Check the Results from the Study Currently Being Solved

[martin_madaj]

Hello guys,

good job with introducing explicit solver to Fusion 360! (However, it would be even better to have it in Nastran In-CAD and/or Simulation Mechanical.)
I have a question: how can I review the results from the explicit simulation that is currently being solved on cloud? The explicit simulations can be quite complex and time consuming, so I'd like to see the rolling results to check whether everything goes right.

Thanks, Martin Madaj.

[Andrew.Sartorelli]

Hi @martin_madaj,

Nice to see you are exploring the simulation capabilities in Fusion 360. You can get updates from an Event Simulation by using the "Solve Status Information Interval" or Heartbeat.  I believe this should send results back at periodic interval.

Regards,

Andrew

[martin_madaj]

Hi Andrew,

nice to see you're everywhere where is the slightest trace of simulation. 🙂 Your answer sounds logical, I'll see, if it works.

Thank you! Martin M.

[Andrew.Sartorelli]

Hi Martin,

Wherever there is Autodesk Nastran, you will find me 🙂

Keep in mind that Event Simulation, Non-linear Static, and Shape Optimization are still in Tech Preview. If you find anything you thinks needs improving or usability issues definitely let us know!

Andrew

[jeremy.wiesner]

Hello Martin,

Unfortunately at the present moment there is no way to retrieve results or a status update from an in-progress cloud solve. The Solve Status Information Interval option for Event Simulation controls how often data is written to the solver output log, but that log is currently only being returned to you when the solve completes. You can view this output log by right clicking "Results" in the browser tree and then "Solver Data". When the dialog opens, go to the pull down menu at the top of the window and select "Solver Output".

Adding the ability to retrieve intermediate results and in-progress solve status is right at the top of our priority list for future updates, so stay tuned. As you continue to try out the Tech Preview study types, we would love to know if you have any ideas for improvement so please keep us updated!

[martin_madaj]

Hi Jeremy,

thank you for the information. I assume (hopefully) you are going to add some more details about the simulation jobs, like Total CPU Time and Wall Clock Time, simulation time estimation or current time step increment. I think it would be useful to know how long did the job took to solve or whether the time step is not decreasing too much.

Best regards,
Martin M.

[jeremy.wiesner]

The heartbeat that is written to the Event Simulation log file includes the following items:

• Increment Number
• Total Time
• Time Step Size
• Kinetic Energy
• Internal Energy
• Viscous Energy
• External Work
• Total Energy
• Elapsed Wall Clock Time
• Remaining Wall Clock Time

You can view this data now in the "Solver Output" as I described above, but unfortunately it is only available after the solve finishes. In the near future we'll provide a way for this information to get back to you while the solve is running. We also have plans to give you access to a runtime estimate before you even start the solve.

[martin_madaj]

Sorry, my mistake. I didn't notice the drop down list in the Results - Solver Data window which took me to the Solver Output with all the items you described in your last post.

Martin M.

[martin_madaj]

Hello again,

I tried to run a sample simulation of an orthogonal cutting (just to test an element deletion feature). The Job Status window had claimed that the simulation was started succesfuly and that it had been running. So i closed the Fusion. This morning I tried to check the status and/or results but the only result I got is a yellow exclamation symbol telling me that the results are "Sick", Job Status window is empty, Solver Output is empty and the Solver Data window is telling me that there was an error with the simulation because of the time limit of 12 hours for the tech preview. I'm OK with it.

But what is unclear for me is whether the Solver Data message refers to the last simulation job that was scheduled for solve or to the previous one which also ended up with the error message. There is no time stamp in Solver Data Summary and I simply don't know if the simulation failed because of the time limit or wrong initial conditions and when did that happen.

Regards,

Martin M.

[mcmillr]

Hi Martin, unfortunately no data is returned in the case of a timeout, so you would be seeing the results of the previous run. You can see the timestamp in the Solver Data/Solver Output at the very top, it looks like:  "This Analysis was run on: Fri Nov 4 21:46:49 2016".

