mathematical behaviour of pde's
TRANSCRIPT
Mathematical
behaviour of PDE’s
Arvind Deshpande
Arvind Deshpande (VJTI) 23/7/2012
Governing equations of fluid flow
(Navier Stokes equations)
)(&
)()()()(
)()()()(
)()()()(
)()()()(
)(0)(
fstateEquationsoCvTiRTP
ergyInternalEnSkgradTdivpdivVVidivt
i
momentumZSgradwdivz
pVwdiv
t
w
momentumYSgradvdivy
pVvdiv
t
v
momentumXSgradudivx
pVudiv
t
u
MassVdivt
i
Mz
My
Mx
Arvind Deshpande (VJTI) 33/7/2012
Governing equations of fluid flow
(Euler equations)
)(&
)()()(
)()()(
)()()(
)()()(
)(0)(
fstateEquationsoCvTiRTP
ergyInternalEnSpdivVVidivt
i
momentumZSz
pVwdiv
t
w
momentumYSy
pVvdiv
t
v
momentumXSx
pVudiv
t
u
MassVdivt
i
Mz
My
Mx
Arvind Deshpande (VJTI) 43/7/2012
Comments on governing equations
1. They are a coupled system of nonlinear partial differential equations and hence are very difficult to solve analytically.
2. For the momentum and energy equations, difference in the conservation form and non conservation form is the just the left hand side.
3. Conservation form contains terms on the left side which include the divergence of some quantity.
4. In CFD literature, entire block of equations are called Navier-Stoke’s equations.
5. All equations for inviscid flow are called Euler’s equations.
6. Conservation form of the governing equations provides a numerical and computer programming convenience in that the continuity, momentum and energy equations in conservation form can all be expressed by the same generic equation. This helps to simplify and organize the logic in a given computer program.
Governing equations are “bread and butter” of CFD-learn them well.
Arvind Deshpande (VJTI) 53/7/2012
General Transport equation in
Differential form
SgraddivVdivt
)()(
)(
Rate of
increase of φ
of fluid
element
Net rate of flow
of φ out of the
fluid element
Rate of increase of φ
due to diffusion
Rate of increase
of φ due to
sources
Arvind Deshpande (VJTI) 63/7/2012
General Transport equation in
Integral form
dVSdAgradndAVndVt
CVAACV
)(.)(.
Rate of increase
of φ fluid element
Net rate of
decrease of φ
due to convection
across the
boundaries
Net rate of increase
of φ due to diffusion
across the
boundaries
Net Rate of
creation of φ
Arvind Deshpande (VJTI) 73/7/2012
General Transport equation in
Integral form
)()(.)(.
)()(.)(.
unstadydVdtSdAdtgradndAdtVndtdVt
steadydVSdAgradndAVn
t CVt At At CV
CVAA
Arvind Deshpande (VJTI) 83/7/2012
Shock capturing
In flow fields involving shock waves, there are sharp discontinuous changes in flow field variables, P,ρ, V, T etc. across the shocks.
Shock capturing methods are designed to have the shock waves appear naturally within the computational space as a direct result of the general algorithm, without any special treatment to take care of the shocks themselves.
Ideal for complex flow problems involving shock waves for which we do not know either the location or the number of shocks.
Numerically obtained shock thickness bears no relation whatsoever to the actual physical shock thickness and precise location of the shock is uncertain.
Conservation form of the governing equations is suitable for shock capturing
Arvind Deshpande (VJTI) 93/7/2012
Shock fitting
In shock fitting methods, shocks are explicitly introduced into the flow field solution. Analytical relations are used to relate the flow immediately ahead of and behind the shock and governing equations are used to calculate the remainder of the flow field between the shock and some other boundary such as surface of a aerodynamic body.
Shock is always treated as discontinuity and its location is well defined numerically.
For a given problem, you have to know in advance approximately where to put the shock waves and how many they are. For complex flows, this can be a distinct disadvantage.
For shock-fitting method, satisfactory results are usually obtained for either form of the equations, conservation or non conservation.
Arvind Deshpande (VJTI) 103/7/2012
Boundary conditions
Real driver for any particular solution
Dirichlet boundary condition - Specification of dependent variables along the boundary
e.g. For Viscous flow, Wall boundary condition
V = Vw at the surface (No slip)
For a stationary wall, V = 0
Known wall temperature, T= Tw
Arvind Deshpande (VJTI) 113/7/2012
Newmann boundary condition
Specification of
derivatives of dependent
variables along the
boundary
e.g. 1) if wall temperature
is changing due to heat
transfer from or to the
surface
2) Adiabatic wall
0
nn
T
K
q
n
T
n
TKq
n
n
.
.
Arvind Deshpande (VJTI) 123/7/2012
Robbins Condition
The Derivative of the dependent variable is given as a
function of the dependent variable on the boundary.
