No description available.

High-level parallel programming approaches, such as OpenACC, have recently become popular in complex fluid dynamics research since they are cross-platform and easy to implement. OpenACC is a directive-based programming model that, unlike low-level programming models, abstracts the details of implementation on the GPU. Although OpenACC generally limits the performance of the GPU, this model significantly reduces the work required to port an existing code to any accelerator platform, including GPUs. The purpose of this research is twofold: to investigate the effectiveness of OpenACC in developing a portable and maintainable GPU-accelerated code, and to determine the capability of OpenACC to accelerate large, complex programs on the GPU. In both of these studies, the OpenACC implementation is optimized and extended to a multi-GPU implementation while maintaining a unified code base. OpenACC is shown as a viable option for GPU computing with CFD problems. In the first study, a CFD code that solves incompressible cavity flows is accelerated using OpenACC. Overlapping communication with computation improves performance for the multi-GPU implementation by up to 21%, achieving up to 400 times faster performance than a single CPU and 99% weak scalability efficiency with 32 GPUs. The second study ports the execution of a more complex CFD research code to the GPU using OpenACC. Challenges using OpenACC with modern Fortran are discussed. Three test cases are used to evaluate performance and scalability. The multi-GPU performance using 27 GPUs is up to 100 times faster than a single CPU and maintains a weak scalability efficiency of 95%. / Master of Science / The research and analysis performed in scientific computing today produces an ever-increasing demand for faster and more energy efficient performance. Parallel computing with supercomputers that use many central processing units (CPUs) is the current standard for satisfying these demands. The use of graphics processing units (GPUs) for scientific computing applications is an emerging technology that has gained a lot of popularity in the past decade. A single GPU can distribute the computations required by a program over thousands of processing units. This research investigates the effectiveness of a relatively new standard, called OpenACC, for offloading execution of a program to the GPU. The most widely used standards today are highly complex and require low-level, detailed knowledge of the GPU’s architecture. These issues significantly reduce the maintainability and portability of a program. OpenACC does not require rewriting a program for the GPU. Instead, the developer annotates regions of code to run on the GPU and only has to denote high-level information about how to parallelize the code. The results of this research found that even for a complex program that models air flows, using OpenACC to run the program on 27 GPUs increases performance by a factor of 100 over a single CPU and by a factor of 4 over 27 CPUs. Although higher performance is expected with other GPU programming standards, these results were accomplished with minimal change to the original program. Therefore, these results demonstrate the ability of OpenACC to improve performance while keeping the program maintainable and portable.

A two-dimensional (2-D), mathematical model is adopted to investigate the development of circulation patterns for compressible, laminar, and shear driven flow inside a rectangular cavity. The bottom of the cavity is free to move at a specified speed and the aspect ratio of the cavity is changed from 1.0 to 1.5. The vertical sides and the bottom of the cavity are assumed insulated. The cavity is filled with a compressible fluid with Prandtl number, Pr =1. The governing equations are solved numerically using the commercial Computational Fluid Dynamics (CFD) package ANSYS FLUENT 2015 and compared with the results for the primitive variables of the problem obtained using in house CFD code based on Coupled Modified Strongly Implicit Procedure (CMSIP). The simulations are carried out for the unsteady, lid driven cavity flow problem with moving boundary (bottom) for different Reynolds number, Mach numbers, bottom velocities and high initial pressure and temperature.