As Jeremy mentioned, getting back information during the solve is a high priority for us.

[martin_madaj]

Hi Rob,

thanks for the answer. What confused me is that my Solver Data/Solver Output is all empty for that simulation. Nevertheless I understand that you're still gathering feedback and I hope that this helps you. I'll keep trying with simplier examples (I got a few more in reserve).

Best regards, Martin Madaj.

[martin_madaj]

Hello guys,

I have to get back to this topic because I'm playing with some event simulations again. But when it comes to a little bit more complicated setup that has to be debugged, I cannot get over as the only information from the solver in the case of failure is "Error: Solver Error" and "Error: An error occurred while solving the model." How can one debug the model to make it work if there is no information (neither graphics nor textual) to start with? Or do the error messages above relate to any simulation timeout (set on the cloud server) if the solve would took to much time?

Best regards,
Martin M.

[mcmillr]

Hi Martin, sorry you are running into this. If the solve exceeds 12 hours you might get an error like this. If you get this error in under 12 hours then do you see any information in SolverData/SolverOutput? (context menu under Results). If this is blank, then the easiest way for us to help would be to share your model or a similar example and we can take a look and try to explain what is going on.

[martin_madaj]

Hi Rob,

unfortunately, I cannot reproduce the issue on the model (I have no clue why). But I can share with you another version of the model from another branch (see the private message box for the link). The issue with that other model is not exactly the same but it's interesting for me as well. The issue is that I can successfully execute the simulation and it runs without no obvious problems. As it is going to take some time (and I don't know how much... what about getting the solver time estimations back to user? 🙂 ), I close Fusion 360 and get back after a couple of hours or the next day. When I want to display the results, I get the message "The results have not been computed yet. Would you like to compute the results now?" There are no other error messages, empty Solve Details window, empty Solver Output, only the Solver Data window contains "Error: Solver error" and "Error: An error occurred while solving the model." but I'm unsure whether it is not a remnant from some previous simulation run. Well, the simulation looks like it has never been executed.

Best regards,
Martin M.

[mcmillr]

Hi Martin. We are planning on improving the feedback for long running solves in a not too distant release and this should include an estimate of the time remaining for the event sim solve. We have some other projects we are trying to finish up first but we realize this is very important to solve. I will also log an issue to make sure that we do a better job communicating the status of a failed solve. I can see that it might not be apparent if the failure message your see is from an older solve or from the last one you had run.

I looked at your model and the estimated runtime is about 3 hours. You can cut this down to 2.5 hours if you remove the two fillets. The main reason for the long runtime is the duration, which was set to 0.006 seconds. I realize this may not seem like a long time but don't set this time any higher than you absolutely need to. If I remove the fillets and set this time to the default of 0.001 the solve takes about 31 minutes.

Hope this helps!

[martin_madaj]

Hello Rob,

you're right - I removed the fillets and set the time to 0.001 s and the simulation was very fast. Thanks! I used the model with fillets and the time of 0.006 s because it was based on another simulation from a different explicit code (this simulation in Fusion is just a test case for me).
But it is still unclear to me what was the reason that the simulation looked like it has not been executed or even started at all (with no errors, exclamation symbol next to a study name etc.). Was it a timeout? But 3 hours is not that much...? Did you find anything? Or did I misunderstand your answer?

Martin M.

[mcmillr]

Hi Martin, it is not clear why your original solve didn't work, it should have run for 3 hours and then succeeded. It could be that there was some service that went down temporarily and interrupted your job resulting in a failure. We will look into ways to improve the experience for long running jobs whether they succeed or fail.

[martin_madaj]

Hi Rob,

OK, thank you.

Martin M.

[martin_madaj]

Hello,

I have to add to this point that I have two more different simulation models that ran for a few hours and ended up with no results and with no error messages. (Either "The results have not been computed yet..." or "The results are out of date..." if there are any older results present.)

Regards,
Martin M.