topic Re: Transient Analysis- Time Step Problem in CFD Forum

Transient Analysis- Time Step Problem

kasruko — Fri, 09 Jun 2017 12:13:44 GMT

Hello everybody!

I was running steady state simulation for a fireplace analysis.I wanna learn how velocity and temperature changes, especially.

While i set iteration #800 and Intelligent Solution Control is on, iteration stops at #739. So, i suppose that the results are settled and solution is completed successfully. (If i am wrong, please warn me!!!)

After then, i change the solution mode to transient solution from steady state. I wanna see what happens on fireplace during first 5 hours. No need to see results so frequently, every 15 mins is appropriate for me.

So, i think i should arrange the settings as:
Time step size=0.5
Stop Time=???
Inner Iterations=3
Time Steps to run=36000

But i have a problem! In this case, the total run time will be=0.5*36000=18000 seconds which is equal to 5 hours. [if i am wrong, please warn me 😃 ]

But what should i enter to stop time? Actually it is 18000 seconds...

Besides, i dont know how to see the results for every 15 mins...

I am so confused.

If someone can help me, i will be glad!

Best in advance!!

Re: Transient Analysis- Time Step Problem

Jon.Wilde — Fri, 09 Jun 2017 12:53:08 GMT

Hey,

I can suggest a few things:

Turn off ISC when you run the transient section, this keeps CFD fixed to your prescribed time steps
Set -1 for stop time (to have it ignored), or what might be simpler is set 1000000 to the 'time steps to run' and the 'stop time' you actually want (18,000). CFD will stop at whichever it reaches first
1. The nice thing about this is that you can tweak the time step size to have it run smoother if necessary and not need to change these other values
Set save intervals every 900s, this would give you a result every 15 minutes

Hope that helps,

Jon

Re: Transient Analysis- Time Step Problem

kasruko — Mon, 12 Jun 2017 07:17:17 GMT

Hey @Jon.Wilde,

Thank you for suggestion! Now the simulation is running by following your advice!!

The thing that makes me confused: What do you mean with "Tweaking the time step size to have it run smoother"? Can you explain it please?

Thanks!

Re: Transient Analysis- Time Step Problem

Jon.Wilde — Tue, 13 Jun 2017 07:17:29 GMT

What I mean is, often you need to run a time step sensitivity study as well as a mesh one. To understand if the time step is effecting your results.

For example, if you ran 0.1s rather than 0.5s, the results might improve as you are capturing more data with finer accuracy. Saying that, there is a point that this is a waste of time, as the results don't change with a time step reduction.

So you might want to experiment, see if the results change.

This also works the other way, if you increase the time step, at some point it will simply diverge as you are jumping between two points in time and CFD can't connect the dots.

Check out the mesh sensitivity article I wrote, the process is the same 🙂

Does that make sense? I had planned to write this in a line or two and ended up with an essay.

Re: Transient Analysis- Time Step Problem

kasruko — Wed, 14 Jun 2017 07:03:13 GMT

@Jon.Wilde!

You are totally right! The simulation has been running since friday and it is still on 3925 seconds. There are 14000 seconds more! It will take a month to get the results. Furthermore, maybe i will need to change some data and run it again, after getting the results from this simulation. So, my study will never end with this speed!!

My inputs are like that:

Time step size=0.5
Stop Time=18000
Inner Iterations=3
Time Steps to run=1000000

To get the results quickly, i must change the time step size. I dont think the results will change so significantly, so i can increase time step. Is there any maximum limit of time step?

It would be great getting results in maximum two days 🙂

Re: Transient Analysis- Time Step Problem

Jon.Wilde — Wed, 14 Jun 2017 08:17:31 GMT

Everyone always seems so surprised when I'm right 😉

To answer your question, there is no strict rule, you might need to experiment a little. Try 10s, 30s, 60s - do the results stay the same and do you avoid divergence?

If so, go for it. Unfortunately, you simply need to test out some figures and see... Once you know what works though, you have more experience for the next time 🙂

Thanks!

Jon