CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   One K-e problem (https://www.cfd-online.com/Forums/fluent/28856-one-k-e-problem.html)

 Jie August 24, 2001 12:00

One K-e problem

I use K-e viscous model to calculate one problem. For one low pressure, the model converged, and I writed the interpolation values : velocity, pressure and temperature .

For the 2nd case, I increased the pressure, initialized the flowfield with the data above, then calculated it. I found that it was too difficult for energy and continuty to converge, while other values were Ok.

What's wrong with it?Anybody have ideas?

Jie

 Scott Whitney August 24, 2001 13:33

Re: One K-e problem

There are four possible outcomes with every CFD calculation: 1) The calculations diverge. 2) The calculations oscillate between two or more equally good answers, never converging on any one specific answer. 3) The calculations converge. 4) The calculations converge but the residuals which are scaled by an arbitrary number never reach an arbitrary value (such as 0.0001).

The most difficult problems to solve involve outcome #1. Outcome #2 may or may not be possible with your simulation (some problems have more than one realistic solution). From what I've seen, most people on the Fluent forum have reached outcome #4.

Before we can help, which of the four are you getting?

 Chetan Kadakia August 24, 2001 23:50

Re: One K-e problem

When you say increased the pressure, do you also mean an increased pressure gradient? This makes the problem more complex, and cells values will change more rapidly. As a problem becomes more complex, the more you need to control the solution, or it will blow up on you. First answer Scott's question, and then answer my questions. Second, why do you initialize data of the second problem with the solution of the first? The idea of a good set of initial values are approximate values that are relatively close to the final solution. If your first problem finds the pressure in one cell to be 1 atm, but the same cell may have a value of closer to 2 atm in the second problem.

 Jie August 27, 2001 16:00

Re: One K-e problem

I think the problem is #2 case.And the oscillation is very small, approximately 1/000 of the residual. So maybe my problem is close to #4 case.

 Jie August 27, 2001 16:04

Re: One K-e problem

Hello,

When calculation a nozzle model, if you let the inlet pressure is 100 atm and the outlet pressure is 1 atm, which initial value should you use? In my idea, the better is that first calculate the 10 atm as the inlet pressure, then add it each time by 10 atm, that's to say, I need to calculate is for several times. ' One time one case ' doesn't work here.

--Jie

 Chetan Kadakia August 27, 2001 16:08

Re: One K-e problem

Try it, but I think it will be better to just start with full pressure range and use lower under-relazxation factors. But I'm not the expert here. Anybody else got ideas?

 John C. Chien August 28, 2001 04:00

Re: One K-e problem

(1). You are solving a compressible flow problem. (2). So, you need to use the compressible flow formulation. Not the one for incompressible or low speed flow formulation. (3). The compressible flow formualtion is basically a transient flow formulation, and you are following (or calculating) the flow development. (4). So, you can set your initial flow field in the way you like, and start the transient calculation (does not have to be time accurate). (5). Since the 10atm solution will be completely different from that of the 100atm condition, you will not get faster convergence by using it as the initial flow field. It is like running the flow with 10atm, then suddenly you have explosion upstream and increase the total pressure to 100atm. In other words, you will have a shock wave moving through a established nozzle flow. (6). Since you don't know the final solution, and in compressible flow, the pressure is the controlling factor, the mass flow etc. will have to be re-adjusted, even if you used the 10atm solution as the initial flow field guess. (not a good idea at all) (7). The simplest way to do is to assume the zero flow field (no flow through), and set the inlet total pressure to 100atm, then start the flow calculation. And you are likely to set the time setp or CFL number to a very small value to get it started. (8). It will take a while to reach convergence. (depends on the mesh size) (9). You will likely get shock waves in the supersonic section of the nozzle also. It is a good idea to study the corresponding 1-D flow problem in the gasdynamic text book first.

 All times are GMT -4. The time now is 22:45.