Chapter Two Literature Survey

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 12

Chapter Two Literature Survey

Literature survey
In this chapter general review which deal with numerical method of flow over a body are presented, the review is classified into: 1- Grid generation and body fitted coordinates. 2- Numerical method in solving flow fields problems for compressible viscous and non-viscous fluid flow. 3- Boundary condition. 4- Turbulence modeling. 5- Artificial dissipation. 6-The scope of the present work.

2.1

Grid generation and body fitted coordinates


Any flow problem gives PDE. To solve these equations for

simple uniform boundary, a numerical procedure is needed with regular rectangular grids . Since engineers very rarely come across such boundaries in reality, it is then necessary to find out a procedure which can map the irregular physical reality into the rectangular regions that can easily be used to realize the computational scheme. Generating a regular grid over an irregular computation region makes the specification of boundary conditions easier, which influences the accuracy of the solution significantly regardless of the approximation accuracy inside the computation region. The use of curvilinear coordinate system with coordinate lines coincident with the boundary of physical domain is called body fitted coordinate, which is a coordinate system that coincides with body

Chapter Two Literature Survey


geometry. The types of coordinates are classified as, orthogonal and nonorthogonal coordinates. Orthogonal coordinates are used to obtain high accuracy, but certain terms form the governing equation are dropped [3] so the equation becomes simpler. Conformal mapping, solution of elliptic, hyperbolic or parabolic partial differential equation, and algebraic mesh generation, can generate this type. In the non-orthogonal coordinates, the concentration of grid lines in regions of large gradient can be controlled. It also has capabilities to simplify finite difference equation by transforming the three dimensional geometry in physical space to rectangular shape in the transformed computational space [4]. However both convergence and stability may be reduced due to the rising of additional terms such as crosse derivative, thereby increasing the computational time. In the conformal mapping method many applications in physical problems can be treated but these are restricted to two dimensional [4]. Therefore, our interest will be focused on the grid generation by solving differential equations of orthogonal boundary fitted coordinate systems. In the first type elliptic schemes require all the boundaries of the computational line or plane. The interior grid points between theses boundaries are determined iteratively by solving a set of elliptic partial differential equations. This method can be extended to three dimensional problems. The procedure was popularized by Thompson et al [5]. The spacing of the grid lines in the physical domain can be easily controlled by using the elliptic generation type. One method of variation is to add an homogeneous terms to the right sides of Laplace equation, so the generating system becomes.

xx + yy = P ( , )
xx + yy = Q( , )

(2-1) (2-2)

Chapter Two Literature Survey


P and Q can be chosen an exponential function [6]. Condunov and Prokopov[7] proposed a method of controlling the clustering grid points, by taken P and Q as follows.

P = a( x . y y . x ) 2 Q = b( x . y y . x ) 2

(2-3)

Where a and b are constants. The results of generation potential flow about different airfoils showed excellent comparison with the analytic solutions. Barfield et al [8] achieved control of the interior coordinate line spacing by solving the following system of elliptic PDEs.

X + aX = 0 X + bX = 0
constants a and b.

(2-4)

Then the interior coordinate line spacing is changed by varying the Many researchers proposed a method for controlling grid using different grid clustering[9][10][11]. Hyperbolic schemes required only the inner and outer boundaries for closed domain. Marching outward from the inner boundary until the outer boundary is reached generates the grid. The procedure is non iterative but provides no control over the position and shape of outer boundary. Steger and Chausse [12] presented a hyperbolic grid generation scheme for airfoil. The parabolic grid generation which combines the attributes of both elliptic and hyperbolic grid generation, was derived by Nakamura [13]. It was found that this procedure is suitable for coordinate line spacing and orientation and has been applied successfully to the simple wing-body configuration.

