Hi, all

I am simulating H2 sonic jet using LES (1300m/s). I experienced problems, obtaining very large mach number. The simulation terminated after a few timesteps. I suspect it may be due to the selction of timestep. As I am aware time-step is restricted by the smallest element size. I found it is rather annoying that CFX doesn't offer automatic timestep in transient simulations (even in RANS). Any suggestions or comments are very much appreciated!

Jianping

Hi,

Doing a sensible LES on a jet of that velocity would be very difficult. The turbulence lengths and time scales would be microscopic (ie huge mesh required with tiny timesteps). Are you sure LES is a sensible technique on this type of model. RANS sounds much more sensible.

Automatic timesteps is planned to be coming with CFX 5.8, coming out later this year. In the meantime you can write your own CEL expression to control the timestep.

Glenn Horrocks

Hi,

Thanks for replying. Is there a simple way to work out the time-step using the CEL expression?

The smallest element I used is about 3mm in length. And the inflow velocity is 1300 m/s. The time step I used is 1e-006. Is this a sensible selection of mesh and timestep. With the above setting, there is still the same problem, saying the newton pressure iteration number (100) is too small for convergence. If I change this, would it be helpful?

Also, I tried to reduce the speed to 600 m/s. Unfortunately, it didn't help either (just ran a bit more iterations then shot out). I still want to have a go with the LES, of course with much lower jet velocity.

In the meantime, I tried RANS, with a timestep of 1e-005. So far it works fine.

Thanks for your help and suggestions

Jianping

Hi,

When deciding the mesh length scale for your simulation you need to consider the length scales of the turbulence. You cannot tell based on the jet velocity alone. If you have a RANS simulation you trust you can work out estimated turbulence length scales from that model and it should give you an idea of the sort of mesh size you should aim at. The turbulence length scale is a function of k and epsilon/omega.

It sounds like you are new to LES modelling so I suggest you do some reading before proceeding down that track. A good textbook to have a look at is "Turbulence modelling for CFD" by Wilcow (I think this is right). It has a good discussion on turbulence in general, and some stuff on LES modelling.

Glenn Horrocks

or do google for prof. lars davidson , from his page you can download les lecture note (or note on turbulence) ..at least they helped�@ me a lot in understanding turbulece ...