CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Numerical Method (

Andrew Hayes June 15, 2005 10:04

Numerical Method
I have been working on a simple numerical program to calculate the velocity profile between two plates. I assume the initial Pressure, the exit Pressure=0, plate length,x, and distance between plates,h. I can then calculate an inlet velocity. If I solve the implicit finite difference model in steady state form I get a typical parabolic velocity profile. And, when I average the output velocities I get a velocity very close to the inlet velocity - some error due to small grid size. But, when I move to the time dependent model I cannot get the velocities to average to the inlet velocity any more. I do, however, get the velocity profile shapes (square in the beginning to parabolic). When I evaluate the dP/dx term in the NV equation I simply calculate the deltaP for the whole pipe and then multiply it by the distance at which the velocity profile is being evaluated [(P/x)*dx], then I divide that by the distance, dx. That worked with the steady state model, but doesn't work with the time dependent model.

Any suggestions would be very helpful.


Andrew Hayes June 15, 2005 14:09

Re: Numerical Method
NV should be NS..navier-stokes..sorry

Mani June 16, 2005 14:52

Re: Numerical Method
Provide some details on how you made your model time dependent. The flow itself is still supposed to be steady, since you use steady boundary conditions, correct?

Andrew Hayes June 17, 2005 08:30

Re: Numerical Method
it is developing flow between 2 plates, 1-d, no-slip boundary condition at the wall. I know the inlet Pressure, P, the length of the plates, x, and the distance between them, h. So, if I were to define a time-step and look at the velocity profile at the first time step it should be square in form, which I get. Then, as the time increases it becomes more parabolic. I get that too. But, I am under the impression that if I average the velocities at all the grid points at a certain time step I should come up with my Vin, which I don't get.

my equation: (standard Navier-Stokes equation)

(dUx/dt) = -(1/rho)(dP/dx) + nu(d^2Ux/dy^2)

Jim_Park June 17, 2005 09:03

Re: Numerical Method
The advection terms are negligible at steady state, as is the second of the momentum equations. I'd think that, during a startup transient, a lot of things are happening in the entrance region of the flow regime that are not 1-d.

In that case, the terms missing from your equation could screw things up during the startup. Put another way, you'll get that Uinlet = avg(Ux) when dUx/dt = 0 everywhere. Perhaps not before.

If you haven't done so, run your simulation out to convergence and see if that solves your problem.

Andrew Hayes June 17, 2005 09:17

Re: Numerical Method
I have run it at steady state (dUx/dt = 0) and have gotten the parabolic profile, and when I sum the velocities at each grid point I get the Vin. That was my only question; about the velocities at each grid point at each time step averaging to the Vin. The profiles look as the should, but the velocities do not average.


All times are GMT -4. The time now is 19:38.