Os mecanismos de instabilidade hidrodinâmica têm um papel importante no processo da transição do escoamento de laminar para turbulento. A análise da instabilidade hidrodinâmica em uma cavidade com tampa deslizante foi realizada através da decomposição em modos globais (biglobal) para avaliar o efeito da compressibilidade neste fenômeno. O escoamento base foi obtido através de simulação numérica direta (DNS). Para tal, foi desenvolvido um código DNS compressível com discretização espacial por diferenças finitas compactas de alta resolução espectral e capacidade de processamento paralelo, com um método de decomposição de domínio que mantém a precisão das diferenças finitas compactas. O escoamento base é usado para montar o problema de autovalor oriundo das equações de Navier-Stokes linearizadas para a perturbação, discretizadas por diferenças finitas explícitas. O uso de diferenças finitas em conjunto com a implementação em matrizes esparsas reduz sensivelmente o uso de memória. Através do algoritmo de Arnoldi, a ordem do problema de autovalor é reduzida e os autovalores de interesse são recuperados. Os resultados indicam o efeito estabilizante da compressibilidade nos modos dominantes da cavidade e revelam modos inerentes ao escoamento compressível, para os quais a compressibilidade tem efeito desestabilizante. Dentre estes modos compressíveis, estão presentes modos de propagação sonora em dutos e modos relacionados à geração de som na cavidade. / Hydrodynamic instability mechanisms play an important role in laminar to turbulent transition. Hydrodynamic instability analysis of a lid-driven cavity flow was performed by global mode decomposition (biglobal) to evaluate compressibility effects on this phenomenon. The basic flow was calculated by direct numerical simulation (DNS). A compressible DNS code was developed with spectral-like compact finite difference spatial discretization. The code allows parallel processing with a domain decomposition method that preserves the compact finite difference accuracy. The basic flow is used to form the eigenvalue problem associated to the linear Navier- Stokes equations for the perturbation, which were discretized by an explicit finite difference scheme. The combination of sparse matrix techniques and finite difference discretization leads to a significant memory reduction. The order of the eigenvalue problem was reduced using the Arnoldi algorithm and the eigenvalues of interest were calculated. Results show the stabilizing effect of compressibility on the leading modes and reveal some modes intrinsic to compressible flow, for which compressibility has a destabilizing effect. Among these compressible modes, there are some related to sound propagation in ducts and to sound generation inside the cavity.

Os mecanismos de instabilidade hidrodinâmica têm um papel importante no processo da transição do escoamento de laminar para turbulento. A análise da instabilidade hidrodinâmica em uma cavidade com tampa deslizante foi realizada através da decomposição em modos globais (biglobal) para avaliar o efeito da compressibilidade neste fenômeno. O escoamento base foi obtido através de simulação numérica direta (DNS). Para tal, foi desenvolvido um código DNS compressível com discretização espacial por diferenças finitas compactas de alta resolução espectral e capacidade de processamento paralelo, com um método de decomposição de domínio que mantém a precisão das diferenças finitas compactas. O escoamento base é usado para montar o problema de autovalor oriundo das equações de Navier-Stokes linearizadas para a perturbação, discretizadas por diferenças finitas explícitas. O uso de diferenças finitas em conjunto com a implementação em matrizes esparsas reduz sensivelmente o uso de memória. Através do algoritmo de Arnoldi, a ordem do problema de autovalor é reduzida e os autovalores de interesse são recuperados. Os resultados indicam o efeito estabilizante da compressibilidade nos modos dominantes da cavidade e revelam modos inerentes ao escoamento compressível, para os quais a compressibilidade tem efeito desestabilizante. Dentre estes modos compressíveis, estão presentes modos de propagação sonora em dutos e modos relacionados à geração de som na cavidade. / Hydrodynamic instability mechanisms play an important role in laminar to turbulent transition. Hydrodynamic instability analysis of a lid-driven cavity flow was performed by global mode decomposition (biglobal) to evaluate compressibility effects on this phenomenon. The basic flow was calculated by direct numerical simulation (DNS). A compressible DNS code was developed with spectral-like compact finite difference spatial discretization. The code allows parallel processing with a domain decomposition method that preserves the compact finite difference accuracy. The basic flow is used to form the eigenvalue problem associated to the linear Navier- Stokes equations for the perturbation, which were discretized by an explicit finite difference scheme. The combination of sparse matrix techniques and finite difference discretization leads to a significant memory reduction. The order of the eigenvalue problem was reduced using the Arnoldi algorithm and the eigenvalues of interest were calculated. Results show the stabilizing effect of compressibility on the leading modes and reveal some modes intrinsic to compressible flow, for which compressibility has a destabilizing effect. Among these compressible modes, there are some related to sound propagation in ducts and to sound generation inside the cavity.

