Browsing by Subject "Computational fluid dynamics"

Now showing 1 - 8 of 8

Computational Analysis of Energy Exchange Mechanisms in Turbulent Flows with Thermal Nonequilibrium
(2018-01) Neville, Aaron
The design of hypersonic vehicles is signiﬁcantly aﬀected by the state of boundary layer. Hypersonic boundary layers can be laminar or turbulent, and in chemical and vibrational nonequilibrium, each with diﬀerent length and time scales. In turbulent boundary layers, heating augmentation can be an order of magnitude or higher above laminar heating rates. The scales of the internal energy relaxation processes can be of the same order or greater than the turbulent ﬂow scales, and can interact with turbulent motion. Understanding how turbulent motion and internal energy relaxation interact is relevant to ﬂow control. Fundamental ﬂows are studied to understand how energy is exchanged between turbulent motion and internal energy relaxation. Speciﬁcally, high-ﬁdelity DNS of vibrational energy relaxation eﬀects in compressible isotropic and temporally evolving shear layers are presented. The energy exchange mechanisms are analyzed by decomposing the ﬂow into incompressible and compressible energy modes. By varying the vibrational relaxation rate, the tuning of the relaxation rate to the turbulent ﬂow is studied. Vibrational energy relaxation is demonstrated to be coupled to the turbulent ﬂow through the compressible modes of the gas. Compressions and expansions generate ﬂuctuations in the thermal state, and the vibrational energy lags behind these ﬂuctuations. Energy is then transferred to or from the vibrational energy mode at a rate proportional to the relaxation time, and the ﬂuctuations are damped. Damping of turbulent quantities are strongest when the vibrational relaxation rate is on the order of the turbulent large structure acoustic rate. Wavenumber speciﬁc damping is also observed in isotropic ﬂows when the relaxation time is on the order of the acoustic frequency of the wave. The eﬀects of vibrational relaxation are shown to increase with compressibility. However, the overall eﬀect on turbulent kinetic energy is weak due to the incompressible mode containing signiﬁcantly more energy than the compressible modes.
Large Eddy Simulation of crashback in marine propulsors.
(2011-06) Jang, Hyunchul
Crashback is an operating condition to quickly stop a propelled vehicle, where the propeller is rotated in the reverse direction to yield negative thrust. The crashback condition is dominated by the interaction of the free stream flow with the strong reverse flow. This interaction forms a highly unsteady vortex ring, which is a very prominent feature of crashback. Crashback causes highly unsteady loads and flow separation on the blade surface. The unsteady loads can cause propulsor blade damage, and also affect vehicle maneuverability. Crashback is therefore well known as one of the most challenging propeller states to analyze. This dissertation uses Large-Eddy Simulation (LES) to predict the highly unsteady flow field in crashback. A non-dissipative and robust finite volume method developed by Mahesh et al. (2004) for unstructured grids is applied to flow around marine propulsors. The LES equations are written in a rotating frame of reference. The objectives of this dissertation are: (1) to understand the flow physics of crashback in marine propulsors with and without a duct, (2) to develop a finite volume method for highly skewed meshes which usually occur in complex propulsor geometries, and (3) to develop a sliding interface method for simulations of rotor-stator propulsor on parallel platforms. LES is performed for an open propulsor in crashback and validated against experiments performed by Jessup et al. (2004). The LES results show good agreement with experiments. Effective pressures for thrust and side-force are introduced to more clearly understand the physical sources of thrust and side-force. Both thrust and side-force are seen to be mainly generated from the leading edge of the suction side of the propeller. This implies that thrust and side-force have the same source - the highly unsteady leading edge separation. Conditional averaging is performed to obtain quantitative information about the complex flow physics of high- or low- amplitude events. The events for thrust and side force show the same tendency. The conditional averages show that during high amplitude events, the vortex ring core is closer to the propeller blades, the reverse flow induced by the propeller rotation is lower, the forward flow is higher at the root of the blades, and leading and trailing edge flow separations are larger. The instantaneous flow field shows that during low amplitude events, the vortex ring is more axisymmetric and the stronger reverse flow induced by the vortex ring suppresses the forward flow so that flow separation on the blades is smaller. During high amplitude events, the vortex ring is less coherent and the weaker reverse flow cannot overcome the forward flow. The stronger forward flow makes flow separation on the blades larger. The effect of a duct on crashback is studied with LES. Thrust mostly arises from the blade surface, but most of side-force is generated from the duct surface. Both mean and RMS of pressure are much higher on inner surface of duct, especially near blade tips. This implies that side-force on the ducted propulsor is caused by the blade-duct interaction. Strong tip leakage flow is observed behind the suction side at the tip gap. The physical source of the tip leakage flow is seen to be the large pressure difference between pressure and suction sides. The conditional average for high amplitude event shows consistent results; the tip leakage flow and pressure difference are significantly higher when thrust and side-force are higher. A sliding interface method is developed to allow simulations of rotor-stator propulsor in crashback. The method allows relative rotations between different parts of the computational grid. Search algorithm for sliding elements, data structures for message passing, and accurate interpolation scheme at the sliding interface are developed for arbitrary shaped unstructured grids on parallel computing platforms. Preliminary simulations of open propulsor in crashback show reasonable performance.
