Вы находитесь на странице: 1из 82

Computational Fluid Dynamics

(CFD): Why, What and How


Design Laboratory I
Autumn Semester 2014-2015

Department of Chemical Engineering


Indian Institute of Technology Bombay
15th September 2014

Why ?????

Equations (PDEs) governing fluid flow are known


Very well for single-phase laminar flow
Not so well for turbulent/ multiphase flow but improving
gradually

Geometries for which equations are to be solved


May be simple (for illustration purpose)
Most of the times very complex (real problems for industry)
Analytical solution
Possible for very simple geometries
Not possible for complex geometries

Laminar flow in a pipe


Header for flow distribution

Air-pulsed
liquid column

Stirred Tank

What is the way out


Carry out experiments (how many and at what scale)

Different impellers

Solve the flow equations by using numerical techniques

Computational Fluid Dynamics


Numerical solution of the equations governing fluid flow (and
other associated phenomena) for the geometry of interest

CFD vs Experiments
Reduction in lead times of new designs
Ease of implementation (transportable experimental
facility)
Study of systems under hazardous conditions and beyond
their normal performance limits
Practically unlimited details of results give great insight into
the physics of the problem leading to knowledge based design
Experiments are nevertheless indispensable for validation
purpose

Application Areas
Aerodynamics
Turbomachinery
Chemical Process Engineering
Oceanography
Environmental Engineering
Microfluidics

Marine Engineering
Combustion
Electronics
Hydrology
Meterology
Biomedical Engineering

What ?????

Components of a CFD Software


Preprocessor
CAD interface for making geometry
Specification of boundary and zone types
Grid generation

Solver
Specification of boundary conditions
Specifications of materials
Specification of flow models (e.g. single/multi phase,
laminar/turbulent)
Selection of appropriate numerical and discretisation schemes
Setting under-relaxation parameters and convergence tolerances
Writing user defined functions (e.g. drag models in two-phase
flow)
Solve the problem

Postprocessor
Graphical display of results such as vector plots, contour plots,
stream lines etc. for better understanding and effective
communication of results (Colorful Flow Diagrams)

Commercial Solvers
FLUENT (ANSYS)
CFX (ANSYS)
STAR-CCM (CD-Adapco)
CFD-ACE+ (ESI)
FLOW3D (FLOW SCIENCE)
PHOENICS (CHAM)
COMSOL MULTIPHYSICS (COMSOL)
EFD Lab (FLOMERICS)

Open Source Solvers


OpenFOAM
Su2

Commercial Pre/Post Processor


SOLID WORKS
STAR-WORKS
ICEM CFD
GRIDGEN
GAMBIT
TECPLOT

10

Equations Embedded in CFD Solver


Mass conservation Continuity equation
Newtons Second Law of Motion Momentum equation
Closure laws / equations (e.g. turbulence equations, drag models)
Energy conservation Energy equation
Other equations are also involved if the problem is a multiphysics
one

Continuity Equation

r
+ div( u ) = S m
t

Momentum Equation
( u )
P xx yx zx
r
+ div( uu ) = +
+
+
+ S Mx
t
x x
y
z
( v )
P xy yy zy
r
+ div( vu ) = +
+
+
+ S My
t
y x
y
z
( w)
P xz yz zz
r
+ div( wu ) = +
+
+
+ S Mz
t
z x
y
z

11

Energy Equation
( i )
u
u
u
r
r
+ div( iu ) = P div u + div(k grad T ) + xx + yx + zx
t
x
y
z
v
v
v
+ xy + yy + zy
x
y
z
w
w
w
+ xz
+ yz
+ zz
+ Si
x
y
z

Newtonian Fluid
u
u w
u v
r

+
+
2 x + divu
y x
z x

v
u v
v w
r
+
= + 2 + divu
y x
y
z y

w
v
w
w

r
+
+ 2 + divu
z
z x

z y

12

Navier-Stokes Equations
( u )
P
r
+ div( uu ) = + div( grad u ) + S Mx
t
x
Momentum equations for
( v )
P
r
+ div( vu ) = + div( grad v ) + S My a Newtonian fluid
t
y
( w)
P
r
+ div( wu ) = + div( grad w) + S Mz
t
z

