Sunteți pe pagina 1din 52

CFD for Aerospace Engineers

Stefan Hickel & Axelle Vir

This course
is not a course on conservation laws and continuum mechanics
Viscous Flows,
cannot treat turbulence and transition phenomena in depth
Turbulence A,
does not give detailed instructions on the analysis or
development of numerical methods
CFD 1,
is also not a Photoshop course
but rather:
is an application oriented, comprehensive introduction to the
state-of-the-art in CFD. We discuss the essential building blocks
of CFD in the lecture and practice the CFD process in tutorials.

Stefan Hickel CFD for Aerospace Engineers

Content of the lecture


Introduction to Computational Fluid Dynamics
Basic equations and dimensionless numbers
Discretization of the fluid domain and grid generation
Turbulence and turbulence modeling
- RANS
- LES
Numerical solution of the Navier-Stokes equations
- Finite-volume method
- Grid and boundary conditions
- Explicit and implicit discretization methods
- Errors of numerical and physical models
Visualization of computational results
Critical evaluation of results and approximation errors

Stefan Hickel CFD for Aerospace Engineers

Content of the tutorials


Practical exercises with ANSYS CFX and ICEM
CFD process chain with ICEM and CFX
Krmn vortex street
(un-/structured, in-/compressible, un-/steady)
Transonic flow over an airfoil

Independent practicing and experimentation


ICEM tutorials
CFX tutorials

Stefan Hickel CFD for Aerospace Engineers

Who, where and when


Its complicated
Instructors:

Stefan Hickel S.Hickel@tudelft.nl


Axelle Vir
A.C.Vire@tudelft.nl

Lecture:

every Monday at 15:45 in CT hall G


additionally:
1st week: on Wednesday at 13:45 in LR hall A
7th week: on Wednesday at 10:45 in LR hall B

Tutorials:

2nd, 3rd and 4th week: Friday at 15:45 in LR hall 007


5th week: Thursday at 15:45 in LR hall 007
6th week: Wednesday at 10:45 in LR hall 007

Stefan Hickel CFD for Aerospace Engineers

Assessment
The final grade consists of two parts:
- Your knowledge about the lecture content will be evaluated
in an exam (50% of the grade).
- Your ability to set-up, run and critically analyze a numerical
simulation will be evaluated through a take-home
assignment (report 10 pages, 50% of the grade).

Stefan Hickel CFD for Aerospace Engineers

Introduction to Computational
Fluid Dynamics

Short History of Computational Fluid Dynamics


Richardson 1922: Weather Prediction by Numerical Process

First numerical weather prediction


Discretized Europe as chess board
with line spacing of about 200 km
The simulation was done by hand
and diverged
Recognized turbulence as reason
for failure
Simulation of entire world should be
possible with 64000 computers
Stefan Hickel CFD for Aerospace Engineers

Short History of Computational Fluid Dynamics


derivation on the blackboard:

Stefan Hickel CFD for Aerospace Engineers

Short History of Computational Fluid Dynamics


Prandtl 1925: oldest turbulence model still in use today
Rosenhead 1931: first successful simulation surface instability
Simulation was done by
hand
Rollup process visible
only 4 time steps

Stefan Hickel CFD for Aerospace Engineers

10

Short History of Computational Fluid Dynamics


Apelt 1951, und Payne 1958:
Finite difference simulation of cylinder flow

Stefan Hickel CFD for Aerospace Engineers

11

Short History of Computational Fluid Dynamics


Smagorinsky 1963:
first large-eddy simulation (LES) with eddy viscosity model
stabilization of weather simulations
Harlow & Welch 1965:
free surface flows Marker-and-Cell method
Jones & Launder 1972:
first two-equation turbulence model for Reynolds averaged
Navier-Stokes (RANS) simulations k- model
Peskin 1972:
Immersed Boundary Method
non-continuous problems on Cartesian grids