Methods for the Modeling and Simulation of Sprays and Other Interfacial Flows
(2019-09) Wenzel, Everett
Interfacial multiphase flows involve the motion of at least two fluids separated by surface tension. Atomizing interfacial flows, colloquially known as sprays, are among the most important fluid dynamic systems because of their ubiquity; power generation, delivery of aerosolized medicines, and productive produce farming all depend fundamentally on the detailed control of sprays. Atomization remains poorly understood because of a historical and persisting inability to accurately and affordably measure the dynamics inside and near the spray orifice outlet -- it is therefore desirable to be able to numerically simulate sprays with high fidelity. This dissertation presents computational methods that aim to improve current shortcomings in the modeling and simulation of sprays. Accurately characterizing the interfacial curvature of poorly-resolved liquid structures is addressed by deriving a series of finite particle methods for computing curvature. The methods are verified in analytical curvature tests, and validated against the oscillation frequency of ethanol droplets in air. The finite particle method, leveraging dynamic length scale modification, is demonstrated to out-perform the widely-used height function approach. Tracking the location of interfaces is also addressed, for which a coupled Eulerian-Lagrangian point mass particle scheme is introduced that preserves a well-distributed particle field, can be applied to an arbitrary number of fluids, and does not limit the simulation time step. The Eulerian-Lagrangian method is demonstrated to out-perform contemporary geometric volume of fluid methods at resolutions relevant to spray simulation in a variety of analytical phase tracking tests, and is dynamically evaluated by simulating extending three-phase elliptical regions, droplet dynamics, and Rayleigh-Taylor instabilities. The Eulerian-Lagrangian method is then extended to an approach for consistently and conservatively solving multiphase convection-diffusion problems -- this extension is verified via two analytical heat transfer problems, and robustness is demonstrated by simulating heated air blast atomization. Each of these tests conserves thermal energy and preserves boundedness of the temperature field. This dissertation concludes by outlining paths for consistently and conservatively solving the multiphase Navier-Stokes equations and the multiphase large eddy simulation equations in the coupled Eulerian-Lagrangian point mass particle framework.
Modeling and control of cadmium zinc telluride grown via an electro-dynamic gradient freeze furnace.
