http://www2.icfd.co.jp
Computational Fluid Dynamics
• http://fmad-www.larc.nasa.gov/aamb/
• Fluid Mechanics and Acoustics Division
• NASA Langley Research Center in Hampton, VA.
Computational Fluid Dynamics: What is it?
• The physical aspects of any fluid flow are governed by the following
fundamental principles:
(a) mass is conserved
(b) F = ma (Newton‘s second law)
(c) energy is conserved.
• CFD is the art of replacing the governing equations of fluid flow with numbers,
and advancing these numbers in space and/or time to obtain a final numerical
description of the complete flow field of interest.
• The high-speed digital computer has allowed the practical growth of CFD.
The Role of CFD in Modern Fluid Dynamics
Pure Pure
experiment theory
Computational
Fluid
dynamics
Fig. The “three dimensions“ of
fluid dynamics
• To whom:
The completely uninitiated student
• To provide what:
(a) an understanding of the governing equations
(b) some insight into the philosophy of CFD
(c) a familiarity with various popular solution techniques;
(d) a working vocabulary in the discipline
• At the conclusion of this course:
will be well prepared to understand the literature in this field,
to follow more advanced lecture series,
and to begin the application of CFD to your special areas of
concern.
I hope...
1.2 The Basic Equations of Fluid Dynamics
Fluid and Flow
Gas, air, fuel, oxigen, CO2
Fluid: Liquid water, oil, liquid metal that flows
(Solid) (geophysical flows)
Eulerian and Lagrangian framework
Eulerian description of the flow
the velocity of the fluid u at the position x at time t
u (x , t ) u (x , t )
Lagrangian labels
the state and motion of the point particles
a = (a1,a2,a3) Lagrangian labels
u (x , t ) = u ( X (a, t ) , t ) = U (a, t ) a
b U (a, 0 )
U (a, t )
Mathematical Model
Finite Control Volume
A closed volume in a finite region of the flow: a controle volume V;
a controle surface S, closed surface which bounds the volume
Control surface S S
Control volume V V
volume dV
The Material Derivative
The velocity of a point particle = the rate of change of its position X
(Integral form)
Second Order
Linear PDEs
Navier Stokes
Equations
b2 – 4ac = 0 parabolic
b2 – 4ac < 0 elliptic
2.1.2
General Behaviour of the Different Classes of
PDEs
• Hyperbolic Equations
• Characteristic curves
Region II
Domain of influence Left-running characteristic
y
a P Region I
Influenced by point p
-Region of influence-
Initial data upon which p
depends
Right-running characteristic
c
Region influenced by
x
point c
Hyperbolic Equations
(x,y,z)
Characteristic surface
y
P Volume
Initial data in the yz Volume which influences influenced by point p
plane upon which p point p
depends
z
Parabolic Equations
• Only one characteristic direction,
• Marching-type solutions
y y
x=0 x=t
Elliptic Equations
• No limited regions of influence;
• information is propagated everywhere in all directions (at once).
y
b c
a b x
• Boundary conditions
• A specification of the dependent variables along the boundary. Dirichlet condition
• A specification of derivatives of the dependent variables along the boundary. Neumann
condition
• A mix of both Dirichlet and Neumann conditions.
