Dr Ugur GUVEN Aerospace Engineer PhD Nuclear Science and Technology Engineer MSc Discretization of Equations The first step in solving any CFD problem is to discretize the equations ID: 582703
Download Presentation The PPT/PDF document "Finite Difference Methods" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Finite Difference Methods
Dr.
Ugur
GUVEN
Aerospace Engineer (
P.hD
)
Nuclear Science and Technology Engineer (
M.Sc
)Slide2
Discretization of Equations
The first step in solving any CFD problem is to discretize the equations.
Usually all fluid dynamics equations are in a partial differential equation format
These differential equations must be transformed into algebraic form, so that they can become solvable by the computer
The most common Discretization technique for Partial Differential Equations is the Finite Difference MethodsSlide3
Taylor Series
The primary background of any discretization using Finite Differences depends on using the Taylor series.
In the Taylor series, you can approximate a solution to any function at (x +
dx
) as long as you know the initial valueSlide4
Taylor SeriesSlide5
Forward Differences
Lets write the X component of velocity at Point (
i,j
) for a 2d flow.
Lets solve for the first term derivative of the above equationSlide6
Forward Differences
Hence after the first term, the remaining terms can be classified as truncation error and can be discarded (if you accept the magnitude of the error)
Then, we get a first order algebraic equation that approximates the partial derivative and it is called a forward differenceSlide7
Rearward Differences
Lets now write the Taylor series for the point u(i-1,j) in a 2d flowSlide8
Rearward Differences
Thus, if we solve the above equations to get a partial derivative of u to x, then we will have a first order accurate algebraic form of the partial derivative in the rearward difference form.Slide9
Central Differences
However, the problem with both the rearward differences as well as the forward differences is the fact that they are first order algebraic representations of a partial derivative.
Hence, the accuracy is greatly decreased due to this and we will need to look for ways to increase the accuracy of the partial derivative to reduce the overall error that will be formed in the equationsSlide10
Central Differences
Lets subtract the following equations to try to increase the order of accuracySlide11
Central Differences
Hence, solving the equation would cause the following equation to be formed:
Hence, the following is a second order accurate central difference representation:Slide12
Central Differences
Hence, the following equation is a more accurate representation of a partial derivative of u over x and thus it has less errors. Moreover, it takes data from both sides of the grid over point (
I,j
)Slide13
Differences with Respect to Y
Using the methodology defined above, it will be possible to create forward, rearward or central differences with respect to y in partial differentiationSlide14
Partial Derivatives Using Finite DifferencesSlide15
Writing Partial Derivatives in Algebra
Hence, using the methods of finite differences, you can easily transform first degree partial derivatives so that you can create an algebraic equation.
For example, transform the following partial differential equation using finite differencesSlide16
Second Order Partial Derivatives
To find the second order partial derivatives with respect to x, lets add the Taylor series expansion for u(i+1,j) and u(i-1,j)Slide17
Second Order Partial Derivatives
Summation of the above equations gives:
As a result, the second order partial derivative is written in central differences as:Slide18
Second Order Partial Derivatives
Hence, the second order partial derivatives written in the central difference notation would be:Slide19
Mixed Partial Derivatives
By writing a mixed partial derivative , we will be able to write second order equations as well.Slide20
Second Order Partial DerivativesSlide21
Transformation of Partial Differential Equations
Transform the following partial differential equation into an algebraic equation by using Finite DifferencesSlide22
HomeworkSlide23
Engineering Example 1
Calculate the shear stress and the heat transfer at the wall with the following data by using finite differencesSlide24
Example 1
The shear stress and the heat transfer at the wall is given by the following equations in first dimensional problems.Slide25
Example 1
First Order Difference Solution, solve for the other pointsSlide26
Time Marching Solution
In the solution of fluid dynamics equations, it is usually customary to solve flows changing over time. At each âtâ step of the flow, the flow properties will change accordingly. This is called Time Marching Solution. Slide27
Example of Time Marching Solution
Lets assume the unsteady, one dimensional heat conduction equation with constant thermal diffusivitySlide28
Example of Time Marching SolutionSlide29
Example of Time Marching Solution
Now, we are going to write the equation only for T, since we are interested in finding the temperature at different time points for x distance. Hence, the solution for Temperature is solved by Time Marching SolutionSlide30
Time Marching Solution
Time marching means that T at all grid points at time level n+1 are calculated from known values at time level n. Then n+2 is calculated pretty much the same way as n+1 levels are used for calculation. Slide31
Explicit and Implicit Approach
If a difference equation contains only one unknown and all other variables are know, then that is called an explicit solution.
If more then one unknown variable exists, then you will need to solve a set of algebraic equations simultaneously. Hence, this is called the implicit approach.Slide32
THANK YOU
Please download this lecture as well as other lectures and numerical examples from my website:
www.cfdlectures.co.cc
Dr.
Ugur
GUVEN
drguven@live.com