Stefan Hickel CFD for Aerospace Engineers

12

Short History of Computational Fluid Dynamics


Caretto 1972: SIMPLE algorithm
(SIMPLE := Semi-Implicit Method for Pressure-Linked Equations)
iterative computation of pressure and velocity field
Chorin 1973: projection method for incompressible flows
pressure Poisson equation is solved for correcting preliminary
velocity field
Cebeci & Smith 1974: RANS turbulence model
Baldwin & Lomax 1978: RANS turbulence model

Stefan Hickel CFD for Aerospace Engineers

13

Short History of Computational Fluid Dynamics


Leonard 1979: QUICK (Quadratic Upwind Interpolation) scheme
Patankar 1980: SIMPLER (SIMPLE-Revised)
Rhie & Chow 1983: Rhie-Chow interpolation
Van Doormal & Raithby 1984: SIMPLEC (SIMPLE-Consitent)
Wilcox 1988: important two-equation RANS turbulence model
k- model

Stefan Hickel CFD for Aerospace Engineers

14

Short History of Computational Fluid Dynamics


Germano 1990: dynamic eddy viscosity model for LES
Menter 1994: RANS turbulence model SST model
Spalart & Allmaras 1994: successful one-equation model for
aerodynamic flows SA model
Stolz & Adams 1999: LES turbulence model ADM
Wallin & Johansson 2005: explicit algebraic Reynolds stress
model for RANS EARSM

Stefan Hickel CFD for Aerospace Engineers

15

ANSYS-CFX
70th:UK Atomic Energy Agency develops software for flow
simulation and safety analysis
80th:the AEA-Technology puts FLOW3D on the market
1995renamed CFX-4
1998AEA-Technology merges with ASC - Advanced Scientific
Computing, the developer of TASCflow
1999release of CFX-5
2002merger with ANSYS
2006 merger of ANSYS and Fluent
2015 ANSYS CFD 16.0 - and there is still CFX and Fluent...
Stefan Hickel CFD for Aerospace Engineers

16

Why computational fluid dynamics simulations?


Objective:

Study details of a particular fluid dynamics


phenomena in a controlled environment

Alternative:

Simulate fluid flow experimentally in a wind or water


tunnel at similar conditions
Example aircraft: for simplicity consider only
two similarity conditions for exterior aerodynamics
Reynolds number

Re = 0U 0 L 0

Mach number

Ma = U 0 c0 = U 0

Stefan Hickel CFD for Aerospace Engineers

RT0

17

Why computational fluid dynamics simulations?


Realistic experimental fluid dynamics simulations can easily
become complicated, very expensive, and/or very dangerous...

National Full-scale Aerodynamics Complex (NFAC), 80x120 ft wind tunnel, NASA Ames
Stefan Hickel CFD for Aerospace Engineers

Pictures: http://www.nasa.gov/centers/ames/news/features/2012/nfac-25_prt.htm

18

Why computational fluid dynamics simulations?

Computational Fluid Dynamics simulations


provide unequaled insights into the
dynamics of fluids.
all field quantities are everywhere simultaneously available
and can be accessed directly without measurement errors
this leads to a better understanding of fluid flow phenomena
cheap (if you happen to already have a large computer)
hazard free

Stefan Hickel CFD for Aerospace Engineers

19

Simulation a complete gas turbine

Medic, You & Kalitzin(2006)


Stefan Hickel CFD for Aerospace Engineers

20

Zoom into a turbulent boundary layer

LES of pressure induced separation of a flat-plate boundary layer

Visualization of coherent structures with Q-criterion,


separation bubble with iso-surface u=0.
Hickel & Adams (2007)
Stefan Hickel CFD for Aerospace Engineers

21

Rayleigh jet break-up

Albina, Muzaferija & Peric (2000)


Stefan Hickel CFD for Aerospace Engineers

22

Cavitation at a NACA hydrofoil

Iso-surfaces 5%

pin = 600 bar

pmax 250 bar


Sheet and cloud cavitation on suction side
Experiment(*)

--

Simulation(**)

