I am having immense difficulty solving an unsteady gas dynamics problem. It is a pressure vessel filling process. The PV is evacuated at t = 0. It is then filled. The flow is chocked at the inlet. I need to find P(t), T(t). I am using TASCflow 2.7.3 (comemrical CFD code) whose documenatation is pathetic.

I desperately need tips/hints/ideas because I have almost given up.......IF YOUR HELP IS USED AND WORKS, Then I will give you writing a letter that says I will pay you $100 dollars and you WILL recive it. I dont go back on my word.

Please e-mail: sherards@hotmail.com or call: 416-978-6443 or 416-536-0178 in Toronto, canada.

Thank you.

(1). If the flow is chocked at the inlet. Then the mass flow rate is fixed with the inlet total pressure unchanged. (2). The size of the PV will give you the volume, the total mass dumped into the PV at any time will give you the pressure and the temperature I hope ( from the original temperature and pressure). Thermodynamically, the same amount of mass originally at the Po,To state is now expanded adiabatically to a new state which occupies a much larger volume(the size of the PV). It is similar to the car engine during the expansion process.(3). As long as the inlet nozzle remains chocked,Do you still need a CFD code? It's based on my intuition. I could be wrong. (check out an engineering thermodynamics text book)

John is right, if the inlet is supersonic, then the information from inside the vessel cannot go upstream accross the shock and therefore, as long as the inlet is supersonic, the inlet values are not affected by the state and evolution of the flow inside the vessel. When a flow is supersonic, all the information (characteristics of the flow) is travelling downstream and the inlet does not 'see' or 'feel' the flow downstream (from the shock) and the vessel.

The density and pressure will increase with time and the temperature might increase as well depending on the assumption on the energy processes. Therefore, at a given time the temperature might be such that the sound speed is so large that the inlet flow becomes transonic and then subsonic. Then the picture changes, because now the inlet flow then 'knows' about the gas in the vessel (characteristic going upstream). If you know (from experiments) that the flow at the inlet stays supersonic, then of course you don't worry about that.

I hope this helps.

PG

Isn't a shock tube a chamber with a diaphram that divides it into two parts? In this case, one sub-chamber being at 10 atm, the other at 1.

I don't understand the 'inlet' reference. The gas is already inside the chamber. Indeed, when the diaphram is ruptured, a rarefaction goes into the high pressure part, a shock into the low pressure part.

I don't know who wrote that famous text, 'any compressible fluids book,' but it should be checked for the solution to this problem.

Sorry, but I can't help with the poster's problems with his CFD code. : (

The inlet will remain chocked as long as the pressure in the PV is lower than the critical point. As the pressure keeps increasing, the inlet will become unchocked. So, there will be two stages to fill the PV: Stage One, chocked inlet. If the inlet nozzle is smooth so that the flow can be assumed isentopicly come to a rest in the PV, you don't need to run CFD code, simply follow John's instruction until the pressure reaches a point when the inlet becomes unchocked. However, if there is a sudden change in area or others that make the flow complicated after the inlet nozzle (strong reverse flow, separation, shock waves etc.), and the pressure in the PV is strongly ununiform, you may need to run CFD code. In Stage One, as the inlet is chocked, you may impose constant mass flow inlet boundary condition. Be caution: you need to continuosly minitor the pressure after the inlet nozzle because when the pressure reaches to a certion point (depend on inlet total conditions), the nozzle will become unchocked so that the mass flow will vary with time. This is the secone stage, then you may apply constant total pressure inlet condition in your CFD model.

Again, if your geometry is simple and the flow is uniform, you can solve both stages by one-dimensional analysis, and no CFD code is realy necessary.

(1). Shock tube problem is a standard test case for transient, compressible flow code to test the handling of the shock capturing capability of the numerical scheme. (2). The problem has analytical solutions and can be solved by transient method of characteristics. Some gasdynamics books cover the subject in great depth. (3). The shock tube applications includes the study of chemical reactions, simulation of hypersonic flows, etc. (4). The finite difference solution of the transient equation with shock is quite straight forward. There were many papers published in dealing with such 1-D traisient problem. I have solved such 1-D shock tube problem in finite-difference formulation using BASIC language on an early 1MegHz micro-computer. (5). It is a good idea to visit some gas dynamics book for this type of 1-D shock tube problem.

