequency of intake modeled as an open-closed organ pipe. I eliminated buzz by implementing 20% bleed on both upstream and downstream of the throat. I studied the effect of rate of change of back pressure on the start/unstart of the intake. Two situations are considered. The first case is that of intake, where the back pressure remains below the critical value. The second set of simulations is an attempt to model the situation where the back-pressure at the exit of the intake exceeds the critical value, and logic is incorporated in the feedback loop of the fuel modulation to start the intake. I studied the effect of back pressure on the performance of the intake via the total pressure recovery and distortion index. In this work, I identified the regions in which intake can operate in the instability region. Also, I studied the effectiveness of the boundary bleed to suppress these instabilities so that the performance of the propulsion system is good and eliminate catastrophic loss in thrust.

Figure 4: unsteady flow (Mach number) in a mixed compression intake at various time instants during little (left) and big buzz cycle.[J1,J3,J4]

Fluidized beds in the conversion of fossil hydrocarbons to electricity

Multiphase flows are a ubiquitous feature of our environment, whether one considers rain, snow, fog, avalanches, mudslides, sediment transport, and debris flows. Very critical biological and medical flows are also multiphase, from blood flow to semen to the bends to lithotripsy to laser surgery cavitation and so on. These are classified according to the state of the different phases or components and, therefore, refer to gas/solids flows, or liquid/solids flow, or gas/particle flows or bubbly flows, and so on. Some are defined in terms of a specific type of fluid flow and deal with low Reynolds number suspension flows, dusty gas dynamics and so on. Others focus attention on a particular application such as slurry flows, cavitating flows, aerosols, debris flows, and fluidized beds and so on. Fluidized beds are generally used in petroleum, pharmaceutical, chemical, mineral, and fossil fuel plants. Gasification of a feedstock in a fluidized bed increases the efficiency of the power plant and reduces greenhouse gases.

MFiX (Multiphase Flow with Interphase eXchanges)[1], a general-purpose computer code developed at the NETL for studying dynamics of fluid along with thermal effects and chemical reactions in fluidized beds, spends more than 70% of the total execution time in solving the sparse linear system. It lacks advanced iterative solvers, preconditioners, and direct solvers. Therefore, the integration of such solvers and preconditioners in Trilinos [2] with MFiX can improve the capabilities of MFiX and can solve larger and more complex problems. Trilinos, an open-source, object-oriented framework, contains robust algorithms and enabling technologies for solving large-scale multiphysics problems. The library consists of more than 60 packages, including preconditioners, linear, nonlinear, direct, transient and optimization solvers, multi-grid preconditioners and many more.

My postdoctoral training at UTEP allowed me to develop an interface (see Fig. 5) to integrate the advanced linear solvers in Trilinos with MFiX. Trilinos provides a framework for simulating large-scale, complex multi-physics engineering and scientific problems. The interface is written in Fortran and C/C++. The interface has been verified and validated on various fluid bed problems (see Fig, 6). This interface can be applied to integrate the software’s written in two different programming languages, such as Fortran and C/C++. Two fluid, discrete element and particle in cell approaches are used to simulate the flow in fluidized beds. I also tested the performance of the linear solvers, which are based on the Kokkos programming model, on various computer architectures. The iterative solvers in the integrated multiphase flow solver are more than 1.5 times faster than built-in solvers in MFiX. Further, I designed and developed a framework to integrate MFiX with Dakota for uncertainty quantification, sensitivity analysis, and optimization of multi-physics problems. However, I also developed an exascale capable pore-network simulator and integrated it with Dakota.

As a postdoc at UWyo, I developed a framework (see Fig.7) to implement the P-1 radiation model in MFiX for Eulerian-Eulerian (Two fluid-TFM) and Eulerian-Lagrangian (Discrete element-DEM and Particle in cell-PIC) models. The P-1 is a simplification of the spherical harmonics method. The framework has been verified on various flue gas conditions. The results with the verified solver are validated against the benchmark results available in the literature. The verified and validated gray and non-gray weighted sum of gray gases models are evaluated for dry and wet flue gas conditions. In addition, they are used to simulate gas-solid flow in a riser (see Fig.8). Significant differences between the temperatures is observed in regions where the gasification of the coal occurs.

[1] M. Syamlal, W. Rogers, and T. J. O’Brien, “MFIX documentation: Theory guide,” National Energy Technology Laboratory, Department of Energy, Technical Note DOE/METC-95/1013 and NTIS/DE95000031, 1993.

[2] M. A. Heroux et al., “An overview of the Trilinos project,” ACM Transactions on Mathematical Software (TOMS), vol. 31, no. 3, pp. 397–423, 2005.

Figure 5: A framework to integrate MFiX with Trilinos[J8].
Figure 6: Flow in a bubbling bed with Two-fluid (TFM), discrete element (DEM) and particles in cell (PIC) methods. [J8]

Data-driven framework for Uncertainty Quantification/Sensitivity analysis/Optimization

Turbulence is significant in many engineering applications, and modeling of these flows poses unique challenges due to complex nonlinear interactions that can involve, for example, effects of multiple physical processes, density differences between fluids, and broad ranges of spatial and temporal scales. Machine learning has demonstrated to be an effective and promising approach to investigate and solve problems in many areas of physics, including turbulence. Machine learning offers computers to learn from datasets without explicit programming instructions and application of machine learning is ubiquitous in our everyday life. The datasets required for training are widely available these days. In recent years, machine learning has driven advances in many different fields. This is due to the improvement in the machine learning models, availability of large datasets for training and readiness of computational resources for training the models with large datasets.

