# transient problem

 Register Blogs Members List Search Today's Posts Mark Forums Read

 December 4, 2002, 14:14 transient problem #1 susan Guest   Posts: n/a Hi, I am doing simulation on glow discharge. it is transient problem and I want to catch details at some specific instance. The discharge physics made the time step cannot be large,currently I used 10e-9 as the maximum. To make the codes more efficient, I tried to use adaptive time step based on the change of the main physical quantity.But the problem is that every time dt changed, there is some oscilation in the main quantity and to damp the ossicaltion I have to put more stringent convergence creteria, which in turn affected the efficiency of the code. does anyone have some idea to solve this problem? any suggestion is highly appreciated! Thanks a lot!

 December 5, 2002, 00:49 Re: transient problem #2 versi Guest   Posts: n/a My experience: firstly, compute dt according to CFL number, then dt=min(1.2*dt_prev,dt) dt=max(0.8*dt_prev,dt), where dt_prev is previous dt

 December 5, 2002, 03:34 Re: transient problem #3 Fred Souliez Guest   Posts: n/a You will have to trade efficiency for stability. I think you need to introduce some relaxation factor in your time step, i.e. use a weighted average of the latest time step with that of the previous step. I don't know of many other ways around that problem. Also, the efficiency of your code will not scale linearly with that relaxation parameter, so that you really need to use it cautiously. Good luck!

 December 5, 2002, 12:06 Re: transient problem #4 susan Guest   Posts: n/a Hi, versi: Thank you for your advice. But I am confuesd with CFL number. If i use semi-implicit scheme, do I still need to abide by the stability requirement of CFL number. Thanks a lot!