2.2 The Dynamic Levels of Approximation
2.2.1 Inviscid Flow Model: Euler Equations
Steady inviscid supersonic flows
Wave equation
2.2.2 Parabolized Navier-Stokes Equations, Boundary
Layer Approximation
Unsteady thermal conduction
Boundary-layer flows
Basic derivations
Discretization errors
Types of solutions:
Time integration Iterative methods
Explicit and implicit
Initial value problem Boundary value problem
I.V.&B.V. problem
Stability analysis
2 Finite Difference Method
2.1 Basic Concept
> to discretize the geometric domain > to define a grid
> a set of indices (i,j) in 2D, (i,j,k) in 3D
> grid node values
The definition of a derivative
2.2 Approximation of the first derivative
2.2.1 Taylor series expansion
Expansion at xi+1
Expansion at xi-1
Truncation errors
Third order BDS, third order FDS and fourth order CDS
2.3 Approximation of the second derivative
Approximation from xi+1, xi
second derivative
2.4 Approximation of mixed derivatives
Discretization error
Richardson extrapolation
for sufficiently small h
• 4.1 Introduction
4.2 Approximation of surface integrals
4.3 Approximation of volume integrals
4.4 Interpolation practices
Linear interpolation
Quadratic Upwind Interpolation (QUICK)
Higher-order schemes
5 Spectral Methods
5.1 Basic concept
A discrete Fourier series
t Solution at
time t=t
time t=0
t=0 Initial condition Steady solution
Boundary condition
Two-level methods
Two-level methods
Predictor-Corrector method
Mutipoint methods
Adams-Bushforth methods
Runge-Kutta methods
The second order Runge-Kutta method
approximate
PDE FDE
Æ0
Solution of PDE Solution of FDE
?
Ans.: „Even if we solve the FDE that approximate the PDE appropriately, the solution
may not always be the correct approximation the exact solution of PDE.
8.1 Consistency, stability and convergence
Consistency
Æ0
PDE FDE
Stability
Convergence
Lax‘s equivalence theorem
consistent
PDE FDE
stable
Æ0
Solution of PDE Solution of FDE
convergent
L : linear operator
Truncation error
8.2 Von Neumann‘s method
Fourier representation of the error on the grid points
FTCS method for 1D convection equation
Fourier series of the error
Amplification factor G
A= A21/A11
Forward elimination
upper triangular matrix
Back substitution
Solution of
Ax = Q (0)
Factorization into lower (L) and upper (U) triangular matrices
A = LU (1)
Into two stages:
Ux=y (2)
Ly=Q (3)
9.0.3 Tridiagonal matrix
Thomas algorithm / Tridiagonal Matrix Algorithm (TDMA)
More generally,
PA = M – N, B = PQ (7)
P : pre-conditioning matrix
An alternative to (5): -M u n
M (u n+1 – u n ) = B – (M –N) u n (8)
or
Mdn =rn
d n = u n+1 – u n : correction
9.1 Jacobi, Gauß-Seidel, SOR method
Poisson equation
u=f
(u i+1,j –2 u i,j + u i-1,j)+(u i,j+1 –2 u i,j + u i,j-1)=f i,j h2
M = LU = A + N
Minimization problem
A: positive definite
Steepest descents
Conjugate gradient method
condition number of A
preconditioning
C-1AC-1Cp=C-1Q
9.3 Multi Grid method
Spectral view of errors
Fourier modes
Restriction
Interpolation
Coarse grid scheme
V-cycle, W-cycle & FMV scheme
9.4 Non-linear equations and their solution
9.4.1 Newton‘s method
linearization
new estimate
y
y = f(x)
x0
o
x1 x
Newton‘s method
System of non-linear equations
linearization
Newton‘s method
9.5 Examples
Transonic flow over an airfoil
11 Incompressible Navier Stokes (NS) equations
11.1.0 Incompressible fluid
Reynolds number
ρul ul
Re = =
η ν
u : velocity scale, l : length scale, v : kinematic viscosity
ex.
U = 10 m/s, L = 1 m, v = 0.15 St, Re ~ 105.
U = 0.1 m/s, L = 100 m, Re ~ 105.
11.2 The pressure Poisson equation method
Governing equations
Navier-Stokes equations
For convenience,
i.e., ~u‘ is neglected. By assuming that the final pressure correction is app‘ :
In the absence of any contribution from source terms, if a steady solution is sought,
ap = 1 - av
which has been found nearly optimum and yields almost the same convergence rate as
SIMPLEC method.
11.5 Other methods
11.5.1 Streamfunction-vorticity methods
Stream function
Kinematic equation
should be small.
12 Some remarks on incompressible flows