Sign in to follow this  
Followers 0
vztkiers

AdH steady state strategy

3 posts in this topic

Hi !

I am relatively new to SMS and to AdH. I model a complex meandering river and overland flow situation. My model is 25 km long and combines parallel rivers in an extreme flood event (steady state). The floods cover the whole width of the valley (3.5 km) in many sections.

I believe that I have been able to create a reasonable good mesh.

I run a steady state simulation, I am into the first couple of hours of my calculation with these BC settings:

IP MIT 200
IP NIT 1
IP NTL 0.01
IP ITL 0.01

TC STD 0.1 10.0


I allow for mesh refinement, for my riverbeds are really steep. This means that my mesh has quality issues of the “Maximum Slope” kind. Other issues I corrected, and I hope the mesh refinements by AdH will refine and handle these slope issues. I actually started with 290500 nodes, the model now runs with 320300 nodes.
I set the IP NIT to 1, for I had issues with extreme and sudden time step reduction (factor 10000) during non-linear iteration steps. Setting the IP NIT to 1 solved this for the moment.
AdH is calculating currently and will calculate for a while to come.

I have basically identified three parameters to change in my BC file to make the calculation converge to the steady state solution.

  1. I will at one point have to reduce the NTL and ITL (AdH will state it has converged to a solution).
  2. I will at one point allow for non-linear iterations, moving to IP NIT 6
  3. I can allow for bigger maximum time steps in the TD STD card

How should I prioritize between these parameters? Will anyone help me by sharing his or her experiences?

My dilemma is that If I allow for a bigger time step, my model might fail to converge, while if I decrease NTL and ITL, I do produce betters solutions, but that seems to be at the cost of calculation time.
I imagine my most optimal calculation allows for big errors (say 0.01) and big time steps, for at the moment there is still a lot of water in my model that is not on the right location, not on the right water level, but I haven’t yet figured out how to set my calculation cards just for such a calculation.
Especially towards the end of my calculations I am under the impression that I spend a lot on calculation time by reducing NTL and ITL, simply to keep AdH running. I need to reduce these cards, for AdH will state it has reached a valid solution when it errors stay within the set tolerance, but that does not mean it has reached the steady state solution.

Anyone with ideas? Useful experiences you are willing to share?

Thanks in advance!

Edited by vztkiers

Share this post


Link to post
Share on other sites

One option is to run the model in Unsteady mode, but for a long enough period of time that it becomes stable. I find this works better than running in Steady as it goes a lot quicker.

I also turn off mesh adaption as it slows everything down considerably and just make sure I have good mesh.

I should also point out that one of the most important things for Steady models is having a really good initial water surface elevation. This generally means running the model with higher tolerances for a while until the water surface is not changing any more and use that as the initial condition for a run with lower tolerances. Another thing I have done is use a separate model (SRH-2D, RiverFlow2D, or another) to come up with the initial condition.

Share this post


Link to post
Share on other sites

Thanks, for these suggestions. Running in Unsteady Mode seemed to offer what I was trying to accomplish: running with relative high tolerances and progressing with an acceptable time step.

But now a couple of days later I can add that my initial conditions are, with only some hours into the solution, simply not good enough for Unsteady Mode. The model runs in Unsteady Mode well up to a certain period in time when the model no longer can converge, and my non-linear iterations use smaller and smaller time steps in an attempt to converge. These calculations fail at these moments. Of course I review the elements that the cause stability issues, but most of the time I have to conclude that there are no real issues with these elements, instead the hydraulic conditions in that element at that moment are problematic, most likely due to the not-so-good initial conditions.

Actually running a steady calculation with my the IP NIT to 1 is at the moment a more robust approach, even though it is a slow progress.

(Just to give an update on my own question)

Edited by vztkiers

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now
Sign in to follow this  
Followers 0