Most materials used in many industries such as plastic, food, pharmaceuticals, electronics, dye, etc. exhibit viscoelastic properties under their processing or flow conditions. Due to the elasticity of such materials, deformation-stress in addition to their hydrodynamic behavior differ from simple Newtonian fluids in many important respects. Rod climbing, siphoning, secondary flows are all common examples to how a viscoelastic fluid can exhibit quite distinctive flow behavior than a Newtonian fluid would do under similar flow conditions. In industrial processes involving flow of viscoelastic materials, understanding complexities associated with the viscoelasticity can lead to both design and development of hydrodynamically efficient processes and to improved quality of the final products. In the present study, the main objective is to develop two dimensional finite volume based convergent numerical algorithm for the simulation of viscoelastic flows using nonlinear differential constitutive equations. The constitutive models adopted are Oldroyd-B, Phan-Thien Tanner (PTT) and White-Metzner models. The semi-implicit method for the pressure-linked equation (SIMPLE) and SIMPLE consistent (SIMPLEC) are used to solve the coupled continuity, momentum and constitutive equations. Extra stress terms in momentum equations are solved by decoupled strategy. The schemes to approximate the convection terms in the momentum equations adopted are first order upwind, hybrid, power-law second order central differences and finally third order quadratic upstream interpolation for convective kinematics QUICK schemes. Upwind and QUICK schemes are used in the constitutive equations for the stresses. Non-uniform collocated grid system is employed to discretize flow geometries. As test cases, three problems are considered: flow in entrance of planar channel, stick-slip and lid driven cavity flow. Detailed investigation of the flow field is carried out in terms of velocity and stress fields. It is found that range of convergence of numerical solutions is very sensitive to the type of rheological model, Reynolds number and polymer contribution of viscosity as well as mesh refinement. Use of White-Metzner constitutive differential model gives smooth, non oscillatory solutions to much higher Weissenberg number than Oldroyd-B and PTT models. Differences between the behavior of Newtonian and viscoelastic fluids for lid-driven cavity, such as the normal stress effects and secondary eddy formations, are highlighted. In addition to the viscoelastic flow simulations, steady incompressible Newtonian flow of lid-driven cavity flow at high Reynolds numbers is also solved by finite volume approach. Effect of the solution procedure of pressure correction equation cycles, which is called inner loop, on the solution is discussesed in detail and results are compared with the available data in literature.

Made available in DSpace on 2016-04-01T17:54:43Z (GMT). No. of bitstreams: 0 Previous issue date: 2007-08-31. Added 1 bitstream(s) on 2016-04-01T18:00:22Z : No. of bitstreams: 1 000505957.pdf: 1448129 bytes, checksum: ef6c3d75951b3a630e2efdba48bcb784 (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / O desenvolvimento de modelos numéricos para simulação de escoamentos tem se dado em diversos aspectos: desde os métodos de acoplamento entre pressões e velocidades até as técnicas de estabilização da solução. Neste contexto, este trabalho se propõe a implementar e avaliar o comportamento de um destes algoritmos, o CBS, utilizando-o na simulação de um escoamento isotérmico de fluido incompressível. Para tanto, foram utilizados variantes do esquema de solução com relação à discretização no tempo, com os métodos explícito e semi-implícito, além de algumas outras alterações na discretização no espaço e nos termos de estabilização. A discretização geral do problema foi feita com o método dos elementos finitos utilizando-se uma malha formada por elementos bilineares. O algoritmo e suas variações foram avaliados através de uma série de resultados, para diversos valores do número de Reynolds, de um problema clássico: a cavidade recirculante. Desta forma, foi possível mostrar que todas as variações da solução usando a malha estabelecida apresentaram concordância satisfatória com os resultados da literatura. Além disto, verificou-se que a omissão do termo característico não apresentou diferenças significativas nos resultados para os valores de número de Reynolds testados. Os resultados mostraram também que o método semi-implícito converge com um menor número de incrementos de tempo que o método explícito. / The development of numerical models for fluid flow simulation occurred at many aspects: since coupling methods between pressures and velocities until stabilization techniques. In this context, this work intends to implement and evaluate one algorithm behavior, the CBS, used in a simulation of incompressible and isothermal flows. This work also uses the explicit and semi-implicit methods of the time discretization scheme and some others changes in space discretization and stabilization terms for tests with algorithm. The discretization technique used is the finite element method with the mesh formed by bilinear elements. The algorithm's changes were evaluated using results for several Reynold's number in a classic problem: the lid-driven cavity. All the changes in solution algorithm with the used mesh show good agreement with other results from bibliography. It was verified too, that the omission of characteristics term didn't present significant differences in results with Reynold's number tested. It was observed yet that the solution code using the semi-implicit method converges with less time steps than the explicit one.

