20 Replies Latest reply on May 23, 2012 5:54 AM by Yuan Ding

    Flow Simulation crashes and data loss

      SW2009, all service packs.

      Vista 64

      i7 2.6GHz

      12 GB memory

      Brand new machine.  New fresh install.  No antivirus installed.

      The only apps installed prior to SW were Office and Acrobat.

       

      I've been tracking this since SW2007 on XP Pro 32 on three different machine.

       

      In general terms, Flow Simulation (then FloWorks) makes SW unstable and crash-prone.  In XP32 the biggest issue seemed to be memory fragmentation.  As problems get a little complex and you need more cells all bets are off and memory fragmentation rears its ugly head.

       

      Most of the work we do requires the simulation of parametrically driven assemblies.  A design table at the top level assy drives various features in the underlying parts.  For example, if you are simulating a heatsink you might make the fin count, thickness, length, height, baseplate thickness, ducting geometry, inlet and outlet geometry and other features table driven.

       

      We also like to setup batch runs of 10, 20, 30 or 40 cases to run either overnight, over a weekend or during the week.

       

      We were told that the 64 bit system would solve all our problems...hence the time, effort and expense to build a new one.  Almost nothing changed.

       

      This is what we've seen (not an exhaustive list):

       

      Memory management is broken.  If you repeatedly open and close a SW file with, say, a single flow simulation, load the results, and then immediately exit  --In other words, do nothing-- you can watch available memory shrink by tens of megabytes at a time.

       

      After running a batch of just a few cases is is almost guaranteed that SW will become unstable and crash without warning.

       

      SW will allow you to edit your design table, add configurations, clone Flow projects to the new configurations and SAVE everything....and then crash and loose every bit of work you've done.

       

      Batch processing seems to randomize the list of available studies making it just about impossible to manage new runs without a ton of work.

       

      Program hangs repeatedly, with "(not responding)" being shown on the title bar

       

       

      There's more detail, but I'll spare you the agony.  Just this Friday I personally spent three hours setting-up a batch run for the weekend.  I rebuilt every single configuration and SAVED after every rebuild.  This is very time consuming.  Well, it crashed and lost all the work.  Every single configuration was gone.

       

      Before anyone suggests installation or hardware issues, please note that this behaviour has been seen and documented to various degrees across three hardware platforms and two operating systems.  It'd be hard to imagine how all three systems and two OS's could have problems.

       

      My VAR is in over their head with this.  I am not sure that anyone cares to address the issues based on the responses I am getting.  As a sanity check, I'd like to learn if any of you are suffering with similar FloWorks/Flow Simulation problems.

       

      Thanks,

       

      -Martin

        • Re: Flow Simulation crashes and data loss

          Hi,

           

          I am experiencing exactly the same problem, Different hardware from HP-8400 series to build my own power box with I7 and 12 Gb of RAM. Program not responding and abrupt crashs are only few clicks away.

          • Re: Flow Simulation crashes and data loss
            David Paulson

            Martin,

             

            I am glad that you initiated this post.  Your experience with Flow is much more extensive than mine is, having only used the product for the past 3-4 months.  I have experienced very bad crashing with Flow with relatively simple flow models, and I have yet to attempt to batch process anything.  And the crashes are spectacular.....SolidWorks just dsappears from the screen of my computer.  No hang ups, etc.  Everything is just gone.  But automatic save does seem to work for me.  For the most part the data is not lost.

             

            SolidWorks needs to address this issue on a high priority basis.  And I suspect most VAR's have not as much experience with the product than some of the posters on this forum.  Therefore, this forum is in my opinion the best means for SolidWorks to communicate directly with it's Flow users and fix the bugs.  In fact, I don't recall a post to the Flow Simulation forum by a SolidWorks employee.

             

            Flow Simulation should be a flagship product for SolidWorks.  I hope more Flow users will join your post to get the message to SolidWorks.

              • Re: Flow Simulation crashes and data loss

                Clearly there are serious issues that need to be addressed.  The contrast in reliability and usability between SW and SW with FloSim loaded is night and day.

                 

                On Vista 64 SW/FS will go into "Not Responding" mode even while loading or saving a file.  If you wait long enough it pops in and out of "Not Responding" and eventually gets done.  The same is true while switching from one configuration to another.  The interface becomes non-responsive and you get "Not Responding" on the title bar.  Every so often the whole screen goes white.  As in, full white, no menus, icons, nothing.  Sometimes Windows will pop-up a window telling you that SW/FS isn't responding and asks if you want to wait.  Once you get an feeling of what crashing vs. being "temperamental" looks like you can answer that question correctly and continue working.