Collapse induced shocks at trailing edge


NACA 0015, AoA=6, chord=0.13 m, span=0.3
m, water, uin=30 m/s, pexit=4.5 bar, ref=1.0
Inviscid Flow Simulation (Euler-Equations)
24 million cells, tCFD = 8.5*10-8 s
5 periods ~ 0.05 s ~ 0.5 mio time steps ~ 4 weeks
on 192 cores

Conditions do not match only


qualitative comparison possible!
Stefan Hickel CFD for Aerospace Engineers

(*) Kawanami, Kato & Yamaguchi (1998)


(**) Schmidt & Thalhamer (2011)
23

Cavitation in micro nozzles


low p

(pexit=115 bar)

high p

(pexit=55 bar)

Egerer, Hickel, Schmidt & Adams (2014)


Stefan Hickel CFD for Aerospace Engineers

24

Flow through a fuel injector


Turbulence
coherent vortex structures

Phase Interfaces
fuel, air, and vapor volume

rley, Hickel, Schmidt & Adams (2016)


Stefan Hickel CFD for Aerospace Engineers

25

Supersonic jet

Stefan Hickel CFD for Aerospace Engineers

26

Supersonic jet (over expanded) - CFX

Stefan Hickel CFD for Aerospace Engineers

27

Supersonic jet (under expanded) - CFX

Stefan Hickel CFD for Aerospace Engineers

28

URANS of a pseudo shock system

Giglmaier et al.
Stefan Hickel CFD for Aerospace Engineers

29

Ariane-5 ECA flight 164


ESA, 2005

Stefan Hickel CFD for Aerospace Engineers

30

Simulation of planetary weather and climate

or even larger
Stefan Hickel CFD for Aerospace Engineers

Picture: http://www.nas.nasa.gov/SC14/demos/demo35.html

31

Computational Fluid Dynamics


CFD:
can replace tedious experimental studies
provides solutions in cases for which experiments are too
expensive, too complex, or too dangerous
(very large or very small geometries, very hot or poisonous
fluids, reconstruction of accidents or terrorist attacks, testing
plans for disaster control,)
can predict future developments and their impact
(weather forecast, global warming, )

Stefan Hickel CFD for Aerospace Engineers

32

Computational Fluid Dynamics


CFD:
can replace tedious experimental studies
provides solutions in cases for which experiments are too
expensive, too complex, or too dangerous
(very large or very small geometries, very hot or poisonous
fluids, reconstruction of accidents or terrorist attacks, testing
plans for disaster control,)
can predict future developments and their impact
(weather forecast, global warming, )

But how trustworthy are such predictions?


Stefan Hickel CFD for Aerospace Engineers

33

CFD =

Computational Fluid Dynamics


or

Colors For Directors


?
Stefan Hickel CFD for Aerospace Engineers

34

Periodic Hill Flow


Simple academic example
ERCOFTAC test case 9.2 designed for the evaluation and
cross comparison of numerical methods
Flow separation at a curved surface, and reattachment
between the hills
Relevant for technical applications (e.g., cooling channels)

Stefan Hickel CFD for Aerospace Engineers

35

Periodic Hill Flow


Simple geometry, simple boundary conditions

Complex unsteady vortex structures


Kelvin-Helmholtz instability
Grtler instability
Stefan Hickel CFD for Aerospace Engineers

36

Periodic Hill Flow


LES with academic flow solver (TU Mnchen)
3.2 million cells, computing time ca. 700 CPUh (in 2005)
Reynolds number 10595

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.22 - reattachment at x/h4.4
Stefan Hickel CFD for Aerospace Engineers

37

Periodic Hill Flow


CFX
2D RANS
Zero-equation model (algebraic equation for eddy viscosity)

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.41 - reattachment at x/h8.4
Stefan Hickel CFD for Aerospace Engineers

38

Periodic Hill Flow


CFX
2D RANS
Spalart-Allmaras model

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.41 - reattachment at x/h7.6
Stefan Hickel CFD for Aerospace Engineers

39

Periodic Hill Flow


CFX
2D RANS
k- model

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.87 - reattachment at x/h2.6
Stefan Hickel CFD for Aerospace Engineers