Various widely used open-source or commercial solvers are available to us to simulate complex engineering problems including fluidized beds. However, most of these solvers do not provide meaningful confidence intervals of the simulation results. It could be due to the lack of tools for finding the ranges as well as the expense of the computations. We know that these intervals are beneficial for verification and validation of the numerical results. In general, nonlinearities and transient behavior of the flows increase the computational cost substantially. Further, species and chemical reactions consideration in fluidized bed reactors significantly impact the computational cost. Besides, multiphase flows such as fluidized beds involve numerous uncertain parameters. Techniques such as sensitivity analysis are useful in identifying the few metrics that have the most influence on the quantities of interest. Further, fluidized beds are more complex than single phase flow simulations, and the computational cost increase plays a crucial role in finding a sampling technique and number of samples.

Water Braking Phenomena for the Holloman High-Speed Test Track

The Holloman High Speed Test Track operated by the Air Force Test Center 846th Test Squadron, is the world’s premier rocket sled test track and has the longest facility of its type in the world, making it one of the most unique test facilities in the DoD capable of replicating operational flight profiles, providing accurate and reliable data to the USAF, Army, Navy and other government agencies for Test and Evaluation (T&E) of critical weapon system and aerospace technology at fraction of the cost to flight testing. Operational flight speeds (can reach > Mach 8) is achieved via rail-mounted rocket-propelled sleds where the test object is attached to a forebody sled driven by one or more pusher sleds to accelerate the object. The sleds need to be recovered on the rail via a combination of aerodynamic drag followed by entry into water braking mechanism, a poorly understood phenomenon due to complex nonlinear multiphase flow dynamics interaction. Accurate prediction of the test profile can result in radical changes to designs of specific sleds and provide greater confidence of braking mechanism and recovery of the critical AF infrastructures. In collaboration with the squadron & supported by rigorous verification and validation, we propose to develop a better predictive capability of the water braking phenomena with high-fidelity Computational Fluid Dynamics (CFD) investigation capable of resolving flow separation, boundary layer, sled and rail/concrete interactions, test vehicles and articles for the test.

Laser propagation in atmospheric turbulence

Within the department of defense, there is an expanding interest to characterize the effects of atmospheric turbulence on the optical system. We essentially want to comprehend laser propagation over a long path and understand the effects of atmospheric turbulence. This is because laser propagation through a turbulent medium can result in refractive index fluctuations that cause random phase perturbations on a laser beam that can lead to beam distortions. Improvement and understanding of optical systems capable of propagating lasers over long distances in atmospheric turbulence conditions can enhance the performance in remote sensing, long-range satellite communications, active imaging, and other related optical systems. However, it can be complicated to characterize these random variations in optical systems associated with the temperature variations that are caused by turbulent eddies.

Dry powder inhalers

Dry powder inhalers (DPIs) are a combination of active pharmaceutical ingredient (API) and significantly larger carrier particles, which are widely used for pulmonary drug delivery. The micro-sized drug particles - which have a vigorous intensity to aggregate and poor aerosolization performance and mixed with significantly large carrier particles that are unable to penetrate the mouth-throat region to deagglomerate but help the smaller API particles to entrain in the inhaled airflow. The performance of a DPI depends on entrainment of the carrier-API combination particles and the time and thoroughness of the deagglomeration of the individual API particles from the carrier particles. Since API particle transport is significantly affected by particle-particle interactions, properties of the fluidizing agent, different particle sizes and density, drag force between the particles and fluidizing agent, they present significant challenges to CFD modelers to model regional lung deposition from a DPI.

The evolution of drug and carrier particles in a dry powder inhaler


Dry powder inhalers (DPI) directly deliver the drug to the target areas and provide more drug choices, especially for those poorly absorbed orally. These methods result in the rapid onset of activity with small doses compared to injection and oral administration. Also, DPIs do not contain chlorofluorocarbons (CFC). On the other hand, metered-dose inhalers contain CFC and damages the ozone layer. In DPI applications, drug particles of size less than 5μm are used. In the inhalation process, the particles are mixed with larger carrier particles that are very cohesive and have poor flowability.

Poro-elasticity

Poroelasticity is the study of the transitory interactions between solid deformation and fluid flow within a porous medium. It is characterized by a time dependent, two-way coupling where changes in the state variables of one phase alter those of the other. Applying external load to a saturated porous medium causes changes in the fluid pressures which induces a flow. Likewise, a change in fluid pressures induces stresses that deform the solid skeleton.

Improving the performance and portability of legacy codes via an object-oriented software framework, Trilinos

Trilinos [2], an open-source software library, contains robust algorithms and enabling technologies for solving large-scale flow problems. The library comprises various packages for preconditioning, solving the linear/ non-linear system, uncertainty quantification, optimization, and many more. The library offers different ways for a particular package to interact with other Trilinos packages. The advanced linear equation solver packages in Trilinos offers portability, scalability, excellent performance, and ease of integration with other legacy codes. The intention of developing this library was to provide a framework that gives application writers advanced programming semantics offered by the new C++11 framework like generic template programming and lambdas. Moreover, it provides inherent parallelism to building and distributed matrices through inbuilt classes that offer a multitude of methods enabling scientific computing. I developed an interface to integrate the advanced linear solvers and preconditioners in Trilinos with a legacy Fortran language-based open-source multiphase flow solver. I studied the performance, as well as portability of the iterative methods in the integrated solvers on various supercomputers. I use this framework to integrate other legacy flow solvers with Trillions and simulate large scale engineering problems such as flow-through billions of pores in a porous media and interactions between billions of particles and continuous phase in a fluidized bed. Also, leverage the advanced portable and scalable linear solvers and preconditioners via the framework to improve the performance as well as portability of various legacy flow solvers on various hybrid computer architectures.

I want to thank the institutions that have provided access to their parallel machines: