The only thing I can think of that can cause this behavior sometimes is using "Automatic" solver selection
The problem may be starting out with Direct Sparse solver and then if you are running out of RAM or some other cause, switching to FFEPlus solver for instance
You can confirm if this is the problem by opening the OUT file and checking for solver type. See if that's the same as the solver it chooses when you begin to re-run the analysis
Typically the "Automatic" solver selection is pretty great but if that turns out to be the problem you may want to specify solver for those types of analysis.
Thanks for the suggestion Ryan but no, I have my studies always set for Sparse.
I am led to believe that Sparse is the better solver when dealing with assemblies involving materials of different stiffness and non-pen contacts using friction.
These are models I used from a year or so ago with no problems.
I have never seen this behaviour before 2015.
I have copied the models to the server for a co-worker to try and both times his has not repeated, but since it doesn't always do it for me either this is no certainty that it is just my SW.
Your suggestion about RAM might be worth looking into as I only have 32Gig installed.
It has since occurred on a brand new built from the ground up model so I am at a loss as to why it does it. Maybe time to ask the boss for a new computer
Thankyou very much for your reply, much appreciated.
I do have bolts defined, most of my studies do.
It is restarting after completion of the whole study, sometimes it will happen (eg 80 minutes total) sometimes it wont (eg 40 minutes).
When it does happen I can open the OUT file in notepad and it will say the study started 40 minutes after it actually did and only say it ran 40 minutes when it obviously ran for 80.
This is the type of behaviour I would expect if I was using p or h adaptive and I had iteration set to 2 because it would run twice but only return the run time of the last run, if I had it set to 5 it would take 200mins but still only say 40mins in the solver messages.
I did try p and h adaptive not long ago (I think after I had some tech advice saying it would help refine the mesh where it needed it) but it didn't seem to be helping so I went back to manually refining mesh in problem areas, and have since read that it doesn't work so well on assemblies anyway.
When it does restart if I am quick enough (Ie before it has a chance to start recalculating anything) I can stop it and it will have the same results as if it only ran once.
I have been doing studies like this for several years now but have only had the problem in recent months, I think since I "upgraded" to 2015 or tried p or h adaptive so it could be due to either of those.
The studies don't have any more DOF, elements or nodes than they usually do. I have 32 Gig of RAM and its not getting close to maxing that out.
CPU usage was around 50% (quad core with hyper-threading) in one instance but it seems to do everything except remember it finished and should stop and record the results.
I usually record a log of my studies and record the start time and the run time (so I can show my Boss I actually do stuff around here, and that FEA doesn't take only 5 minutes) so I would like to think if had happened on earlier versions I would have noticed that in my log.
It also happens on my computer at home, so it should not be a hardware issue.
Thanks again for your response, if you think it has something to do with bolts I am keen to hear your theory.
edit: see below study is at 32 mins 54% complete overall, several seconds later it has shot up to 100% (where normally it would end and show results) then is back to 0% complete overall and 7 seconds into a new study. note in the performance history the same bump to show time lapse. note it says sparse before and after. DOF = 367,000 but I have had studies with 2,000,000 DOF run without incident (just a long time).
Thanks for the detailed response. The reason for me asking about bolt connectors is that SOLIDWORKS does need to run a portion of the study twice when there are bolts with pre-load. The two cycles of the solver with bolt connectors are related to the pre-tension load applied on the joints that make up the end points of the bolt. If there is no pre-tension load on bolt, the solver will only solve the equations once. However, when pre-load is applied, two cycles are necessary to get an accurate simulation. The first cycle is to get the exact clamping force applied on the joints. In this cycle, the only loading applied to the model is the pre-tension on the bolts (using the thermal contraction effect). There is no other external loading. We solve the equation and get the contraction of the joint. Then, the solver is able to modify the applied pre-load so that the clamping force on the joint is equal to the pre-load input. In the second cycle, both the modified pre-load and external loads are applied to get final results for the model.
If it is restarting completely then I would suspect it is not the bolts. If it is happening on both of your machines it may be related to the registry if you have the same setting on one machine as the other. Could I suggest a registry reset so that all the user settings are removed and any corruptions or issues in the registry are rectified?
If you type regedit in the start menu it should find a programs called regedit.exe, open this. Don't delete anything in here. Browse to HKEY_CURRENT_USER\Software\SolidWorks, right click the SolidWorks folder and rename it to "SolidWorks Old". This will force SOLIDWORKS to rebuild the registry when you next open the software. Reopen SOLIDWORKS and re-run the study to see if this helps. If you want to get the setting back to how they were you can rename the registry back to how it was (you will need to rename the new SolidWorks folder to something else first)
Are you able to attach the model or the .out file for me to test with? If it is a generic issue it should run twice on my machine too.
You can have a whole Rx file if I can figure out how to send it to you (14 MB), I dumbed the model down enough it only takes 4 minutes to run now so was able to capture the whole thing on Rx, which I have sent to my VAR but you can have a go if you like.
edit: FYI I tried the registry thing (thought I had tried it but that was for something else) and believe it or not it has made it worse, where it had been a bit intermittent now it has happened every time I have run it, even after restoring the previous registry. I thought I was onto something where if I had Compute Free Body Forces option checked it would run twice and if it wasn't checked it would run once, but now it runs twice regardless.
Haven't noticed any studies running twice since February which coincidentally is about the same time we went over to SolidWorks 2016.