Energy Equation for a Newtonian Fluid


( i )
r
r
+ div( iu ) = P div u + div(k grad T ) + + Si
t
2
2
2
2
2
u v w u v u w
= 2 + + + + + +
x y z y x z x
2
v w
r2
+ + + (div u )
z y

13

Turbulence Models in CFD Solvers


Classical models based on RANS equations
Zero-equation model: Mixing length model
Two-equation model: k- model
Standard k- model
RNG k- model
Realizable k- model
Reynolds Stress Model (RSM )
Algebraic Stress Model (ASM)

Advance Models
Large Eddy Simulation (LES)
Direct Numerical Simulation (DNS)

14

Numerical Solution of Governing Equations by the Solver


This is based on conversion of PDEs to algebraic equations following
a procedure called discretization in which values of derivatives of a
variable at a node is expressed in terms of its value at the node and
neighboring nodes
ay '' + by ' + c = 0

y = f (x) yi = f (x i )

' ' h 2 y ' ' 'h 3


y
y i +1 = y( x + h ) = y + y' h + i
+ i
+ .........
i
i
i
2!
3!
' ' h 2 y' ' ' h 3
y
y i 1 = y( x h ) = y y' h + i
i
+ .........
i
i
i
2!
3!
y i +1 y i 1
'
y

y
y
=
i 1
i
y i' = i +1
+ terms of O( h 2 ) Central Difference
2h
2h
y i +1 y i
'
y i +1 y i
'
Forward Difference y i =
yi =
+ terms of O(h )
h
h
y i y i 1
'
y i y i 1
'
15
=
y
Backward Difference i
yi =
+ terms of O(h )
h
h

Numerical Solution of Governing Equations by Solver


y i +1 2 y i + y i 1
y i +1 2 y i + y i 1
''
2
=
y
Central
Difference
y =
+ terms of O( h )
i
2
2
h
h
y i + 2 2 y i +1 + y i
y i + 2 2 y i +1 + y i
''
''
Forward
Difference
yi =
yi =
+ terms of O(h )
2
h2
h
y i 2 y i 1 + y i 2
y 2 y i 1 + y i 2
''
y
=
y i'' = i
+
terms
of
O
(
h
)
Backward Difference i
2
h2
h
''
i

First order upwind


Second order upwind
Power law
QUICK (Quadratic Upstream Interpolation for Convection
Kinetics)

16

Example of 1D Conduction Problem

d 2T h P

(T TA ) = 0
2
dx
kA

BC1 : @ x = 0,

BC 2 : @ x = L ,

d 2
2

(
mL
)
=0
2
dX

hP
m =
kA
2

T = T0

dT
=0
dx

T TA
=
T0 TA

BC 1 : @ X = 0 , = 1
d
BC 1 : @ X = 1,
=0
dx

X=

x
L

17

Example of 1D Conduction Problem


d 2
2

(
mL
)
= 0 BC1 : @ X = 0, = 1
2
dX
d
BC 1 : @ X = 1,
=0
dx

d 2
i +1 2i + i 1
2
2
2 (mL) i = 0

(
mL
)
i = 0
2
h
dX i

2
i 1 2 + (mL ) h 2 i + i +1 = 0 M-2 such equations for all internal nodes

BC1 : @ X = 0

1 = 1 1 equation valid at the 1st node

M 1
d
BC 1 : @ X = 1,
=0
=0 M
h
dx M
M M 1 = 0 1 equation valid at the Mth node
M equations can be used to solve for M unknowns

18

Type of Grid/ Mesh (Structured and Unstructured Meshes

19

Type of Grid/ Mesh (Triangular & Quadrilateral Elements)

20

Type of Grid/ Mesh (Bolundary Layer and Hybrid Meshes)

21

Choosing an Optimum Grid


Coarse Mesh
h is large
Terms O(h) or O (h2) will be large
Discretization errors will be more
Computationally easier to solve

Fine Mesh
h is smaller
Terms O(h) or O (h2) will be smaller
Discretization errors will be less
Computationally difficult to solve
22

