18-Introduction To Fluid Flow

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 32

Lecture 18:

Introduction to Fluid Flow


Last Time…

z Looked at boundary conditions specific to convection-


diffusion problems
» Inflow, outflow, geometric bc
This Time …

z Start looking at the computation of fluid flow


» Sequential schemes
» How to introduce pressure into continuity?
z Look at where to store pressure vis-à-vis velocity
» Checkerboarding problems
z Look at staggered mesh as a remedy for checker-
boarding for structured meshes
Introduction to Fluid Flow

z So far, we have assumed that the fluid flow field is


known
z We now turn to the computation of the fluid flow
z Navier-Stokes equations are special case of
convection diffusion equation with φ = u, v or w, and
Γ=µ and appropriate S
z So why the fuss?
Steady 2D Navier-Stokes Equations

z Can write Navier-Stokes equations as:

z For Newtonian fluid, can confirm that:


Why The Fuss?
z NS equations are non-linear
» Not by itself a problem
» Can do Picard iteration
z More than one momentum equation (x,y,z)
» Not a problem
» Can solve each sequentially
z Source terms?
» Stress tensor sources can be computed
z Main issue is that pressure is not known!
» Must find a way to find it
» What PDE to use?
Issues with Pressure Computation

z Two different problems


» Where to store pressure vis-à-vis velocity
» How to find an equation for pressure
z In 3D, we have 4 equations in 4 unknowns
» 3 velocities, 1 pressure
» 3 momentum equations and one continuity
equations
z In principle, possible to solve
Vorticity-Stream Function Formulations

z Used during 70’s-80’s


» Derive one PDE for stream function Ψ and one PDE for the
single component of vorticity ξ in 2D
» Eliminates pressure as variable
» Only 2 PDE’s as opposed to 3 for primitive variable
formulations using (u,v,p)
z However, no stream function definition possible in 3D flows
z Vorticity-vector potential formulations in 3D
» 6 components (3 for vorticity, 3 for vector potential)
» Fewer components (u,v,w,p) for primitive variables
z Difficult to derive bc’s
z Primitive variable formulations preferred today
Sequential Solution

z Our approach so far has been to solve sets of coupled


PDEs in a sequential manner
z We identify:
» u-momentum equation for u velocity component
» v-momentum equation for v velocity component
» Continuity as PDE for pressure
z Possible solution loop:
» Discretize and solve u over domain, assuming v, p
known from current iterates
» Similarly solve v from discrete v-mom equations
» Similarly solve p from discrete continuity
» Continue until convergence
Equation for Pressure
z Continuity equation ∂ρ
+ ∇ ⋅ ( ρV ) = 0
∂t
z For compressible flows, we could convert this into an
equation for pressure. For a perfect gas, for example:

ρ = P RT

z What about incompressible flows?


» No relationship between pressure and density
» No obvious way to introduce pressure into
continuity equation
Sequential vs. Direct Schemes

z Important to recognize that the problem of pressure


computation for incompressible flows is an artifact of
sequential solution procedure
z Only for sequential procedures is it necessary to
identify a PDE for each unknown variable
» u-mom for u velocity; v-mom for v velocity,
continuity eqn for pressure; energy equation for
temperature etc.
z For direct solution procedures, no such one-on-one
identification necessary
Direct Schemes

z Discretize u,v momentum equations and continuity at all


cells
z Assemble a big matrix: Ax = b
z Solution vector is:

x = [u1 , u2 ,..., u N , v1 , v2 ,..., vN , p1 , p2 ,..., pN ]


T

z No need to identify which unknown gets solved using


which PDE
z Unfortunately too expensive for practical use as of this
writing
Pressure Calculation Issues

z Problem in using sequential procedures to compute


pressure
z No pressure in incompressible continuity equation
» Need to find a way to introduce p into continuity
» Path to solution issue
Pressure Calculation Issues(Cont’d)

z Another issue has to do with storage of pressure and


velocity
» Obvious choice is to store pressure and velocity (u
and v) at cell centroids
» Turns out that this type of “co-located” storage is
problematic
» Gives wiggly pressure/velocity solutions
» Feature of discretization, not path to solution
» Will not go away no matter what “path to solution”
we use
Steady 2D Navier-Stokes Equations

z Navier-Stokes equations:

z For Newtonian fluid, sources given by:


Co-Located Pressure-Velocity Storage

z Pressure gradient is really


a source term, but let us
consider it separately
since it is not known a
priori
z Consider 2D uniform
Cartesian mesh
z Store (tentatively) u,v,p at
cell centroids
Pressure Gradient Term

z Control volume integration requires evaluation of

z Write in discrete form

Note: this is a
vector term!
Pressure Gradient Term (Cont’d)

z Outward-pointing area vectors:

z For u-momentum equation


Pressure Gradient Term (Cont’d)

z v-momentum pressure gradient

z Discretizing other terms as before:


Pressure Interpolation

z We need pressure on faces, but it is available only on


cell centroids
z Interpolating to face:
Pressure Interpolation (Cont’d)

z Linear interpolation leads to following pressure terms

Notice that pP does not


appear in pressure
gradient terms for
control volume
equations at P
Discretization of Continuity Equation

z Consider 2D steady continuity equation

z Integrate over control volume

z Write discrete form:


Continuity Equation (Cont’d)

z Velocity vector:
z Discrete continuity equation:

z Need velocity at faces, have it at centroids


» Linearly interpolate to faces
Continuity Equation (Cont’d)

z Interpolate linearly

Note: u and v at
point P do not
appear in
control volume
equation at P
z Discrete continuity equation:
Checkerboarding

z Velocity fields with checkerboard pattern would be


seen by discrete continuity equation as uniform flow
field
z Pressure fields with checkerboard pattern would be
seen by momentum equation as uniform field
Checkerboarding (Cont’d)

z Say we pick up checkerboard pattern in velocity which


satisfies continuity equation
z This implies a pattern of momentum sources
z Pressure is a part of the solution,
» Can find pressure field that exactly balances
checkerboarded momentum source
» Legitimate solution to discrete equations, but
physically impausible
z In practice, we don’t pick up exact checkerboarding
» Wiggly pressure and velocity solutions
Remedy: Staggered Mesh

z Store pressure at main cell centroids


z Store velocities on staggered control volumes
Staggered Mesh: Momentum Equations

z Pressure difference for ue

z Pressure difference for vn

z No need to interpolate pressure to face


» Available on faces of momentum cv’s
z Pressure checkerboarding avoided
Staggered Mesh: Continuity Equation

z Discretize continuity equation on main control volume:

z No need to interpolate velocities – no checkerboarding

Velocities
available on face
where needed
Discrete Equations

z Continuity:

z U-momentum:

z V-Momentum:
Momentum Equation Neighbors

z Neighbors of ue are shown

z Neighbors of vn?
Closure

z Two different issues with respect to finding pressure


z First has to do with how to get pressure into continuity
equation for incompressible flows
» Have not figured this out yet
z Second issue has to do with where to store pressure
vis-à-vis velocities
» Staggered storage avoids checkerboarding
» Added geometric complexity due to three different
control volume types in 2D; four types in 3D

You might also like