Chapter Two Literature Survey


The grid generation by solution of partial differential equations can be orthogonal and non orthogonal body fitted coordinate system. In the orthogonal coordinate system the transformation of partial differential equation is simpler than the other one. Also the cross derivative terms (non-orthogonal terms) dont arise through this transformation [4]. Also this method has another advantage like fewer computing operation, faster convergence and the stability and accuracy of the solution. The disadvantage of this method is the complexity of controlling the locations of grid point along the boundary. For three-dimensional it is impossible to generate such coordinate. In non-orthogonal coordinates, the advantage is the control ability of grid lines in the regions of large gradients. In the three dimensional geometrys it can simplify finite difference expression to a rectangular shape in the computational space . However, non-orthogonal coordinates introduce additional terms as derivatives which may reduce both convergence rate and stability hence it increases the computing time. Haussling [14] solving elliptic differential equation presented a method to generate orthogonal and nearly orthogonal boundary fitted coordinate system using Poisson equation to generate a curvilinear grid. Papantois and Athanassiadis [15] generated a curvilinear orthogonal grid by solving a set of Laplacs equations in two dimensions. Thames et al.[16] proposed a general procedure to generate a curvilinear coordinate systems for two dimensions by solving a set of Laplac equations, the space of grid line controlled by varying the elliptic generation system for the coordinates .

Chapter Two Literature Survey

2.2

Numerical Method
Number of investigators proposed different numerical approaches

for the prediction of viscous and inviscid flow over different configuration hence computation of fluid dynamics schemes (CFD) contains a wide range of applications. However the survey here is limited to the compressible supersonic viscous and inviscid flows. The use of numerical methods in solving the flow problems started early 1950. It was first used by Courant et al. [2]. In recent years there is a considerable number of solutions to some equations of fluid flow in regions with curved or complex shaped boundaries. Contributions have come from different branches of physics and engineering. Several method have been presented to treat irregular boundaries fluid flow problems [1726]. Lax and Wendroff [27], [28] and [29] solved time Euler and compressible Navier-Stocks equations . The second-order Lax-Wendroff method has led to a whole family of variants when applied to non-linear systems, the most propular of these variants is due to MacCormack [30]. In the present work MacCormack explicit technique is used . Courant et al [2] has introduced for the first time the explicit up-wind scheme . Later several extensions to second order accuracy and implicit time integration have been developed . D.Migdal, et al. [31] found that where conservative form of the Euler equations required fine mesh size, large computer storage and long computational time are also needed. J.D, Anderson Jr., [32] used explicit finite difference MacCormak predictor-Corrector method for unsteady quasi-one-dimensional equations in conservative form . R.A Serra [33] developed a time-dependant numerical method for the solution of inviscid mixed subsonic-supersonic flows with shocks.

Chapter Two Literature Survey


The analysis was based on lax-Wendroff explicit procedure by using special treatments of boundary conditions. Results obtained by finite difference technique and gave good results when compared with experimental data. R.A Delaney and P.Kavanagh [34] used a time dependent method of characteristics for transonic flow analysis in axial turbomachinary cascades. The analysis was based on the unsteady, two dimensional, compressible, inviscid flow equations. The method has an advantage of rigorous treatment of differential domain of dependence of a solution point and accurate numerical treatment at boundaries of the flow fluid. Result showed good agreement with experimental data. G-Morertti [35] solves Euler equations for compressible flow by integration scheme. Steger [36] and Van leer [37] introduced the flux vector splitting methods their study can be considered as a members of the same subgroup of schemes in the line of continuous method which solves over each mesh interval . Van leer [37] solved the Euler equation with the same method which called a flux difference splitting methods, all these methods give a smooth and accurate results when solving Euler equation. Without strong gradients but they can have different behaviors in the presence of shock waves, therefore a particular attention will be taken when the flow is with discontinuity. Various methods have be attempted to control these oscillations by the use of artificial viscosity. Another approach to prevent numerical oscillations without using artificial viscosity, was first introduced by Boris and Book [38]. The method is called total variation diminishing scheme. In this scheme the variation of the numerical solution is controlled by a non-linear way such as to forbid the appearance of any new extreme. This concept, when