Arvind Deshpande (VJTI) 133/7/2012
Inlet and outlet
Inlet – Density, velocity and temperature at inlet
Outlet – location where flow is approximately unidirectional and where surface stresses take known values.
For external flows away from solid objects and for internal flow, at a location where no change in any of the velocity components in direction across the boundary and Fn = -P & Ft = 0
Specified pressure, 0,0
n
T
n
un
Arvind Deshpande (VJTI) 143/7/2012
Other boundary conditions
Open boundary
condition
Symmetry boundary
condition
Cyclic boundary
condition 21
0
0
n
n
un
Arvind Deshpande (VJTI) 153/7/2012
Initial conditions
Everywhere in the solution region ρ, V and
T must be given at time t = 0
The Initial and Boundary conditions must
be specified to obtain unique numerical
solutions to PDEs
Well posed problem
Arvind Deshpande (VJTI) 163/7/2012
First Order 0
yG
x
Second Order 02
2
yx
Third Order 02
2
3
3
xyxx
Partial Differential Equations
Classifications of PDE’s according to order
Arvind Deshpande (VJTI) 173/7/2012
Mathematical behavior of PDE’s
0
0
2
2
22
2
2
cdx
dyb
dx
dya
gfy
ex
dy
cyx
bx
a
If b2-4ac < 0, elliptic equation (Imaginary characteristics)
If b2-4ac = 0, parabolic equation (1 real characteristics)
If b2-4ac > 0, hyperbolic equation (2 real and distinct
characteristics)
Arvind Deshpande (VJTI) 183/7/2012
Eigen value method
det [Ajk-λI] = 0
If any eigen value λ = 0, the equation is parabolic.
If all eigen value λ ≠ 0 and they are all of the same sign,
the equation is elliptic.
If all eigen value λ ≠ 0 and all but one are of the same
sign the equation is hyperbolic.
N
j
N
k kj
jk Hxx
A1 1
2
0
Arvind Deshpande (VJTI) 193/7/2012
Elliptic PDE
Typical Examples are
( Laplace’s Equation – Irrotational flow of
an incompressible fluid, steady state
conductive heat transfer)
02
2
2
2
yx
and ),(2
2
2
2
yxgyx
( Poisson’s Equation)
Note that In both of the eqns, b=0, a=1, c=1 which makes
442 acb which is < 0
The solution domain of Elliptic Eqn has closed ended nature
Arvind Deshpande (VJTI) 203/7/2012
Pictorial Representation of Elliptical Problem
Domain of
dependence
Arvind Deshpande (VJTI) 213/7/2012
Elliptic PDE
Characteristics are imaginary/complex
Information propagates everywhere
Equilibrium problems (div grad φ = 0)
Boundary value problems
Smooth solution
Steady state temperature distribution of a insulated solid rod
Steady viscous flow
Steady, subsonic inviscid flow
Arvind Deshpande (VJTI) 223/7/2012
Parabolic PDE
A Typical Example is
( Heat Conduction or Diffusion Eqn.)
Where is positive, real constant
In above eqn. b=0, c=0, a = which makes 042 acb
The solution advances outward indefinitely from Initial Condition
This is also called as marching type problem
The solution domain of Parabolic Eqn has open ended nature
)(
2
2
divgradt
xt
Arvind Deshpande (VJTI) 233/7/2012
Pictorial Representation of Parabolic Problem
Domain of dependence
Arvind Deshpande (VJTI) 243/7/2012
Parabolic PDE
Information travels in one particular direction (downstream)
Time dependent problems which involve significant amount if dissipation
Initial-Boundary value problems
Smooth solution
Unsteady heat conduction
Unsteady viscous flows
Thin shear layers – Boundary layers, jets, mixing layers, wakes, fully developed duct flows
Arvind Deshpande (VJTI) 253/7/2012
Hyperbolic PDE
A typical example is
( Wave Equation)
Where c2 is real constant and always positive
In above eqn b = 0, a = c2, c = -1 which makes
which is >0
The solution domain of Hyperbolic Eqn has open ended nature
Two Initial conditions are required to start the solution of
Hyperbolic eqn in contrast with Parabolic eqn where only one Initial
conditions is required.
divgradct
xc
t
2
2
2
2
22
2
2
22 44 cacb
Arvind Deshpande (VJTI) 263/7/2012
Pictorial Representation of Hyperbolic Problem
Bo
un
da
ry c
on
ditio
ns
Bo
un
da
ry c
on
ditio
ns
Initial conditions
P(x’,t’)
Domain of
dependence
Domain of
influence
Arvind Deshpande (VJTI) 273/7/2012
Hyperbolic PDE
Characteristics are real and distinct
Information propagates along these characteristics
Time dependent problems which involve negligible dissipation
Initial-Boundary value problems
Solution may be discontinuous
Compressible fluid flows at speeds close to or above the speed of sound
Steady, inviscid supersonic flow
Unsteady inviscid flow