Sherard,

I sympathize with you. Unfortunately, I have heard this story about Tascflow before. It is pressure based, what you need is a density based formulation. It wont do the job.

And yes, Tascflow does have really crappy documentaion, some of the worst I have seen....Almost like it was written by students.

YZ

As the TASCflow guys are located in Waterloo which is just one hour away from Toronto I wonder why you do not just go there and discuss the problem with these guys. They should be able to solve it.

(1) A good question. (2). When CFD software vendors are promoting their products at cfd-online, I think they also need to provide a hotline number so that readers of this forum can send their questions directly for help. (3). It is hard for general public to know what's in the code and how to adjust the parameters. (4). I have just visited the AEA Technology,CFX web site, and I was able to find the CFX-TASCflow code under the category of For the Rotating Machinery Applications. Whether this CFX-TASCflow is identical to the other older version of TASCflow is unknown. (5). My feeling is that if the new CFD-TASCflow is being promoted as a tool for rotating machinery application, then the user's problem is appraently not within the main area of the code application. Also, from the published references available at the website, there was a long list of radial flow turbines and pumps applications. (6). As a user of commercial CFD codes (CFX-TASCflow included), my suggestion to other users or potential users is : Don't use the code if there is no working samples of the code for the problem you are interested in. But if you are already a CFD expert, then I guess you don't need my suggestions at all.

Hi Sherard,

If you are still having difficulties, I would like to see some more of the details ie. the prm, bcf, grd, initial rso files and a more detailed description of the problem geometry, etc.

As for the $100 well, charity is one thing, the real cost for anything even approacing a minimum consulting fee is 1-2 orders of mangnitude above this ....so we will call this charity and this will be a freeby but for the good of broader education!

Why are you using 2.7.3, 2.8 has been out for quite some time and includes a 2nd order in time transient scheme!

I (and a lot of other REALLY GOOD CFD GUYS) think the documentation is in fact excellent!

"It is a poor tradesman who blames his tools"......I don't know the author....sorry!

Good Luck....................................Duane

Hi John,

1. CFX-TASCflow is the new version of TASCflow, which was developed by ASC, headed in Waterloo Ontario but also in Germany and the states. AEA bought them around 96 and the CFX-TASCflow is pretty much based on the original as opposed to CFX which AEA also develops and markets.

2. Yes, they do talk a lot about turbomachines BUT that is by no means the only thing it is good at!

3. TASCflow has an INCREDIBLY robust coupled algebraic multigrid solver that simply is the best I have ever seen! In fact, in my experience, if the solver crashes.....the user has done something wrong. None of this pissing around with inter-equation relaxation factors like the segregated guys and never really knowing if you really have optimal settings! Typically you can get a 10^{4} residual reduction in a few hundred iterations. It does use a lot more memory which some people used to complain about but these guys knew 10 years ago that would not be a problem in the future (today)!

4. The accuracy of the Skewed Upwind differencing with both a first order (mass weighted) and second order (linear profile) advection scheme treatment with the option of corrections for diffusive and source term influences (PAC) IS AWESOME! And I mean for real industrial problems not for the kiddy stuff that lots of people play with for decades in Universites, etc.

Good Luck!.................................Duane Baker

(1). Thank you for the technical information about the CFX-TASCflow.(I think this is how they identify the code today.) (2). I think you are right about all the items you mentioned. So, I am not going to repeat it here. (3). Recently, I have start checking out the code for turbomachinery applications, especially in the area of turbulent flow separations. It is very important in the prediction of the loss and the efficiency. I can't tell you my finding because it is still preliminary. (4). All I can say is the convergence rate is consistent with this type of formulation. ( it is always much faster than the density-based compressible flow codes.) So, if one is interested in the low speed flow solution, it is very important to use the pressure-based codes for fast convergence. (5). The large memory requirement could pose a problem, because one can easily run into 250MEG to 500MEG RAM for a single blade row calculation. While your statements are true from the numerical point of view, I am looking at the over-all system requirement and the solution accuracy of turbulent, separated flows in turbomachinery. In other words, we are trying to find out the limitations of the code. I can only say that, a high quality mesh and a good turbulence model are still the two important issues in CFD or CFD codes.