Improving Product Life with End-to-End Automotive Durability Analysis

Improving Product Life with End-to-End Automotive Durability Analysis


well thank you for obtaining the webinar this morning is three of us says i’ve been using mass transfer a long time ten years in the arrow a defense industry past 20-plus the automotive and just general engineering I’ve seen our command I’ve seen our computing environment changed drastically over the years from the CDC IBM univac era when and then when the pc was released by idea txt 80 and now laptop and tablets up any other associated things that go along with the automation process of CA practices so i want to go to a bit of history because there are students and others on here probably haven’t seen what i have over the years and i want to say that no my view is anything past displacement calculation and CAE is post-processing stresses of course are done inside to get the fem codes but what you do afterwards depends on when you start an edge area mango hand calculations anyone who’s in aerospace probably knows a Bruins handbook in the calculations have to be done for march as a safety encourages general engineering there’s the lot of handbooks Forks has one that everyone probably remembers or stress and strain so the first step in automation is specially these these handbook so that everyone’s using the same equation to get the same answer and that’s the important part which then leads to what you’re doing these engineering calculations I remember that to know the first time you sure what it’s kind of interesting that it becomes kind of boring which means it’s error-prone there’s a lot of physical time involved with it because we always had a checker to sign off on that and that’s probably still around in some areas someone check your hand job please make make sure you’re doing the right thing finite elements of course produce volumes of information and you’re not about to do a stress calculation for every element although you would like to because that would eliminate errors so here’s our first step in automation is someone writes a program and I’ve written them maybe a few of you have two new run this stuff now we’ve got a process that we’re going to do and becomes standard within the company everyone accepts it everyone’s doing it we can do these on programmable calculators you’ve got the private commercial software where some might remember TK saw her it’s still around where they get brought in formulas from forks handbook so that you could just go through an interface and calculate things so enough of that now the CAE process whether you’re an automotive aerospace it’s pretty much defined and it’s become big and everyone follows it so that anyone can sit down and do this every company has their own see a process you’re using all them at zebra products you know you go through a loop of some sort whether you start out with CAD you can bring it into our apex product and cleanup and mesh it let’s send that out later to one of our chem solvers which communicates with act trainer Adams possibly or you can go straighten just do your chem analysis and then views the post-processing for the buoys Pat ransom expert meant an apex can do that VI is for act tran but there is a defined process that you do here you go through this loop and you repeat it over and over again and in some cases you have your own code that you insert into this for the big thing is that you don’t want to do this has become so important today as you don’t want to break it because if you break it you increase time and cost that raises management size to see what’s going on different results and I are not wrong but they’re different so if you’re going to replace some type of calculation in your process you want to make sure that it fits in and doesn’t cause any problems I know this our sales people notice so we don’t come to you with suggestions that will make your life miserable that’s that’s not conducive so if we want to improve things whether it’s a commercial cee or yourself you know you got to say physical wall time it’s engineering costs and computer time and you got to fit in with little or no changes some changes maybe you have to call it a different coat and do something a little bit of file for it but once it’s done it never has to change again so we’re going to talk about the fatigue step in CAE in how we can improve that time shorten it down and not break your process everyone has at least done at least one the P calculations somewhere in your career help your student you do a simple SN curve lip up you might use one of the equation Gerber’s or soderbergh’s do this a couple forward problems into your test problem you might even you’ll have a hard problem that considers the surface conditions or pre-stress after that it becomes more involved in at that point we want to cut considering doing a code to do this is relevant multiaxial spot-weld see welds or her looking what they call en analysis which is strain based it has to do a computer calculation for this also these femme models they know these two these femme are so large that the volume of stress it comes out almost mandatory that you have to filter things so your chair workflow is typically this you will run a finite element model you will produce a volume of FPA output and build a strand is very good at me Seb Coe is very good at producing lots of data for you typically mean you know binary for sop to that’s our format now you’re going to have to translate that which means someone’s guy to write a code to read that whether it’s the commercials the team code or you rip the one because you have your own fatigue code as you’ve done so you turn that into a very large file if you know ascii things have expanded tremendously you’ll have to bring