(2007-12) Lun, Lisa San
In this thesis, numerical models are used to study the effect of novel processing methods to grow bulk, single crystal cadmium zinc telluride (CZT) in a vertical Bridgman (VB) furnace. Additionally, we investigate new mathematical algorithms for improved solving capability of equations that describe such crystal growth systems. A two-dimensional crystal growth model for the simulation of bulk crystal growth in a VB system is presented. This model consists of conservation equations for coupled continuum level transport of heat, mass, and momentum. Thermodynamic relations associated with phase change are also included. The Galerkin finite element method is used to discretize the spatial portion of the governing equations. The resulting sets of nonlinear algebraic equations are solved using Newton's method. Novel processing methods that are not practical to attempt in experiments are investigated using numerical modeling. A two-dimensional, planar, crystal growth model is used to explore the effect of ampoule tilting on zinc segregation in a CZT crystal. Tilting is shown to improve lateral segregation. We also analyze the use of closed-loop control to improve the macroscopic melt-crystal interface shape during growth by changing the furnace temperature gradient. Targeted closed-loop control on the temperature gradient adjacent to the solid only gave the best results and unexpectedly produced a favorable convex shape. A multi-scale crystal growth model is developed by coupling pre-existing codes, one which specializes in modeling the complex crystal growth process and the other which specializes in modeling the heat transfer effects in a furnace. Previously, a coupling algorithm based on the Gauss-Seidel method was used but it converged unreliably [136, 196]. Here, we use an Approximate Block Newton approach where we approximate Newton's method used to solve the two separate models as if they were one monolithic model. A Schur complement formulation is employed to solve the free-boundary problem associated with melt crystal growth systems. With this form, the difficult interface location part of the problem is mapped away from the equations governing transport. We assess the behavior of this method using two-dimensional simulations, but the goal is to improve solvability of three-dimensional problems.
Modeling of Concentrated High Intensity Electric Field (CHIEF) and Its Comparison with Other Non-thermal Liquid Food Pasteurization Technologies
(2015-12) Peng, Peng
Non-thermal preservations of food have received rising attention due to the increase concern of environmental sustainability and the demand of safer food with improved nutritional functionalities. High pressure and electric field treatment are two non-thermal food treatment strategies that have been widely studied. Some representatives of non-thermal technologies that utilize high-pressure and electric field to pasteurize food products include High hydrostatic pressure (HHP), high-pressure homogenization (HPH), and pulsed electric field (PEF). These non-thermal technologies, together with concentrated high intensity electric field (CHIEF) are studied and compared in this thesis research. This study used finite element (FEM) and computational fluid dynamics (CFD) methods to model and simulate the fluid flow, electric field distribution and temperature rise in CHIEF reactor. The simulation was confirmed to be valid by comparing it with experimental results. The model built in this study showed that the performance of CHIEF system was influenced by a set of intrinsic and extrinsic parameters. This model could be used to control and set variables in further optimization of the CHIEF system. Each of the non-thermal technologies discussed in this study has its advantages and unique field of use. HHP, dynamic high-pressure treatment and PEF are relatively mature technologies, while CHIEF system is an innovative and promising non-thermal method that can potentially be used as alternative to PEF.
Multiscale Computational Analysis of Nitrogen and Oxygen Gas-Phase Thermochemistry in Hypersonic Flows
(2016-02) Bender, Jason
Understanding hypersonic aerodynamics is important for the design of next generation aerospace vehicles for space exploration, national security, and other applications. Ground-level experimental studies of hypersonic flows are difficult and expensive; thus, computational science plays a crucial role in this field. Computational fluid dynamics (CFD) simulations of extremely high-speed flows require models of chemical and thermal nonequilibrium processes, such as dissociation of diatomic molecules and vibrational energy relaxation. Current models are outdated and inadequate for advanced applications. We describe a multiscale computational study of gas-phase thermochemical processes in hypersonic flows, starting at the atomic scale and building systematically up to the continuum scale. The project was part of a larger effort centered on collaborations between aerospace scientists and computational chemists. We discuss the construction of potential energy surfaces for the N4, N2O2, and O4 systems, focusing especially on the multi-dimensional fitting problem. A new local fitting method named L-IMLS-G2 is presented and compared with a global fitting method. Then, we describe the theory of the quasiclassical trajectory (QCT) approach for modeling molecular collisions. We explain how we implemented the approach in a new parallel code for high-performance computing platforms. Results from billions of QCT simulations of high-energy N2 + N2, N2 + N, and N2 + O2 collisions are reported and analyzed. Reaction rate constants are calculated and sets of reactive trajectories are characterized at both thermal equilibrium and nonequilibrium conditions. The data shed light on fundamental mechanisms of dissociation and exchange reactions – and their coupling to internal energy transfer processes – in thermal environments typical of hypersonic flows. We discuss how the outcomes of this investigation and other related studies lay a rigorous foundation for new macroscopic models for hypersonic CFD. This research was supported by the Department of Energy Computational Science Graduate Fellowship and by the Air Force Office of Scientific Research Multidisciplinary University Research Initiative.