40

Periodic Hill Flow


CFX
2D RANS
Wilcox k- model

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.26 - reattachment at x/h5.8
Stefan Hickel CFD for Aerospace Engineers

41

Periodic Hill Flow


CFX
2D RANS
Menter SST model

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.26 - reattachment at x/h7.4
Stefan Hickel CFD for Aerospace Engineers

42

Periodic Hill Flow


CFX
2D RANS
Speciale, Sarkar and Gatski (SSG) Reynolds stress model

Pseudo streamlines based on time averaged velocity


Flow separation at x/h0.66 - reattachment at x/h3.6
Stefan Hickel CFD for Aerospace Engineers

43

Periodic Hill Flow


very simple configuration
low Reynolds number and incompressible fluid
inexpensive (RANS) models fail to predict separation bubble
results for expensive (LES) models depend on grid resolution
and numerical method
Noticeable disagreement between codes and models

Be always critical about the results of


numerical simulations!

Stefan Hickel CFD for Aerospace Engineers

44

Another example...
Shocktrain in a divergent duct [Giglmaier et al.]

Stefan Hickel CFD for Aerospace Engineers

45

Computational Fluid Dynamics


Steps and best practice for performing CFD
Physical
Problem

Knowledge and Experience

Physical
Modeling
Physical Model
Mathematical
Modeling
Mathematical Model

Flow Solver
Simulation
Numerical Solution

Discretization
Algebraic Model

Error Estimation

Validation

Algorithmization
Solution Algorithm
Programming
Stefan Hickel CFD for Aerospace Engineers

Sensitivity Analysis
Analysis and
Interpretation
Physical Result
46

Computational Fluid Dynamics


A CFD simulation always consists of three (plus one) steps:
0. Think about the problem
1. Pre-Processing (CFX-Pre and ICEM CFD)
- Generate computational grid
- Define boundary conditions
- Define models for fluid and turbulence

e.g., with ICEM


e.g., Definition File
with CFX-Pre

2. Solution (CFX-Solver)
- Compute the solution

observe convergence

3. Post-Processing (CFX-Post, or Tecplot, Paraview,)


- Visualize the results
- Quantify uncertainties and errors
- Analyze and physically interpret results
Stefan Hickel CFD for Aerospace Engineers

colorful pictures and


new knowledge
47

This course:
Physical
Modeling
Mathematical
Modeling
Numerical
Modeling
Solution
Interpretation

Stefan Hickel CFD for Aerospace Engineers

Modeling and Interpretation


To model means to approximate
and to neglect!
Too complex or unknown processes
are replaced by simpler processes
(with hopefully similar effects).
The chosen level of abstraction
depends usually on rules of thumb
and lots of experience.
Be always critical about the results
of numerical simulations!
48

This course:
Physical
Modeling
Mathematical
Modeling
Numerical
Modeling

Parameters and
dimensionless numbers
Conservation laws
Turbulence models
Boundary conditions
Discrete operators
Solution algorithms
Computational grid

Solution
Interpretation

Programming
Running a simulation
Visualization
Analysis and validation

Stefan Hickel CFD for Aerospace Engineers

49

This course:
Physical
Modeling
Mathematical
Modeling
Numerical
Modeling

Parameters and
dimensionless numbers
Conservation laws
Turbulence models
Boundary conditions

Interpretation

Solution algorithms
Programming
Running a simulation
Visualization
Analysis and validation

Stefan Hickel CFD for Aerospace Engineers

-> Definition file

Discrete operators
Computational grid

Solution

CFX Pre

ICEM CFD
-> Grid file

CFX Solver
-> Results file

CFX Post
-> Pictures and
knowledge
50

Tutorials: cylinder flow

3D wake flow of a circular cylinder, Re = 100


large-scale, unsteady flow separation and vortex shedding:
Krmn vortex street
of great practical interest: cables, pipelines, chimneys

Stefan Hickel CFD for Aerospace Engineers

51

Krmn vortex street

Stefan Hickel CFD for Aerospace Engineers

52

S-ar putea să vă placă și