                 

                Here are two VERY annoying and HUGELY time wasting problems with FS. 

                 

                First, create the model to be simulated.  Keep it simple.  Run a simulation to make sure it all works.  BEFORE running the simulation exit SW/FS and make a complete copy of the directory where the model resides.  The idea is to be able to get back to a "clean" copy with no simulation data whatsoever.  Do this before even creating the FS project.

                 

                Now that it ran, restore the "clean" version of all the files and delete the files you used for the test simulation.

                 

                Now make the geometry more detailed.  If we are talking about a heatsink, for example, give it only three fins and make the 0.25in thick for the test run.  For the real run give it 50 fins 0.025in thick, longer, taller...you get the idea.

                 

                Take the time to parameterize a few of these dimensions.  You want to have control over, say, the fin thickness, spacing and count.

                 

                Create a design table and define 25 different configurations with varying thickness, spacings and counts.

                 

                Now go back to the first config and create a FS project, fully defined.  Rebuild and make sure all is well.  Save all.

                 

                 

                OK, here's the first problem:  Now you want exactly the same study to be applied to all 25 configurations.  How do you do it?  Can you say:  Copy FS project to a range of selected configurations?  Nope.  This is what you have to do:

                 

                You right click on the FS feature tree.

                Choose "Clone Project"

                Select an existing configuration to clone it to.

                Click OK.

                Now you wait while it sparks and sputters and, eventually, switches from the configuration you were on to the configuration you are cloning to.

                On Vista64 you are likely to see the "Not Responding" label come up on the title bar several times.

                Eventually it will tell you that the geometry has changed and that the computational domain has to be redefined.

                You say "OK".

                More sparking and sputtering and "Not Responding"

                Then it tells you that it is a good idea to rebuild

                You say OK.

                More sparking and sputtering and "Not Responding"

                Now you are done.  You have cloned the FS job from configuration 1 to configuration 2

                 

                AND YOU STILL HAVE 23 MORE TO DO THIS WAY!!!!

                 

                On anything but the most trivial models this cloning process takes minutes per config.  I spent three hours one night doing nothing but mindlesly cloning FS projects to be able to run a batch job.

                 

                 

                But it DOES NOT end there.  No.  Every so often SW/FS has a special gift waiting for you.  After all that work you decide that it might be prudent to save before you go on.  In fact, if you are really prudent you save after every third or fourth cloning process...just to be sure.  When you are done, you save a final time....and SW/FS goes into the "spark/sputter/non-responsive" mode before your horrified eyes.  And, it CRASHES.

                 

                After the crash process concludes you restart SW/FS and reload your project only to find out that EVERY SINGLE CONFIGURATION IS GONE and all of your work from the last three hours is gone.  Design table trashed and back to three hours ago.  Brilliant.

                 

                 

                Now, the above doesn't happen every time.  That's what makes it fun.  You never know what SW/FS is planning for you as you work along.

                 

                If you are lucky enough that the above does not happen you can then attempt a batch run.  Here's where the next bit of fun begins.

                 

                You might usually arrange your configurations in a logical sequence.  Maybe config 1 through 8 tests different ideas on fin spacing; 9 through 16 checks out fin height, etc.  So, it stands to reason that you might want to run them in a particular order.

                 

                As you open the batch run dialog you find out that FS has taken your nicely named and ordered simuations and, instead of presenting them to you in the original order --as seen in the feature tree-- it basicly randomizes the list.  So, you might have confit 23 followed by 2 and then 8 followed by 19, etc.  There are no buttons or tools to reorder the stuff other than two up/down buttons that you have to mindlessly click with the mouse to sort the list by hand.  Talk about carpal tunnel.

                 

                Where this becomes an absolute pain is that, even if you take the time to reorder the list, if you exit the batch dialog and come back your reordering is lost.  Sometime one might run just a few cases at a time.  Maybe you do this to check on the merits of a few ideas before devoting more simuation time to similar configurations.  So, you select five or six configs in batch mode and let them run.  When you come back and decide to run another set the list is so jumbled-up that it takes a tremendous amount of work to locate what you want to run and reorganize it.  This is an absolute mess.  It shows no respect whatsoever for the user and purchaser of this software and it clearly indicates that SW must think that our time is free.

                 

                Here's another one:  Every time you open a design table in Excel for editing the design table opens in a small window regardless of whether or not you chose to maximize it the last time you worked on it.  If you are juggling 30, 40 or 50 configurations with dozens of parameters in columns it is an absolute PAIN to have to maximize both Excel and the workbook every time you have to edit the design table.  I use a large 1920 x 1200 pixel monitor.  Why is it that dialogs are the size of a postage stamp?