Fluid mixing is a challenging problem in laminar flow systems. Chaotic advection can play an important role in enhancing mixing in such flow. In this thesis, different approaches are used to enhance fluid mixing in two laminar flow systems. In the first system, chaos is generated in a flow between two closely spaced parallel circular plates by pulsed operation of fluid extraction and reinjection through singularities in the domain. A singularity through which fluid is injected (or extracted) is called a source (or a sink). In a bounded domain, one source and one sink with equal strength operate together as a source-sink pair to conserve the fluid volume. Fluid flow between two closely spaced parallel plates is modeled as Hele-Shaw flow with the depth averaged velocity proportional to the gradient of the pressure. So, with the depth-averaged velocity, the flow between the parallel plates can effectively be modeled as two-dimensional potential flow. This thesis discusses pulsed source-sink systems with two source-sink pairs operating alternately to generate zig-zag trajectories of fluid particles in the domain. For reinjection purpose, fluid extracted through a sink-type singularity can either be relocated to a source-type one, or the same sink-type singularity can be activated as a source to reinject it without relocation. Relocation of fluid can be accomplished using either "first out first in" or "last out first in" scheme. Both relocation methods add delay to the pulse time of the system. This thesis analyzes mixing in pulsed source-sink systems both with and without fluid relocation. It is shown that a pulsed source-sink system with "first out first in" scheme generates comparatively complex fluid flow than pulsed source-sink systems with "last out first in" scheme. It is also shown that a pulsed source-sink system without fluid relocation can generate complex fluid flow. In the second system, mixing and transport is analyzed in a two-dimensional Stokes flow system. Appropriate periodic motions of three rods or periodic points in a two-dimensional flow are determined using the Thurston-Nielsen Classification Theorem (TNCT), which also predicts a lower bound on the complexity generated in the fluid flow. This thesis extends the TNCT -based framework by demonstrating that, in a perturbed system with no lower order fixed points, almost invariant sets are natural objects on which to apply the TNCT. In addition, a method is presented to compute line stretching by tracking appropriate motion of finite size rods. This method accounts for the effect of the rod size in computing the complexity generated in the fluid flow. The last section verifies the existence of almost invariant sets in a two-dimensional flow at finite Reynolds number. The almost invariant set structures move with appropriate periodic motion validating the application of the TNCT to predict a lower bound on the complexity generated in the fluid flow. / Ph. D.

Esta dissertação investiga numericamente a influência dos efeitos viscosos e elásticos em escoamentos de materiais viscoplásticos no interior de uma cavidade dirigida. O modelo mecânico empregado é constituído pelas equações de conservação de massa e pelo princípio da quantidade de movimento linear, para fluidos incompressíveis, acoplado à equação constitutiva. Esta equação modifica o modelo viscoelástico de Oldroyd-B de modo a acomodar que os tempos de relaxação e retardo do material, bem como sua viscosidade viscoplástica, dependam das mudanças de sua microestrutura. A aproximação numérica do modelo emprega o método multi-campos de Galerkin mínimos-quadrados em termos do tensor de tensão extra, do vetor velocidade e do campo de pressão. Os resultados objetivam a determinção do tamanho e localização das regiões aparentemente não-escoadas do material, bem como sua deformação elástica, intensidade de tensão, e a sua vorticidade no interior da cavidade. Os resultados claramente indicam que o padrão do escoamento é fortemente influenciado pela variação dos efeitos elásticos (variação do tempo de relaxação adimensional, θ0 * ), viscosos (variação do índice de power-law, n) e cinemáticos (variação da velocidade adimensional, U* , do escoamento) no interior da cavidade. / This dissertation investigated numerically the influence of viscous and elastic effects on flows of viscoplastic materials within a lid-driven cavity. The mechanical model used is made up of mass and momentum balance equations, coupled with the constitutive equation. This equation modifies the viscoelastic Oldroyd-B model to accommodate both relaxation and retardation times, and viscosity function, dependent on the microstructure changes. Numerical approximations of the model make use a three-field Galerkin least squares method in terms of the extra stress tensor, velocity vector and pressure field. Computations focus on the determination of the size and position of apparently unyielded regions as well as the elastic deformation, stress intensity, and the vorticity within of the cavity. Results clearly indicate that the flow pattern is strongly influenced by the elastic (variation of the dimensionless relaxation time, θ0 * ), viscous (variation of the power-law index, n) and kinematic (variation of the dimensionless flow velocity, U* ) effects within the cavity.

A method to solve the Navier-Stokes equations for incompressible viscous flows is proposed. This method is SIMPLE (Semi-Implicit Method for Pressure Linked Equations) algorithm to iteratively solve the two-dimensional laminar steady momentum equations and based upon finite volume method on staggered grids. Numerical tests are performed on several cases of the flow in the lid-driven cavity, as well as of the flow after a backward-facing step with SIMPLE and SIMPLER (SIMPLE Revised) methods. Finally, results are compared qualitatively and quantitatively with numerical and experimental results available in the literature for different Reynolds numbers to validate the methods.