ELLIPTIC GRID GENERATION

Computational Fluid Dynamics (Assignment No. 03)

Submitted To: Associate Prof.

Zhonghua

Han
Prepared By:

Waseem Sakhawat
(MS Student)

Student ID:

2012420071

School of Aeronautics Northwestern Polytechnical University, Xian, China

Contents
Introduction................................................................................................................... 3 Mathematical Background............................................................................................. 3 Program Scheme........................................................................................................... 4 Implementation............................................................................................................. 6 Input Parameters:......................................................................................................... 6 Program Output............................................................................................................ 6 Conclusion..................................................................................................................... 7 References.................................................................................................................... 7

Introduction
In this report an "O" type grid is generated for a two-dimensional airfoil using elliptic grid generation techniques. The term boundary-fitted coordinates (BFC) is commonly used to describe such grids. These grids are used when the computational domain does not line up with commonly used Cartesian or polar grid. Also periodic boundary condition is used rather than the fixed or derivative boundary conditions.

Mathematical Background
Elliptic grid generation is one of several methods used to generate structured grids for odd geometries. Algebraic methods and hyperbolic systems of equations are sometimes used, particularly for external flows. In this work Laplace equations are solved to generate the mesh. Rather than work in the physical domain where the geometry is complicated and the equations are simple, we prefer to work in the computational domain where just the opposite is true. Once we have done our work in the computational domain, we bring the results back to the physical domain for viewing. When transformed from the physical (x, y) space to the computational ( , ) space, and the Laplace equations 2 = 0 and 2 = 0 become:

x 2 x + x = 0 y 2 y + y = 0
Here,

(1) (2) (3) (4)

= x2 + y2
Also

= x x + y y

= x2 + y 2

(5)

The boundaries in the physical region will be mapped to the computational plane to become boundary values for the two elliptic equations. The above equations will then be finite differenced in the computational space and solved iteratively to fill in the interior values. The results will then be plotted up for inspection in the physical domain.
3

Program Scheme
The accompanying sketches of the physical and computational regions should help you with the mapping of the boundaries and, in particular, the handling of the periodic boundary at the trailing edge.

Physical Domain

Computational Domain

The program draws one particular airfoil (e.g NACA 0012) in the physical plane. The outer extreme in the physical plane is a circle of diameter perhaps ten to fifteen times the chord. These values need to be assigned along the upper boundary in the computational plane. At the "cut" emanating from the trailing edge, y = 0.0. for all points and linearly distribute the x values between the trailing edge and the outer boundary. Once the boundaries are mapped to the computational plane the interior is filled using interpolation technique. Interpolation, gives a very good initial grid which then may be improved using the elliptic system. Besides the major time saving expected from using a good first "guess" with the iterative solution for the elliptic system, using interpolation first has another, very important side effect. The transformed equations are coupled and non-linear, and an iterative solution from just arbitrary initial conditions may converge to another, non-useable solution. In next step the equations introduced earlier are to be finite-differenced using standard centered differences and the resulting equations for x and y is to be solved using an iterative techniques. The iteration scheme in discretized form is given by

(6)
4

Gauss-Seidel iteration is easy to implement. The difference form of Equation 1 is solved easily for x (i, j) and that of Equation 2 is solved for y (i, j). Obviously the, and used here all have to be updated constantly as your iteration proceeds. Then successive overrelaxation method (SOR) (a method of solving a linear system of equations derived by extrapolating the Gauss-Seidel method) is applied. This extrapolation takes the form of a weighted average between the previous iterate and the computed Gauss-Seidel iterate successively for each component. xi ( k ) = x i
(k )

+ ( 1 ) xi ( k 1)

(7)

where x denotes a Gauss-Seidel iterate and is the extrapolation factor. The idea is to choose a value for that will accelerate the rate of convergence of the iterations to the solution.

For this assignment the values of x and y are considered to be fixed at j = 1, the airfoil surface and j = jmax, the outer boundary of the computing region. (There may actually be reasons to allow these points to be part of the solution.) A periodic boundary is to be implemented at the trailing edge, so that it acts essentially as part of the interior. A simple, frequently used method of implementing a periodic boundary is to add a fictitious extra column. The iteration runs from i = 2 to i = 37. At the end of a sweep the values of x and y at i =37 are transferred to i = 1 (and used in the calculation at i = 2) and those at i = 2 are transferred the other way to i = 38 and used in the calculation at i = 37. This simple swap obviates the necessity of setting up separate forms of the difference equations at the ends. Probably your best development strategy is to forget the periodic boundary to start out with and implement it only after you have your program working reliably. That is, dimension only 37 x whatever corresponding to the schematic shown and assume that values of x and y are fixed for all values of j at i=1 and i=37. With these boundary conditions
5

your grid should show a slope discontinuity at the cut and you will see why a periodic boundary is preferred.

Implementation
The above programming scheme explained is now implemented by generating an O-type grid for NACA 0012 airfoil by solving the above mentioned Laplace equation. The program is written by using MATLAB. An input file is created in which the desired points in the I and the J directions are specified. Also the outer boundary radius the tolerance for the program termination and total number of iterations are specified.

Input Parameters:
Ip Jp br tol itr = Points in the I direction = Points in the J direction = Boundary radius = tolerance = Total Number of Iterations

Program Output
Two cases are selected for demonstration of code. Case 1: Ip = 66, Jp = 24, br = 15, tol = 1e-6, itr = 1000

Grid Far View

Grid Zoom View

Case 2: Ip = 131, Jp = 51, br = 20, tol = 1e-6, itr = 1000

Grid Far View

Grid Zoom View

Conclusion
Overall, an elliptic grid was shown to provide desired results for discretization. It succeeded in smoothing out otherwise rough edges. The simplest elliptic partial differential system and one that does exhibit considerable smoothness is the Laplace system. It guarantees a one-to-one mapping for boundary-conforming curvilinear coordinate systems on general closed boundaries.

References
[1] [2] [3] Abbott, I.H, and von Doenhoff, A.E., Theory of Wing Sections, Dover, New York, 1959. Anderson, D.A., Tannehill, J.C., and Pletcher, R.H., Computational Fluid Dynamics and Heat Transfer, McGraw-Hill, New York (1984). Hoffman, K.A. and Chiang, S.T., Computational Fluid Dynamics for Engineers - Vols. I & II, Engineering Education System, Wichita (1993).

[4] Thompson, J.F., Warsi, Z.U.A. and Mastin, C.W., Numerical Grid Generation - Foundations and Applications, North-Holland, New York (1985).