Performance portability strategies for Computational Fluid Dynamics (CFD) applications on HPC systems
(2013-06) Lin, Pei-Hung
Achieving high computational performance on large-scale high performance computing (HPC) system demands optimizations to exploit hardware characteristics. Various optimizations and research strategies are implemented to improve performance with emphasis on single or multiple hardware characteristics. Among these approaches, the domain-specific approach involving domain expertise shows its high potential in achieving high performance and maintaining performance portability. Deep memory hierarchies, single instruction multiple data (SIMD) engines, and multiple processing cores in the latest CPUs pose many challenges to programmers seeking significant fractions of peak performance. Programming for high performance computation using modern CPUs has to address thread-level parallelization on multiple cores, data-level parallelization on SIMD engines, and optimizing memory utilization for the multi-level memories. Using multiple computational nodes with multiple CPUs in each node to scale up the computation without sacrificing performance increases programming burden significantly. As a result, performance portability has become a major challenge to programmers. It is well known that manually tuned programs can assist the compiler to deliver the best performance. However, generating these optimized codes requires deep understanding in application design, hardware architecture, compiler optimizations, and knowledge in the specific domain. Such manually tuning process has to be done for each new hardware design. To address this issue, this dissertation proposes strategies that exploit the advantages of domain-specific optimizations to achieve performance portability. This dissertation shows the combination of the proposed strategies can effectively exploit both the SIMD engine and on-chip memory. High fraction of peak performance can be achieved after such optimizations. The design of the pre-compilation framework makes it possible to automate these optimizations. Adopting the latest compiler techniques to assist domain-specific optimizations has high potential to implement sophisticated and legal transformations. This dissertation provides a preliminary study using polyhedral transformations to implement the proposed optimization strategies. Several obstacles need to be removed to make this technique applicable to large-scale scientific applications. With the research presented in this dissertation and suggested tasks in the future work, the ultimate goal to deliver performance portability with automation is feasible for CFD applications.
A stochastic particle method for the investigation of turbulence/chemistry interactions in large-eddy simulations of turbulent reacting flows
(2013-12) Ferrero, Pietro
The main objective of this work is to investigate the effects of the coupling between the turbulent fluctuations and the highly non-linear chemical source terms in the context of large-eddy simulations of turbulent reacting flows. To this aim we implement the filtered mass density function (FMDF) methodology on an existing finite volume (FV) fluid dynamics solver. The FMDF provides additional statistical sub-grid scale (SGS) information about the thermochemical state of the flow - species mass fractions and enthalpy - which would not be available otherwise. The core of the methodology involves solving a transport equation for the FMDF by means of a stochastic, grid-free, Lagrangian particle procedure.Any moments of the distribution can be obtained by taking ensemble averages of the particles. The main advantage of this strategy is that the chemical source terms appear in closed form so that the effects of turbulent fluctuations on these terms are already accounted for and do not need to be modeled.We first validate and demonstrate the consistency of our implementation by comparing the results of the hybrid FV/FMDF procedure against model-free LES for temporally developing, non-reacting mixing layers. Consistency requires that, for non-reacting cases, the two solvers should yield identical solutions. We investigate the sensitivity of the FMDF solution on the most relevant numerical parameters, such as the number of particles per cell and the size of the ensemble domain. Next, we apply the FMDF modeling strategy to the simulation of chemically reacting, two- and three-dimensional temporally developing mixing layers and compare the results against both DNS and model-free LES. We clearly show that, when the turbulence/chemistry interaction is accounted for with the FMDF methodology, the results are in much better agreement to the DNS data. Finally, we perform two- and three-dimensional simulations of high Reynolds number, spatially developing, chemically reacting mixing layers, with the intent of reproducing a set of experimental results obtained at the California Institute of Technology. The mean temperature rise calculated by the hybrid FV/FMDF solver, which is associated with the amount of product formed, lies very close to the experimental profile. Conversely, when the effects of turbulence/chemistry coupling are ignored, the simulations clearly over predict the amount of product that is formed.

University Digital Conservancy

Browse by Subject

Browsing by Subject "Computational fluid dynamics"