10

Chapter Two Literature Survey


applied to central schemes, leads to particular forms of artificial viscosity without empirical constants. R.J Stiles [39] described a method of solving the equation governing unsteady two-dimensional inviscid flow. Interior points were computed using MacCormack method, while all boundary points were computed by a reference plane characteristic method. Analytic transformation of the physical space (x,y,t) to the rectangular computational space ( , , t ) was made to perform differencing procedure . The same procedure was used by Clinc [40], the interior points were computed using MacCormak predicator-corrector method . Since 1950[2] numerous finite-difference algorithms have been developed for solving the compressible viscous flow. Most of these research use MacCormaks technique to solve the system equations [41]. As mentioned above there are two schemes for solving Euler and Navier-stokes equation, classified as explicit and implicit method. In the explicit method all the new properties are evaluated from the known old properties. Solving Euler equation and Navior-stokes equation, by explicit MacCormack predictor-corrector technique, was first introduced in [30] this method is a second-order accuracy in both space and time . There are other explicit methods like time marching finite volume method and others[42], the MacCormak method is the best one for the following reasons. 1- It is of second order accuracy. 2- The ability to extended to two and three dimensional problems. The implicit finite difference approach the equation are backward difference in time and the no linear terms at the new time are linearized

11

Chapter Two Literature Survey


about their values of previous time level. Many researchers solve numerically the viscous and inviscid flow problems, using conservative set equation, for external and internal flow like MacCormak [41]. The advantage of this method is in saving computer time and storage requirements[43].

2.3

Boundary conditions
In order to obtain a unique solution of a partial differential equation

a set of supplementing conditions must be provided to determines the arbitrary function. The supplementary conditions are classified as boundary and initial conditions. Initial conditions are required for which the dependent variable is specified at some initial state. Boundary conditions are required for which the dependent or its derivative must be satisfied on the boundaries of the domain. The types of boundary condition are [44]. 1234The Dirichlet boundary conditions if the dependent variable along the boundary is prescribed. The Neumann boundary conditions if the normal gradient of the dependent variable along the boundary is specified. The Robain boundary condition, if the imposed boundary condition is a linear combination of the Dirichlet and Neumann types The mixed boundary condition if a certain portion of domain is the DirIchlet B.C. and another portion is a Neumann B.C.This is known as a mixed B.C. The B.C. treatment are perhaps the most important factor for accuracy of the computation flow results. In general the boundary condition can be classified according to the type of the flow, viscous on

12

Chapter Two Literature Survey


inviscid, laminar or turbulent, external or internal and subsonic or supersonic . or according to the type of governing equation elliptic or parabolic or hyperbolic or on procedure of solution like explicit, implicit or other classification .

2-4 Turbulence Modeling


The mechanism of turbulence is a very complicated subject. The mean velocities and turbulent viscosity are the main variables in solving any turbulent flow. But there is no rational theory to determine them, so many researcher try to find and create mathematical basis for investigation turbulence motion and with the aid of empirical hypothesis to calculate the property of turbulent flow, these are the turbulence models. Not only the exchange of momentum responsible for the mixing motion , but the heat and mass transfer also. Most of models which are used to predict supersonic and hypersonic flows, are classified as [43][44]. 1-Algebraic stress model. 2-Zero equation model. 3-Two equation model K. model. 4-Reynolds stress equation model. The first type is the algebraic stress equation model, which is an economical method of incorporating the effects of anisotropy into the calculations of Reynolds stresses. The model is not as well validated as the K. model but can be used in flows where the latter is known to perform poorly and where the transport assumptions do not compromise too severely the calculation accuracy[43] .