Pressure Velocity Coupling for Incompressible Flows


There is no explicit equation for pressure
Pressure field needs to be assumed to obtain the velocity field by solving
momentum equations
The resulting velocity field should satisfy the continuity equation
If continuity equation is not satisfied, the initial guess of pressure field
is not correct and should be revised
Revision should be done in a systematic way using an algorithm called
Pressure-Velocity Coupling Scheme
In such a scheme the continuity equation is converted into a pressure
correction equation
Different Schemes have been proposed
SIMPLE (Semi-Implicit Method for Pressure Linked Equations)
SIMPLER (SIMPLE Revised)
SIMPLEC (SIMPLE Consistent)
PISO (Pressure Implicit with Splitting of Operators)
23

Example 1: Solid Suspension in an Industrial Scale Agitated Tank

T = 2300 mm
Singh et al. (2011). Flow simulations of industrial scale agitated tanks, BARC Newsletter, 319, 8-15
24

Example 1: Solid Suspension in an Industrial Scale Agitated Tank

CFD simulations revealed that flow in the bottom of the tank is upward
instead of being downwards which is not good for solid suspension
25

Example 1: Solid Suspension in an Industrial Scale Agitated Tank

A modification in which impeller was lowered was evaluated using


CFD. Flow at the bottom was now downwards but the flow rates were
26
small.

Example 1: Solid Suspension in an Industrial Scale Agitated Tank

A modification in which outer blades of the lowest impeller were


removed was evaluated using CFD. Flow at the bottom was now
downwards and high.
27

Example 1: Solid Suspension in an Industrial Scale Agitated Tank

A modification in which outer blades of all the impellers were removed


was evaluated using CFD. Flow at the bottom was now downwards and
quite high.
28

Example 2: L-L Two-phase Flow Patterns in Microfluidic Devices

Continuous phase: water


Dispersed phase: kerosene

Experimental flow patterns from Zhao et al.,


AIChE Journal, 52 (2006), 4052-4060

29 rd
Singh et al. (2012). Numerical simulation of liquid-liquid two-phase flow at microfluidic junctions,
3
European Conference on Microfluidics, Heidelberg, Germany

Example 2: L-L Two-phase Flow Patterns in Microfluidic Devices

Continuous phase: Olive oil


Dispersed phase: water

Experimental flow pattern from Mehrotra et al.,


Appl. Phy. Lett., 92 (2008)
30

Example 3: CFD Modeling of Pump-mix Action

Singh et al. (2008). CFD modeling of pump-mix action in continuous flow stirred tanks, AIChE Journal,
54(1), 42-55.
31

Example 3: CFD Modeling of Pump-mix Action

Straight Blade
Paddle (SBP)

Rushton Turbine
(RT)

Top Shrouded Turbine


with Non-radial
Rectangular Blades
(TSTNRB)

Both side Shrouded


Turbine with Radial
Trapezoidal Blades
(BSTRTB)

Top Shrouded Turbine


with Radial
Trapezoidal Blades
(TSTRTB)

Top Shrouded Turbine


with Radial
Rectangular Blades
(TSTRRB)

Top Shrouded Turbine


with Curved Blades
(TSTCB)
32

Example 3: CFD Modeling of Pump-mix Action

Wealth of data obtained from CFD


simulations was analyzed to understand the
pump-mix action
Based on understanding of pump-mix
action new designs of pump-mix impellers
were conceptualized

33

Example 4: CFD Modeling of Flow Electrolysers

34
Singh et al. (2012). Numerical simulation of flow electrolyser: effect of obstacles, Electrochimica Acta,79,
57-66.

Example 5: Mixing at Microfluidic Junctions

Several 1-2 microfluidic junctions were


conceptualized and evaluated vis--vis conventional
1-2 microfluidic junction

Singh et al. (2014). Numerical simulation of mixing at


1-1 and 1-2 microfuidic junctions, Chem. Eng. Proc.,
Article in Press.
35

Example 6: Numerical Simulation of Capillary Rise

36

Example 7: Formation of Drop on a Single Hole in a Plate

Experimental figures from Soleymani et


al., Chem. Eng. Res. Des., 86 (2008)