                 

                Want more?  Why is it that dialogs that have long strings don't resize to use more screen area or remember how you expanded them?  Take the "View external relations" dialog", you can never read the full string.  You have to grab it and slide it wide every time.  And, even then, you have to pull it off the edge of the dialog for it to work.  Here I am with a 24 inch 1920 x 1200 pixel monitor and I can't read half the data in most dialog boxes because these programmers are too lazy to put in a little extra effort to intelligently size their controls and store window data in the registry.  Yes, we write software too and I know exactly what it takes to do this work.  It really isn't that difficult but, if you are lazy and just don't care it is easy to avoid it.

                 

                Anyhow, that's enough for today.  I just got done reinstalling SW/FS three times in a row to try to solve a problem...it is fair to say that I am in a foul mood.

                 

                 

                -Martin

                 

                 

                 

              • Re: Flow Simulation crashes and data loss
                Joe Galliera

                Save your SLDASM (or SLDPRT) before running, this is where all of the Flow Simulation setup is stored.  The only other file that you will care about is the flow data file n.FLD (n being an integer) in the numbered folders.  You can make Flow Sim save your results periodically by setting this up in the Calculation Control Options; these files get saved as r_[interation number].fld.

                 

                Did you know that you can rename any r_[iteration number].fld or r_backup.fld to n.fld (n being an integer same as folder #) and continue your calculation from that iteration number forward?

                 

                If you do lose your FLD file, then you have to start from the beginning.

                  • Re: Flow Simulation crashes and data loss

                    11769 wrote:

                     

                    Save your SLDASM (or SLDPRT) before running, this is where all of the Flow Simulation setup is stored.

                     

                    If you read my post carefully you will discover that saving was a part of it.  I realize it is a long post.

                     

                    So, yes, plenty of "CTRL-S" action will not guarantee that you will not loose all of your work if you get one of these FlowSim crashes.  The ONLY thing that will guarantee your data is to fully EXIT SW and, to be prudent, make a backup copy of the files.  Exiting SW alone should do it though, but I've learned to be very paranoid.

                     

                    -Martin

                      • Re: Flow Simulation crashes and data loss
                        David Paulson

                        Martin,

                         

                        I understand that there is an SPR that addresses the Vista X64 and Flow Simulation issue.  I presume that this SPR is one that you initiated??  If so, do you have any feedback on when a "fix" will be forthcoming??

                          • Re: Flow Simulation crashes and data loss

                            Yes, I believe that there are several SPR's stemming from my conversations with SW.  As frustrating as these issues are to a user, I should at least acknowledge that SW is interested in fixing the problems and understanding the issues.  It took a lot to get their attention, but once I did we got down to business.  I think the last online meeting lasted well over two hours with everyone watching my desktop remotely as I demonstrated a half dozen ways to crash it.

                             

                            Anyhow, no idea on a timeline.  I'm sure they have a lot on their plate.  I'll update this thread if I hear anything back.

                             

                            -Martin

                              • Re: Flow Simulation crashes and data loss
                                Derek Bishop

                                Martin,

                                 

                                I believe that the problems that you are experiencing are the same problems that every serious user of the various Simulation modules experiences. It's not a random isolated thing. It is systemic and is due I believe to SolidWorks philosophy on software development. After using the Simulation software for a couple of years with similar experiences and seeing similar comments from other genuine and competent users I believe the problem is that SolidWorks Simulation is not properly tested and debugged before it gets shipped to users.

                                 

                                I am becoming convinced that SolidWorks relies heavily on users to debug the Software once it gets to Beta stage and maybe before. You may have seen my other posts that highlight the issue. Having done Beta testing I can confirm that at this stage there are still a large number of blatant problems in the software that should have been picked up by SolidWorks. The number of users testing this software is minimal so when you start using it on a project you are often the one who has to identify them. That can be a time consuming, costly and frustrating process. Most users I'd suggest don't get paid for doing this work.

                                 

                                Once you identify a problem, if and when SolidWorks will fix the software is another issue. It depends on their idea of the severity of the problem and I believe their marketing policies ie. the areas that they want to develop.

                                 

                                The software is relatively inexpensive compared to other programs like Ansys. It has potential but is being hamstrung by the obvious problems with which it is plagued. If you are doing anything that is a bit unusual, that you are unlikely to use again and the project will only last for a few months or less then it can be a financial disaster for you and your client.