in your material properties that suma doesn’t change that’s pretty standard once you’ve got it set up and then your cyclic load data again those are usually standard and they don’t they don’t change particular you’re in a process you don’t change those into use standard stuff over and over again then you go you your fatigue calculation you might have to write some type of translator to put the output in the format did a graphical interface to use and actually display as a is Rachel I like we see here and print out and so on but these fatigue codes are generally independent of of your workflow you’re using someone from the outside and you’ve got to inserted inside there now in order to do that with a non mass transit you know fatigue type codes we’ve got three ways of going about this the first one is very typical shell script whether it’s a no type scripting or Python whatever your favorite cocoa codeine translator is you’ll build that it’s usually an executable that runs after an ashtray is completed that’s very very common now you’ve got the cost of someone that has to maintain math and keep them keep track of it so IT departments are our don’t like to do that and you’ve got all the disk space that you’ve got to maintain or open this space for these large files that have to be dealt with and then there’s transfer costs and so on depending on how big your models is how helping your you know what your internet setup at your company another way to do this which makes it a little bit easier is we call we’ve got a maje one site mastering called I show which causes net Tran while the third party will execute you the nest rain starts up again prob this is that when a strand sits idle because the batch program ready bitch Corbett’s can hold that memory and disk eating that up well that’s a detriment also you’re not using the efficiency that we have inside mass transfer file at i/o handling look very very good at doing this we handle very large models and we were doing it for 30 years the third way of doing this and this has been a little more recent with some changes in the net Tran file management that we have is compiling the executable library right with inside mastering it says relatively no but the advantage here is are you using Emma seen a strand file and IO handling capabilities licensee becomes easier because it’s embedded with the net without clicking our licensing system so it becomes a feature or if you do this yourself you don’t have the license of your own code but also it’s the familiar dance troupe master and be defiled to supply information over the years this has just become a standards of all data file and it never changes from version to version it gets improved with additions but you know it’ll always be upwards compatible you so what we’ve done is we’ve turned it into a female library call which is the mastering scripting language is d man so it’s fatigue solved and we’ve had a long long relationship with encode so what it is is the dt lie that you see in encodes of the peaks offer and we call it out and we build all the files for them we it all set up we’re using the nest tran databases scratch and the data management to maintain this information that goes inside there and if you look in the installation you’ll see there’s the encode libraries that we’re calling it using so vast strands running we’re using the memory we’re not tying it up while sitting there idle we give that over to the dt line and let it executes so now we’ve done is we’ve compressed things when you run your final model and we support the three methods of doing fatigue analysis statics normal bones in both transient all of the teeth fatigue material properties are in bulk data for men and we run the dt live right within a strand or lowes database does not change you aren’t going to make you turn that into anything different or unusual will read the DAC file the RPC as they are so we’re not going to cause any difficulties in that part of your process and the fatigue output look so I’ll show you later out you’ll be pleased that it’s basically stuff that you are looking for not only our standard stuff puts some additional one demo model that we’ve shown over here is this is a nav estar camp for a truck it’s at that time is a relatively small space requirements have dropped dramatically you see on the chart here anyone who’s used up we call nef which is our embedded fatigue as seen at least a nine percent reduction in file storage if you’re running a simple model like a keyhole standard one that everyone execute their fatigue go to show that we do get the same answer you know your cpu type will be about the same but as your model size is grow and I know there are very large large models out there that is the tendency of FBA is to build huge models your CPU time starts to starts to drop we’re seeing twenty thirty percent reduction in time sometimes larger depending on the inefficiencies of the old method or these deputies solver is outside so it’s it’s a lot simpler to do which is great for your process it’s faster smaller it’s also portable too and once you have everything built because it’s a post-process you have the data structure in there to launch the team you can run it over and over again and you don’t hit the touch it so this this model at one time it ran through MSC fatigue what we use a DG live solver inside pan train is the pre and post but still we has these intermediate files I had time it took about 120 minutes to run all the way through 191 individual files he has to be processed and stored on the disk when we move this over to the embedded fatigue we drop their solution time to 25 minutes only 32 megabytes up to space compared to what 282,000 before they’re obviously the results are the same they have to because we’re