Example 8: Numerical Simulation of Air Pulsed


Liquid Column

Example 9: Animation of Stream Lines Past the Hand of a


Swimmer

Source # http://www.fluent.com/about/news/pr/pr11.htm

39

Example 10: CFD Simulation to Estimate Force while


Diving

A free-surface simulation of
the forces experienced
when diving, from CFD
simulation performed in
ANSYS CFD software.

http://www.deskeng.com/virtual_desktop/?p=6002)

40

Popular websites related to CFD


http://www.cfdreview.com

41

Popular websites related to CFD


http://www.cfdreview.com

42

Popular websites related to CFD


http://www.cfd-online.com

43

Popular websites related to CFD


http://www.cfd-online.com

44

Popular websites related to CFD


http://www.cfd-online.com

45

Popular websites related to CFD


http://www.cfd-online.com

46

Job Prospects

47

Job Prospects

48

How ?????
Flow Simulation and Virtual Tracer
Experiment in a Stirred Tank using CFD

Why Tracer Experiments ?


Real equipment do not behave ideally
For example, a stirred tank, though expected to behave as a perfectly
mixed vessel, may not give required duty due to problems like dead zones,
bypassing etc.
The tracer experiments are done to quantify the non-ideality and
represent the real life equipment as a series and/or parallel combinations
of ideal vessels
This process is called compartment modeling
Since the performance equations for ideal vessels are known the
performance of the compartment model and hence real equipment can be
predicted
CFD can be used to carryout the virtual tracer experiments
Following adequate validations, CFD based virtual tracer experiments
can be use to optimize the design. For example, the degree of mixing in a
stirred tank depends on impeller type, impeller diameter, impeller location
in the tank, number of blades . conducting real tracer experiments for
all these parameters will be expensive and time consuming. On the other
hand CFD based virtual tracer experiments can be conducted easily

Description of Geometry and Process Conditions


Tank diameter (T) = 100 cm
Tank height (H) = 100 cm
Impeller diameter (D) T/3
Clearance (C) = T/2
Impeller blade width = 0.2D
Impeller blade length = 0.25D
Impeller disk diameter = 0.75D
Number of baffles = 4
Baffle width = T/10
Baffle height = H
Hub diameter = 3 cm
Hub height = 3 cm
Shaft diameter = 1.5 cm
Baffle, disk and blade thickness = 5 mm
Flow rate = 20 lph
N = 300 rpm

Steps Involved in Simulation


Preprocessing in GAMBIT
Creation of as real as possible geometry of the tank
Specification of the boundary types
Mesh generation and export of mesh to FLUENT
Processing in FLUENT
Scaling of grid
Selection of turbulence models
Selection of discretization schemes
Specification of boundary conditions
Simulation of flow
Specification of a user defined scalar
Virtual tracer experiments
Postprocessing in FLUNET/ EXCEL
Visualization of velocity vectors
Comparison of theoretical and predicted F curve

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

CFD Simulation of Stirred Tanks


Impeller tends to establish the circulatory flow patterns
Baffles tend to break this pattern
A point in the tank may be more influenced by the baffles or by the
impeller
Tank volume is split into two parts
Inner volume rotating with the impeller
Outer volume stationary with baffles
The solution will depend on the location of the interface between
the two parts
Usually the interface is located between the impeller blade tip and
baffle tip
The equations in the inner volume are solved in a frame of
reference rotating with the impeller (moving mesh OR moving
frame of reference)
The equations in the outer volume are solved in laboratory frame
of reference

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Preprocessing in GAMBIT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Processing in FLUENT

Postprocessing in FLUENT

Postprocessing in EXCEL
1
0.9
0.8
0.7

0.5
0.4

Predicted

0.3

Theoretical

F( t ) = 1 e

t /

0.2
0.1
0
0

100

200

300

400

500

600

700

800

900

t (sec)
0.3
0.25
0.2

F(t)

F(t)

0.6

0.15

Predicted
0.1

Theoretical

0.05
0
0

10

20

30

t (sec)

40

50

I see and I forget


I hear and I remember
I do and I understand
- Confucius (551 479 BC)

Вам также может понравиться