                                 

                                I make a couple of suggestions. The first is to get hold of decent software that has a good reputation in the industry and that is properly supported by engineers and software developers. Talk to other users and read the forums. Most experienced engineers have an instinctive understanding of what is and is not acceptable. Give the Software company a few of your problems up front and see if the software has the capabilities you require.

                                 

                                The second suggestion is that SolidWorks needs to give better definition on the capabilities and limitations of Simulation and make sure that the software lives up to its claims. It needs to do less things better. Its a question of integrity which we all know is sadly lacking in big companies and particularly those that develop software. If SolidWorks choose to persist in the current vein they owe it to potential customers and users to make their policies on these matters clear.

                                 

                                Kind regards,

                                 

                                Derek Bishop

                                  • Re: Flow Simulation crashes and data loss

                                    I can't say that I disagree with you.  It is really frustrating to come into this as capable and accomplished engineers with deep experience in both very complex software and hardware design and be reduced to "you need to reinstall your OS" or "there's something wrong with your machine" or "we think you have to downgrade to XP 64".

                                     

                                    Save notebook computers, I can't remember the last time I purchased a prebuilt system and let someone else install any software on it.  I was wire-wrapping 8080 and 6502 computer boards before (my guess) a lot of the SW developers were off diapers and have built and used countless PC's since the days of the very first PC clones...I was designing and building parallel-processing computers from scratch (not bolting cards together but designing the actual circuitry as well as writing the OS from scratch) twenty years ago....

                                     

                                    So, in this context, to be told that the solution is to rebuild a machine or buy a store-built machine or downgrade on OS (Vista sucks, so it wouldn't be a downgrade...but that's the terminology MS uses) just rubs me the wrong way.  The plain and simple truth is that there's something seriously wrong with the software and they problably don't know what it is and don't have the clock cycles to devote to figuring it out.  In that environment the easiest thing to do is to put it on the user and shotgun the problem by making them comply to your version of a reality that is just not sustainable.  Computers change and evolve.  I can't continue to fix problems by buying the HP, Dell or whatever overpriced box that SW might choose to call their "safe" configuration.

                                     

                                    I guess I might be out of touch...I just can't figure out how you'd go about writing code that is so hardware sensitive unless you start poking into very low level code.  And, on something like an i7 based motherboard there really isn't that much in the way of variance.  The important bits like the memory controller are fully integrated into the processor...

                                     

                                     

                                    Anyhow, this, so far, has cost us tens of thousands of dollars.  I am not sure where it is going from here.

                                     

                                    -Martin

                                      • Re: Flow Simulation crashes and data loss

                                        Martin,

                                         

                                        I do not think downgrading to XP64 will help.  In the past month I have gotten away from some of the structural and thermal analysis I was doing and was performing much more floworks stuff and all of a sudden I was getting the infamous BSOD (Blue Screen of Death).  All the other simulations packages ran fine under XP64 for the past year.  I don't think I ever got a BSOD until I started heavily using the floworks.  The system I am running on is a A Dell T5400 with a 2.6 GHz Quad core Xeon with 8gig of ECC.  This is one of the most stable workstations I have ever used and there is no reason that Floworks should be crashing at the frequency I am seeing.

                                         

                                        I how SW is listening.

                                         

                                        Best Regards

                                         

                                        Jim C

                                          • Re: Flow Simulation crashes and data loss

                                            I am also having an issue with this. I have tried multiple times to get flo works to solve a problem and have had it crash on me each time after it completes the mesh and just starts to solve.

                                             

                                            I'm using solidworks 2009 x64 SP4.1 on a

                                            Vista ultimate 64 bit

                                            Intel core2 quad 2.66 GHz

                                            8Gb Ram

                                            Nvidia Graphics

                                             

                                            The model I'm (trying) to run is the tutorial file with the flow analysis of a ball valve.

                                • Re: Flow Simulation crashes and data loss
                                  Yuan Ding

                                  Hi Joe:

                                   

                                  I am a new learner of the Solidworks Flow-simulation. And I met a problem these days,

                                  The problem is below:

                                   

                                  The fluid(normal water) flows through a  decompression valve,

                                  we only know the velocity of the inlet(6.1m/s), and the valve's Diameter=0.168m.

                                  My question is how to definite the turbulent flow parameter?

                                   

                                  Thank you very much if you can give me some suggestion!

