Whenever I have to commit a delivery date for thermal simulation results, it is already a second nature for me to state the date with "if all goes smoothly".
Those asking for a date almost always want the results sooner, and somehow believe that as long as they asked or a shorter delivery date, the thermal simulation will magically be completed faster.
This keeps happening regardless of how many times I state a realistic turnaround time, and that is with me already committing to an optimistic time frame assuming everything goes smoothly, without any major hiccup.
Of course I wish that everything would go smoothly, there would be no unexpected issue, and my thermal model and simulation would be perfect and needing only run once. I sincerely hope so, honestly I do, who wouldn't want smooth sailing and least amount of work?
But reality is that thermal simulation is seldom issue-free, things don't always go smoothly, and regardless of how many times you pushed for results, the time needed for simulation doesn't magically shorten.
Usually when a divergence occurs in simulation, I can resolve it within two debug revisions, often just needing one. And if I manage that, I can usually still meet my committed optimistic time frame.
Usually, not always. But people just ignore the "if all goes smoothly" part, demand results and pile on the pressure when things go badly.
Took me thirteen or fourteen debug revisions to get a model to run without divergence issue in my recent project, can really do without the constant demand for results while I was trying my best to make it work.
The original model has no problem whatsoever, it's the computing resource that is lacking. Ran out of memory instead of divergence issue with the original model, forcing me to reduce the mesh size, which led to the divergence issue.
It's liberating when I got a temporary access to a powerful computer to run my original model and it ran without a problem, giving the constantly-hounded for, heavily-demanded results. My work has no problem, thank you very much, just the computing resource I have to work with is not up to par. Give me the right tool!
Since it's only a temporary access to the super-computer, I have to lower the mesh size of my original model for it to work on the computing resource I normally have. Basically I have to make my perfectly working model into a lower quality coarser mesh version with lost of accuracy and lots of effort for the 'crappyfication' process.
Anyway, ya, been stressed by this ordeal recently, because things simply don't always go smoothly.
Posts
