sudden opening of gates
 

Hi guys, I´m trying to run a problem in which, in a simplistic way, a small gate at the inlet of an horizontal channel(and in the outlet of a tank with higher level of liquid than in the channel) is opened suddenly. What I expect is a travelling wave but instead an overflow in the very first iteration is what I get. I´ve checked carefully the initial and the boundary conditions and I can´t see anything wrong. The dimesions are tipically of 1 m and the level of the tank is about 10 m. Air and water are the liquids. Whatever suggestions will be wellcome. Thx

Re: sudden opening of gates
 

Hi pi,

If you reckon your physics is OK then I bet it is the timestep size. Try smaller timesteps.

Glenn

Re: sudden opening of gates
 

Hello Glenn, the characteristic time is 0.1 s and for the first ten iterations I use 0.01 s timestep, which should be enough, am I right?. As both the tank and the channel are open to the ambient the initial condition has a discontinuity in the pressure (same as the pressure difference at the two sides of the virtual gate) which should drive the following motion. Initially the level of the tank is 10 m higher than in the channel and the mesh in the discontinuity region, right where the door would be is about 1 m, being 20 m the height of the channel. With this numbers can you estimate the timestep?. My next move will be refine the mesh in the door area, right in the tank outlet and channel inlet, and will try with a smaller timestep. By the way, the error was in FINMES, fatal overflow in linear solver if it is of any help.

thanks in advance pi

Re: sudden opening of gates
 

Hi pi,

It is a common mistake to work out a characteristic timescale of a simulation, divide it by ten (or some other small number) and expect that to be a reasonable timestep size for the simulation. It does not work.

This goes especially for compressible, multiphase, chemistry (especially especially combustion) and other highly non-linear simulations. For solver stability you need very small timesteps.

It is very difficult to estimate a good timestep before the simulation. Assumng the rest of the simulation is OK, just have to keep making it smaller until it converges. Then you need to do a timestep independance check, and also be careful about using very small timesteps with turbulence models and round-off errors, but that's another story...

Regards, Glenn

Re: sudden opening of gates
 

Dear Glenn,

after keep trying with increasingly smaller timesteps unsuccesfully I set the solver relaxation fluids parameter to 0.75 and linearly exact numerics to true. With this I got a start up but after few iterations a message of insuficient catalogue size shows. Do you know what is it about?

Another question. The ceiling of the tank is open to atmosphere. If I close it and set a wall condition I foresee that will be problems because as the water level goes down an empty space void will be on the top. Can you figure out any roundabout to avoid this (beside to make a small hole in the ceiling of the tank)

have a nice weekend

pi

Re: sudden opening of gates
 

Hi pi,

I would not set the solver relaxation fluids parameter or the linearly exact numerics parameters. Leave them as defaults, it is highly unlikely this is your problem.

Other things: 1) Use upwind differencing, first order time differencing; 2) If your model is turbulent, use the zero equation turbulence model; 3) How small a timestep did you test down to?.

The catalog size sometimes gives you this error. You just have to allocate a larger catalog. I can't remember how to do this as I haven't had to do it for some time. Can somebody show us how to set a bigger catalog size?

It is very difficult to help without knowing the details of your simulation. It would help to give more details of what you are trying to do.

Glenn

Re: sudden opening of gates
 

Hello Glenn,

I´m trying to reproduce a 3D version of the Maxwell problem which can be seen in the CFX presentations. In that simulation it is a 2D of a reservoir with let's say water, that discharges directly to a channel. I don't know if you have seen this video. Anyway, the initial condition it is a 30 m heigth tank connected to a channel with quiet water at 20 m level (all the heigths from the bottom). The channel is 150 m long. To make a picture of it imagine a L rotated 90º CCW. More or less this is the shape of the domain. I try to get waves non parallel to the coast line. The video is 1.4 Mb so if you don´t have it I can send it to you to get a better feeling.

In my case the tank is open to the ambient (which is air). In the start you will see exactly like a imaginary gate is suddenly open through a discontinuity in the pressure field (which comes from the different water level at both sides of the "gate", in the outlet of the tank, at the inlet of the channel, where the L elbow is.

The purpose of the tank is to produce the waves but I´m thinking, at the sight of all these problems, if it´d be easier to produce them by means of a pressure perturbation through one of the boundary condition. On the other hand I´d like to know what the hell happens there and why the run does not work. One never knows when or where you go to find a similar problem and these difficulties gives you experiece to solve it.

Thank you very much for your help

pi

Hello dear

I just saw your problem and learned that you got its solution. I have just started my MS and am working in the same field. It would be very nice of you if you could please help me. Feel free to contact me on

mechovator@yahoo.com

With best regards

Mechovator

Hi,

I am to model 2 parallel gates with not-equal openings in an open channel.
here I have some questions hope someone may assist since I am new to CFD tools:

1. How much is OpenFOAM appropriate for this purpose? I doubt about this when I read lots of work using OpenFOAM in machinery fields but very little in open channel flow problems.

2. If OpenFOAM is not suitable for such a problem, which open source codes is suggested?

Thanks, ali.

Try the OpenFoam forum.

Hi Glenn,

Is the forum you said anywhere else?
I believed this forum concerns OpenFOAM!
If yes, would you send me the link plz.

Thanks,
ali.