                                • Re: Flow Simulation crashes and data loss
                                  Bill McEachern

                                  Hi Guys,

                                   

                                  I feel your pain but I gotta tell ya I have been using this code (Flow/Floworks) for over 10 years I don't get any of this sort of grief. I had some issues with interface preformance on an LED light fixture I did back in March but it had over 1200 selected items in the specification tree (not individual ones - probably only about 400 of those) and it was brutally slow but some what understandably. It seems faster now on that front.

                                   

                                  Like today I just did the inlet for an electrostatic precipator - went off without a hitch.

                                   

                                  Maybe you should have a SWX TTM drop by next time they are in the area and see what's up as I don't get those kinds of issues, then again I am not rying to run vast number of analysis in a batch run either. Though I just did like 10 and it work fine though they weren't very big cell count wise.

                                   

                                  I find the flow product far more robust than the FEA products but Flow is far less adventurous on the development side. there is a cae to be made for that and I wouldn't be opposed to that sport of track either for the FEA and motion products. The 2010 beam/shell auto contact is pretty awesone and I did a 200 beam package building with shells for the six walls and it went off without a hitch and not one manual contact specified in the model. That was pretty neat. Yeah I am sure some of the issues reamin but it is getting better.

                                   

                                  I have no clue though why you are getting the level of grief you are seeing though.

                                  Bill

                                    • Re: Flow Simulation crashes and data loss
                                      David Paulson

                                      Bill,

                                       

                                      It would be helpful to those of who are experiencing this issue if you would post your system configuration and operating system.  There is a know SPR issue with Vista x64.  Does anyone know if Windows 7 will fix the issue??

                                        • Re: Flow Simulation crashes and data loss
                                          Bill McEachern

                                          My system is a:

                                          Zeon 3210 (not exactly state of the art)

                                          with 8GB of ram

                                          winXP64 sp2

                                          nVidia FX 550 (had an FX1750 or something but it died an noisy death)

                                           

                                          I hear Windows 7 is suppose to be good but I wouldn't believe any of that till it ships.

                                            • Re: Flow Simulation crashes and data loss
                                              David Paulson
                                              I think that most of the issues are related to Vista x64, and that is the subject matter of the SPR.  But it has been quite some time this SPR has gone unresolved, don't you think??  I am very tempted to instal Windows 7 x64 when it released on October 22.
                                                • Re: Flow Simulation crashes and data loss
                                                  Ian Hogg

                                                  Hi David,

                                                   

                                                  If there's a particular model associated with the SPR that I can try and reproduce the crash with, I have just installed Windows 7 on my machines so I can give it a whirl. Let me know the SPR number.

                                                   

                                                  In general, I had little issue overall with Flow on my 64bit box (Dell T4700 Dual Quad Core, 16Gb RAM). The most prevalent problem I've had was with batching solves and running multiple models in parallel (it would crash every now and again).

                                                   

                                                  Ian

                                                    • Re: Flow Simulation crashes and data loss

                                                      The model that gave me trouble was the valve example in the flow works tutorial.

                                                       

                                                      It would mesh fine, then it would crash (every time) when it tried to solve.

                                                       

                                                      Dan

                                                       

                                                      Vista64

                                                      Intel Core2 Quad 2.66

                                                      8gb Ram

                                                      Nvidia Graphics

                                                        • Re: Flow Simulation crashes and data loss
                                                          Ian Hogg

                                                          Hi Dan,

                                                           

                                                          I did see your post, and before I switched my 64bit machine to Windows 7, I did run that tutorial without any issue. I was using 2009 sp4.1 when I tried it. Since I couldn't reproduce the issue on that model I didn't go any further with it (ie it seemed to be more machine specific). It's one of the challenges when you get intermittent issues to diagnose the root cause. If an SPR exists with a reproducible error, then it's easy for me to test it and see if it runs or not.

                                                           

                                                          This is where it's valuable to use the SW Rx tool. I try to use this to document issues I hit and get them logged so that QA and development know the exact machine config, SW version, graphics drivers, etc. With that, they can try and reproduce the error as close to the original system as possible.

                                                           

                                                          If you repeatedly get the issue, have you escalated it through your VAR? Between the steps to highlight the problem (especially the movie) and your system config, they (or SW tech support) may be able to identify the root cause.

                                                           

                                                          Cheers,

                                                           

                                                          Ian

                                                            • Re: Flow Simulation crashes and data loss

                                                              I havent tried to get help through a VAR because I'm using an educational copy of SW through my school - and there arent too many people at the school that use vista 64. If the problem persists on another model besides the tutorial valve (it seems to work fine on other models I'm using - that i have created), I will try to post a log from the RX tool.

                                                               

                                                              Thanks,

                                                              Dan