13

Chapter Two Literature Survey


The second type zero equation model which is one of early turbulent models also called mixing length model. In this model most of the kinetic energy of turbulence is contained in the largest eddies and the turbulence length scale l is therefore characteristics of these eddies, which interacts with the mean flow. This algebraic model is attractive since it needs storage space and less computational time and its widely used in solving Navier stokes algorithms. Baldwin and Lomax [45][43] and Cebeci and smith [46], both of them developed Prandtle mixing length models. These two models are the most widely used turbulence models in external aerodynamics calculations in the aerospace industry. David Degani[47] investigated the supersonic boundary layer over a concave surface using Baldwin Lomax and Cebeci and Smith models in solving parabolized Navier Stokes equations and he obtained a good results in calculating the turbulence viscosity. Fayadh[48] use the Baldwin Lomax in supersonic flow a good results obtained in solving external supersonic flow . The third type (two-equation model) like the k. model. Which focuses on the mechanisms that effect the turbulent kinetic energy. This model was used and predicted better approximations for the flow field within pipes and ducts, Leschziner and Rodi[49] used k. model with strong curvature duct, Rodi [50] used this model with strong adverse pressure gradient. The advantage of this model is that it needs only initial or boundary conditions and is the most widely validated turbulence model. But its more expensive to implement than mixing length model because it need two extra PDEs and its failure in some cases . The last type is the Reynolds stress equation model, which is the most complex classical turbulence model also called the second-order model. It gives good and accurate calculation for many complex flows such as wall jets, and non circular duct but it requires high computing

14

Chapter Two Literature Survey


costs and not widely validated as the k- model or the mixing length models. This model originates from work reported by Launder [51] [43]. In the present work Baldwin Lomax (zero-equation model was chosen because of its widespread use in compressible flow calculations especially when the boundary conditions Chang [48].

2-5 Artificial Dissipation


The solution of the Euler equation using numerical techniques requires the use of either a differencing method with inherent dissipation or the addition of dissipation terms to a non dissipative scheme. This is because the Euler equations do not provide any natural dissipation mechanism (such as viscosity in the Navier-Stokes equations)that would eliminate high frequencies which are caused by non-linearities and especially shocks. A variety of numerical method to solve Euler equation use basically central difference approximation to the spatial derivatives and the addition of some form of artificial dissipation. The work of Thomas H.[52] , Roe [53] , Van Leer [54] , Osher and Chakravarty [55], and Hartens [56] all use the numerical dissipation. There are two main reason for the addition of artificial dissipation . 1-To control the odd-even uncoupling of grid points typical of central differencing. 2-To control strong nonlinear effects such as shocks. The dissipation models are also applied in viscous computations where it effects the accuracy and stability of fluid dynamic results . In the present work artificial dissipation was used to control the stability and accuracy of the results.

15

Chapter Two Literature Survey

2-6 Scope of the present work


An important conclusion can be drawn from the previous literature review, it is clear that their is a lack of information about solving supersonic viscous an inviscid flow. In the present work the method use allows the use of arbitrary coordinate system, and it involves an economical mesh refinement technique , which preserves second order accuracy and offers a great flexibility of use. During the review of different numerical technique in solving CFD problem, MacCormak method is the easiest in understanding and programming. Therefore in the present work Explicit MacCormak predictor corrector technique was used due to. 1- Its perfectly satisfactory for many CFD problem. 2- Second order accuracy. 3- Minimum storage capacity. 4- Boundary condition can be easily to specification. 5- Fully conservative (solution of PDE,s in conservative form). During the present work, we shall take the simplified case, where the nozzle exit flow is assumed uniform stream, due to the very small size of the vane. Hence the nozzle exit boundaries was considered a far-field B.C. Explicit time marching MacCormak predictor-corrector technique (due to the above reason) was used to solve the flow over the vane, with body fitted coordinate to deal with complex geometry. Elliptic and hyperbolic grid generation were used. The vane shall be rotated from 0 to 35 degree, to show the effect of the angle of attack on the fluid flow properties. For each degree of rotation a re-mesh technique was used to generated the new mesh and this is down by the computer. Fortran language was used in building the computer program to solve the present problem.

16

You might also like