[ansys_matt]

I ran a simulation with exporting data set to every iteration to see if it looked correct the first 20 iterations. Everything ran just fine. So I went back in and set the export to only write a file every 10 iterations since I am going to run this forever. I also reinitialized the solver. When I ran it again, this time it failed to converge on the 12th iteration!?

How is that possible? I thought that reinitializing the solution simply took it back to where it was before the last solution began. So how could it be that fluent finds a different (divergent) solution when I changed nothing except how often it exports the data??

I am not using adaptive mesh or time stepping.

[RodriguezFatz]

You must have changed something. Are you sure that you used the same initialization?

[pakk]

Originally Posted by ansys_matt

I thought that reinitializing the solution simply took it back to where it was before the last solution began.

No, reinitializing means that Fluent throws away the solution, and starts with an almost empty solution.
In most cases, it is the same, but now always.

If you initialize a case (iteration number zero), do a simulation for 100 iterations, and then reinitialize, Fluent will be in the same state as in iteration number zero, which happens to be where it was before the last solution began. (if you do initialization in the same way)

If you initialize a case (iteration number zero), do a simulation for 100 iterations, change some parameters and do a simulation for 50 iterations, you end up in iteration 150. If you then reinitialize, Fluent will go back to iteration 0, not to iteration 100, so you are not back at where you were before the last solution began.

(Excuse me if you already understood this, but I thought it'd be best to explain it anyway.)

[ansys_matt]

If you initialize a case (iteration number zero), do a simulation for 100 iterations, and then reinitialize, Fluent will be in the same state as in iteration number zero, which happens to be where it was before the last solution began. (if you do initialization in the same way)

That is what I thought too. Apparently changing the export interval can change the solution outcome. Strange.

[sescobar]

It is very unlikely that export interval influences the solution. Something changed in your model set up or initial condition.
Changing between double and single precision may also lead to different solutions for the same case and initial condition. I don't know if this us your case.