Fluid Forces and Convergence

[cadmonkey]

G'day all,

This is basically 3 questions...

1)
I was wondering, is it possible to output forces at each iteration, opposed to at the end of the simulation or at the end of each timestep?

I'm using the "calculate fluidic forces" solver or whatever it is, using this...

Solver 2
Exec Solver = Always
Equation = Fluidic Force
Procedure ="FluidicForce" "ForceCompute"
Calculate Viscous Force = True
End

Solver 3
Procedure = "SaveData" "SaveScalars"
Exec Solver = Always
Filename = forces.dat
Variable 1 = Time
End

So I've made the "Exec Solver" command "always", but its still only outputting after each timestep for a transient case or at the end of simulation in the case of a steady state. Can I make it do it on each iteration instead?

2)
Also, I've tried getting forces using the "calculate loads" in the navier-stokes solver (rather than the separate fluidic force solver), but it only gives me a single number instead of x and y components (I've tried Variable 1 = Flow Solution Load 1 and Variable 2 = Flow Solution Load 2, still only 1 number gets output, which is the boundary sum). The number it gives is also a bit different to the resultant force found using the fluidic force solver. Is there a way to get it to give me both x and y components?

3)
Lastly, is there a way to get ElmerSolver to output fluid forces to the convergence graph so I can check out how well the forces are converging as it iterates?

Thanks heaps for your time,
Daniel

[raback]

Hi Daniel

1) Perhaps you just have one physical solver? To see the convergence with the nonlinear iteration you should set the "Nonlinear System Max Iterations" of the physical solver to one, and ensure the convergence by the "Steady State Max Iterations" loop. Then you should also see the output after each iteration.

2) I cannot immediately see why you would only get one value. Does the SaveScalars solver complain anything? The results are expected to be different as different integrations points are used. The system using the matrix residual is more accurate if it is applicable. It cannot be that, for example when a node has contributions from two or more boundaries on its force.

3) Unfortunately not. You could perhaps follow the file written by SaveScalars by LiveGraph, or something similar. Actually, it would not be a bad idea to let the user define that some operator computed by SaveScalars would be outputted so that it is recognized as a value to follow in the convergence monitor.

-Peter