I have just started trying to use CFD and am a bit inexperineced with it. Can anyone with experience in modelling free jets help me?

I'm trying to use the segregated solver, with the relizable reynolds stress model for a under-expanded compressible free jet.

Is there a rule of thumb for the type of grid for compressible free jets? I'm using hexahedral elements in fluent 3D but the iterations are very slow, I have approx. 80,000 elements in one quadrant of a plain jet, is this stupidly large?

I was told it is difficult to capture shock waves in CFD, is this true? and if so is it because of the solver or the grid resolution or both as I'm having problems getting anything more than the first shock cell in an under-expanded jet?

When I can get the solver to start (if it doesn't diverge, aside: how do underflow and overflow occur and how can I avoid them?) if i monitor the residuals I am unsure how to respond, using the under-relaxation factors, in repsonse to a given behavior (of the residuals).

I have been keeping the under-relaxation factors low to try and avoid divergence. Is there any problems with this? will the sloution eventually converge but just take longer?

Eventuallly the residuals will flatten out a stay effectively constant. Even though some of the residuals may be a order or so larger than the set critera, if they stay constant has the the solution conveged? as when this happens dropping the under-relaxation factors drops the residuals in the next couple of iterations, then th residuals remain constant again.

I know if the residuals begin to oscillate I should reduce the relaxation factors, but what if they don't oscillate and steadily rise (very slowly)?

Regular spikes in the residuals have also occured, what do they mean and is there anything I can do about them?

I realise these may be very basic questions, but any help would be appreciated as I would know if I'm going in the right direction, thank you.

(1). I am not sure why you are using this code to solve the under-expanded jet problem. (you are very lucky that it is not a timed bomb.) (2). So, you really have to find the reason to use this code first, otherwise, you will have un-answered questions. (3). I will give you some suggestions to see whether you can find the reason to use the code. (4). First, try to set the ambient pressure equal to the exit pressure. In this way, you will have the constant pressure free jet problem. In this problem, you should try to answer all of your questions, except the shock wave problem. (5). If you can not find the reason to use the code using this problem and be able to answer most of your questions by your self, you should stop using this code. There is a mis-match between the code and the problem (and the user as well). (6). After that, (assuming that the use of the code is justified) try to reduce the ambient pressure slightly and run another case. Repeat this process many times, until you reach the desirable pressure ratio. You should also try the mesh refinement for each case and make sure that the solution is mesh independent (including the formation of shocks). (7). This guideline is applicable to any other codes. (It is easier to get the information from the code developer about the feasibility of using this code to solve your problem. Trial-and-error approach will not teach the code smarter, unless it has learning capability in it.)

John, thank you for such a prompt response

I am running blind (partially) in a sense as I think you've gathered already from your message.

I have been trying to attain the desired pressure ratio by gradually increasing the nozzle NPR in steps. You're advice reassures me that I'm in roughly the right direction.

You're comment that you're not sure why I'm using Fluent for a under-expanded jet is a good question, as I don't know either, my CFD experience is very limited (I've only started looking at them to try and see if there is any correlation between case solutions from CFD and experimental measurements I've taken for my MPhil/PhD). The Fluent code is the only code I am aware of on college, the college has fluent 4.5 and has only recently this year aquired Fluent V5. I'm unclear as to why you think I'm lucky it hasn't time bombed? is this a very bad code to be using for compressible jets?

I agree with you it would be easier to obtain information from the code developer about the feasibility of the problem on their code. I have not spoken with Fluent, but from their web site it seems that for academic sites, only the registered license holder is entitled to support (unfortunately I'm not) and so (as I can see) help is limited to the manual or sites such as this. (The college has a firewall and I haven't been able to get much success trying to access newgroups.

Thanks again for you advice, I'll give it a try

Ricky

(1). I can not say that a code is good or bad, because a library of subroutines and modules is very similar to a school library with books, tapes, and CDs. (2). When you need a book, you first do the book search on the terminal. If the library does not have the book, then it can not solve your problem. (3). One way you can do is to post a message about your problem with detailed conditions of this under-expanded jet, to see whether anyone has solved the similar problem using this code. If someone has obtained good solutions using this code, then I would say that the code is good for your problem. Otherwise, there is no answer. (in many cases, the developer of the code doesn't even know the problem you are trying to solve.) (4). Not all Chinese food are the same. The general rule is to check at least three independent sources first about your problem before you buy the product. (5). I am not trying to make the answer difficult to understand. I am just saying that you are violating the common sense rule of using a commercial CFD code. (6). Without support, a commercial CFD code is just like a bomb ready to explode any time in the school environment. (it is like a gun without a safety lock and training) (7). By the way, if I am getting good solutions from a code-A for my problem-B, then the code is a good code. If it can not produce good solutions for my problem-B, it is not a good code even if it can produce good solutions for problem-C. This is the common sense rule.

I think you raised issues in your description of the kind of help you solicit.

(1)Unless the mach number is less than one (0.5 or less I am not very sure of the limits) I would use the coupled implicit solver instead of the segregated solver. The coupled solver is the solver that used to be called Rampant which was used to solve compressible flow problems.I would start with a low Courant number say 0.1. AS soon as you gain confidence that the solution is stable I would gradually increase it to 5 or higher. The realizable turbulence is not a bad model for jet calculations.

(2) If your jet is say parallel to one of the axes, hex mesh may be superior to the tet given the same discretization scheme. The hex mesh will have the tendency to minimize numerical diffusion into your solution. But if your jet is coming out of its source at an angle, It probable does not matter which mesh you use (tets or hexes or hybrid).

(3) I have captured shocks (expansion waves etc.)

(4) and (5) may be related to the fact that you may be using the wrong solver. See comments under (1)

(6) This is a very important point. I usually monitor the residuals as well as observe the changes in magnitudes of several other variables (say pressure, velocity, temperature etc) in different location in my domain. Your residuals may remain constant while the dependent variables you are monitoring change. So constancy in your residuals does not mean that your solution has converged. In addition to changing the under-relaxation factor, careful grid adaption can be of help if your residuals stay constant.

(7)Continually rising residuals may be sign that you have an illposed problem (wrong boundary conditions etc), your grid density is inappropriate or you chose the wrong solver (segregated versus coupled). There may be some other reasons.

(8)Spikes in the residuals may exist for different reasons.I have seen them sometimes because my starting solution (patch) is inappropriate. By getting a better starting solution the spikes disappear. Sometimes I may start with a transient solution approach with small time steps so that I can get a good starting solution for my steady state calculations.

[karthick]

Hi friends!
i wud lik 2 model a underexpanded jets from a nozzle of 10mm diameter. i have generated contours for mach no 0.5--the results are good.
But i wud like 2 model a sonic underexpanded jet. i am using a rectangular control volume of 1500 vs 100 mm

i splitted the bottom left edge of the rectangle. (i.e) from the base i created a point at 5 mm which is the exit of the nozzle. i specified this as pressure inlet in boundary condition----i got good results,


now i wud like to model a underexpanded jet---wat pressure ratio should i start with. what schemes should i specify?

i would like to capture the expansion waves in the jet exit...


pls do help me!!!!!!!!!!!!!!!!!!!