E=Hydraulics-pipe flows, channel flows, intuitive relations
Theoretical=Formulation of equations
CFD=to start with Laplace eqn , human agents computing in 1900 etc
Cycle continues
If theoretical and computational solution do not match with experimental, we need to modify theoretical formulations
For simple flows the cycle has stabilized
For complex flows : like unsteady, hypersonic or combustion, multi-phase, bio-flows etc formulations are still evolving
Multi scale :
DNS, : no assumption, only continuum
LES : sufficiently large to reduce computation , but smallest size is such that unsteadyness of the given type of flow is captured. E.g in case of flow in weather prediction, the size could be several meters, and in case of flow around a wing, the size would be of the micron order.
LES with wall layer: walls constrain fluctuations ( V, V’)
RANS: I recognise my computiing power. Now instead of fluctuations, I only model the effect of fluctuations. E.g Kinetic energy ( u’2+v’2+w’2 )
this is 1 eqn turbulence model
rate at which K is destroyed : epsilon ( as K cannot be growing always . So it has to be balanced with growth)
Current flavour : RANS
Next 5 years : LES with wall layer
Eulers eqn are used in Aerospace only : as no viscosity , only compressibility matters
Incompressible/Invsicid =Laplace : outdated
Kolmogorov scales – smallest scale in continuum
Fluid= gas+liquid
Conservation of mass = conservation of volume ( rho is constant)
More correctly it should be
(u/x + v/y + w/z) = 0
In eulerian : there is another problem
Lagrangian: cell moves with flow. Deform yes, but volume remains
In euler: I draw in volume space, flow moves across
Then why eulerian: permits conservation .
Temperature is like color, no role in flow
Thus fluid mechanics separated from Heat transfer where conduction and radiation are considered.
Convection is only used thru empirical formulae.
Now with CFD : heat+fluid flow together
Now CpT (energy per unit volume) is variable along with mass, momentum
Earlier T was only measured and hence variable but that is only the effect.
How do I get pressure guess.
There was no way how to guess better p
Another issue: the eqn only shows gradient of p, but not the absolute value of p. This would create both mathematical anamoly and physical meaninglessness.
In fact , in incompressible flow , there is at least some semblance of meaning.
But in compressible flow this will be very crucial.
In fact , only about 30 years ago, a hack was developed by Patankar of IITB , along with Spalding at UK
SIMPLE : algorithm will be explained by next lectures ;
P=RT is standard eqn of state. But we do not have T as variable. We consider energy i.e e internal energy
Thus
It is reduced to
p = (-1)(E- ½ u2)
Kg/m3 ( ) (
Special methods to take care of 2nd law. So ensure entropy is not negative. Shock results in entropy growth
Incompressile flow methods cannot handle this.
That is why separate ays for compressible, and incompressible
One may ask: Is incompressible a special case of compressibel? NO
The model is different. In deed even the physics is. Sound and matter move at drastically different speeds.
If compressible is made a special case into incompressible : assumptions will be violated
Perhaps , computers with very high capacity may be able to use models that capture these radical differences.
Iggest headache: interpretation
Common mistake: e.g presuure differene
But user may use absolute values.
Say 10 values of pr at inlet, 10 values of pr at outlet.
In fact equations govern only the gradient. So absolute value may be floating
( subtract pr)
Another mistake
Habit of listing the variables and their beautiful plots
In fact ranges of values given to visualizer will show gradients where values may be ( in absolute terms) very much same
( contour plot of density: interpret density) ( incompressible flow should not show any variation, let alone beautiful plot )
Examples : whether viscocity needed
if only lift : inviscid solution is fine
if drag also needed: viscosity is a must. May be RANS
Academically : cavity: how many vortices, whether separation etc
In industry : different concersm
Academic problem of Driven cavity : 2-d, 3-d
If I need Cl, Cd , Cm of 0.001 level accuracy, why go for fine mesh
e.G : an industrial problem: Centre of pressure : every time different. In 3-d , no centre of pressure. So could not give load for structural analysis which takes data of load only as CP.
CFD ready cad : negotiate the correct geometry : e.g landing gear, angle of tail deflection .
Cl, Cd, Cm : Many permutations for arriving at the angle of attack, tail deflections . Add to that flap deflection, landing gear deflection, aileron deflection. The user has to agree.
Currently Automobile: CAD repair : months
Solution: days
Postprocessing : days
Similarly Wing : days
But aircraft: solution may take long
Business cycles : intermediate sign offs
Not flow field but only analysis
Not the pressure distribution all over but the loads on body