using the same song and we only have two individual files to deal with the input file Maryland being the results so we’ve kept everything inside the nast ran scratch database the good thing about this is that right now it would take under 20 mins well that’s you know two cycles per day now we’ve dropped it to eight this means if you’re doing any type of envio or do we type analysis the more you can do in one cycle during the day that’s that’s the goal faster that you can get to the optimized or what you’re going to do with this yep do this wheel carrier or steering knuckle this one took eight hours to do the petite results I was told as the multiaxial calculations which really sucked up the time well we brought into nef it now takes 38 minutes you seem depending on what you want to do when the fatigue which are you can have some huge CPU time games full car body this is typical apparent Detroit reduce solution will 12 the mold transient and actually this one’s for million degrees of freedom that that’s really rather small but even that size you want to do petite filtering that’s standard capability in dt live in wheel we have that enabled inside nef where your first step course is to filter on element stresses and damage then you go through you compress the time history which is another filtrate step and then when the final results come out people back to do a complete fatigue analysis on the remaining elements this type of filtering avoiding the typical error of having to select a group of elements to process but what if you pick the wrong group of elements and the one that’s really critical hasn’t been processed for this type of fatigue filtering avoid set everything gets look down and now we can do this one in in 2.5 hours we’ve got it hooked up with our solution 200 and this simple simple model just an IV with some holes I don’t we want to do a stress and a shape optimization and in the past and for fatigue if we ran this you get thirty-nine percent mass production which is great but if you ran in fatigue analysis post on that then failure in 10 to the fifth cycles so now what you’ve done is you actually built the prototype in induce that cost or you’ve had that cost now test failed you have to go back again include up and click fatigue in the authorization but we do that within master and salt 200 we could be I beam shape in the lower right corner you look at as much mass reduction but you can failure 10 to the 8 cycles which would look deep right Irial is here it’s a shock collar and forcefully has been turned 90 degrees instead of pointing straight up but all the shell thickness is because there’s individual panels that make up these shock towers each one can be a design variable and if we of course you want to minimize weight but will include fatigue on this the summary on this one is can we knock weight down got twenty twenty four percent reduction and we brought light up 2.7 2.8 here’s the individual thicknesses on those pieces that fit together the next question is well you know put those together you can’t use spot loads and yes we have the spot will critique analysis inside nef the common modeling methods out there are bars or beans there is the mscc weld element that we can handle and there’s another technique that uses the solid hex element with the rigid body element number three to connect it down to the pants that way be little spot though which is that red nugget does not have to be exactly on top of the right point are under the woman is the same way with bars and beams they have to be ready read connection which puts a lot of restriction on them so we could include if we have spot welds also as part of the fatigue analysis and we can do seen Welles seen wells are modeled as shown then squad forwards or equivalent there’s root elements toe elements in the yellow which is the actual throat that you want to take a look at so we can process things aside EDF and these different types of scenes can’t be modeled though show all that skin news inside the fatigue analysis that’s inside man strain output formats besides the bar below me to punch file will give you the ascii universal violence csv we will give you the hypermesh Hyp extension if that’s what you want to post-process head the design life FBR will come out but you can get every one of these output files all the same time if you’d like our product pant rain has completely enclosed support and our good friends over at answer also have complete reimpose support for this fatigue input and output so by embedding the NES Inside Man strand we’re not going to break anyone’s process use the same dance our CP files it’s all BTW live petite options nothing different so you get the same answers so whatever use your data storage definitely reduce wall clock time also get to use our SMP voice and for parallel support which essentially gives you under the core access fatigue analysis on an element is essentially independent of each other so each one of these cores or we were called bath processors or CPUs can process groups of elements individually so it is like explicit goat its scales linearly with the number of course was people than just a few elements you don’t need to have 16 cores active that’s kind of a waste but if you have a very large model take a look at every element we really scale linearly with those elements in the number of course we’ve got the direct link to the next train optimizer and if you’re doing an MDL where you’re using let’s say mode frontier in your have a separate mass trans and then fatigue analysis has two different inputs we can now merge those into one you’ll have one thing to execute an advanced or so your overall MBO one much faster you


Leave a Reply

Your email address will not be published. Required fields are marked *