17 Replies Latest reply on Nov 16, 2010 4:17 PM by Matthew Perez

    To Knit or not to Knit?

    Matthew Perez

      Ive noticed that knit with gap control seems to be a "heavy" feature.  On larger surfacing projects I like to knit bodies together for various reasons(selectable edges for one) so sometimes I end up with a lot of knit features which weigh on the models.  Because of this I have found that its better to get to a good "moving on" point and Insert Body into New Part to start with other detailed features.  I dont really feel like this is a good method but it works and doesnt drag on rebuild times.  I know how certain things like sketch fillets vs feature fillets or featured patterns vs sketch patterns differ in terms of stability and updates, but ive never seen the same type of discussion covered for things like knitting surfaces, multiple trims then knit, vs mutual trim and so on.  So id like to open up the "floor" to see if you guys have run across and tips or tricks that have increased your productivity when dealing with these large surface models(600-1000+ features).  Now in a lot of cases its not neccessary to knit the surfaces but i would rather take (lets say) a car as one body then break it back down into its parts(after i have inserted it into a new part file).  I am approaching the current model im working on in a different manner that didnt seem to work which is why I am thinking of other methods.

       

      What better place to ask

        • Re: To Knit or not to Knit?
          Jerry Steiger

          Matthew,

           

          I tend to hold off on knitting surfaces together until I am forced to do so. I'm not quite sure why, as I've never had the feeling that they were particularly heavy in terms of rebuild time. (But I'm not very good about checking into such things.) I guess one reason is that it seems like knitting surfaces together sometimes changes the surfaces near the intersection. Thinking about it now, that doesn't seem like a good reason to wait to see what happens!

           

          Good topic for discussion; I wish I had more to say about it. With any luck, someone else will.

           

          Jerry Steiger

            • Re: To Knit or not to Knit?
              Matthew Perez

              Thanks for the quick reply jerry.  I too am interested to see what comes out of this discussion.  I didnt think they were really heavy either until the current project im working on.  It seems to really drag down the updates.  BUT even though they are resource hungry on the current project there may be other forces at work:)

            • Re: To Knit or not to Knit?
              Robert Stupplebeen

              I typically limit the number of knits (or any feature) in my models.  I pay more attention to when changing from surfaces to solids or back, because this usually has a significant rebuild time.

               

              Rob Stupplebeen

              • Re: To Knit or not to Knit?
                Charles Culp

                Matthew,

                 

                I've been known to use dozens of knits in my parts. What I try and do is never knit everything together until the end. If I'm working on a smaller section, and need to knit so I can fill in a gap, I will only knit those surfaces around the gap. On the last part I made I even did a copy surface of existing surfaces, worked with those, then used delete face and finally knit it to the final knit to solid.

                • Re: To Knit or not to Knit?
                  Dan Eldred

                  i think it's a matter of what geometry you need in your part and when (in the feature tree). i find that i'll need to use a mutual trim to get the proper output or in some cases need to use single trims in order to get to geometry hidden by that which i need to trim away. as far as knits go, i've not noticed any drawbacks to using them at will but i do like to use them at the last possible instance; which is what jerry mentioned. i'd like to think it's a habit formed from not wanting to put anything into the feature tree that doesn't absolutely need to be there.

                    • Re: To Knit or not to Knit?
                      Matthew Perez

                      Thanks for all the comments guys.  I agree.  I try to reduce the number of features in the tree when possible, but with surface i tend to try and knit things too much i think.  I tend to work from one area to the next.  Ill have an idea of the big picture but I work with things as I go if that makes any sense.  Fillets and other features are easy to plan for later but i find(at least in the past) that it was easier to knit as i go, but maybe not:)  What i dont understand is what makes certain knits more intense than others.  Ive uploaded a portion of my feature stats on the model that brought this subject to light.  Ive included two screen shots showing the two "heavy" features then followed by a whole host of knits.  The thing thats odd to me is the surface fill and fillet seem to me like they would be relatively "light" operations.  The fill was done as the second operation.  One edge was part of a surface(no edge conditions added) and the other 3 were sketches.  I've never run into any issues before but seeing these things is making me consider alternative modeling methods to increase the efficiency of my models.  I am going to spend some time figuring out if things like gap control make the knit feature more intense.  Some of the knits needed gap control, others didnt, so investigation is needed there.

                       

                      im slowly working backward through the model and going to travel down the knitless path where possible to see how much that makes a difference.

                       

                       

                      Keep it coming!

                        • Re: To Knit or not to Knit?
                          Charles Culp

                          I wish that 1.5 second knits were my longest rebuild times...

                           

                          As you can see from my tree, there are plenty of knits, but they are not always at the top. The Combine is #1, but that is no surprise to me. This is a feature where I have "hollowed" a very complex part, and the combine merges the exterior geometry with the interior geometry. The three 12 second cuts after that are probably because I have verification on rebuild turned on, I imagine they will go down after I turn that option back off. The same with the Boss Extrudes. The other features will probably remain at the top.

                           

                          In general, knits take longer depending on the total number of faces in the knit (which is why I suggest only knitting a few faces to each other until the very end of the tree). Also, they seem to take longer when there are strange gaps, but I can't quantize that for you, it's more of a 'feeling'. Fillets are always long build times. Remember, all a fillet is is an automated surfacing command. It is the same as Double offset + trim + sweep + knit.

                           

                          Statistics-Knit.png

                            • Re: To Knit or not to Knit?
                              Matthew Perez

                              I like that charles, revA_aftermeeting   Thanks for posting your feature stats.

                               

                              I have been going back through this model and i still cant figure out where the issue is but i have come to the conclusion that its not the knits.  However I am removing all of them i can.  With this model only the rebuilds happen then i get a "lock up" for a bit.  I assumed it was some heavy features but after stripping the model down a bit it appears to be something else.  Its only this model and its happened on different computers running different operating systems, graphics cards etc

                              • Re: To Knit or not to Knit?
                                Jerry Steiger

                                Matthew,

                                 

                                Seeing the feature statistics from Charles got me thinking about what is often the feature that takes the top position in our parts: Delete Face with a patch or fill on a solid body. I believe it is really the knit part of the Delete Face that is taking most of the time. The Delete Face gets used to fix multiple sins in the model. The geometry is always a bit dodgy in the area and the Delete Face is used in an attempt to clean it up somewhat.

                                 

                                Jerry Steiger

                                  • Re: To Knit or not to Knit?
                                    Matthew Perez

                                    Thanks jerry for looking at that.  I can honestly say that ive never used delete face and allowed it to fill or patch anything(aside from playing around with it a bit).  Even though its a few extra steps I always do that myself, more times than not because its not a simple fill in my case.  I find that the delete face command doesnt take up any time in rebuilds so it does make you wonder a bit.  One thing I do wish would happen(semi on this topic) and i know this comes up all the time, but being able to separate a face.  Now i use split all the time.  Some times to get curves on a surface for paths, some times for edges for operations like loft.  The downfall is whenever you split a face, if the surface it belongs to is knitted to something else, this split line is erased essentially.  In some cases i have had to go as far as creating a copy surface, trimming one part of the copy(to keep what i want) and trimming the other part of the original.  Where the split works just fine having it removed after a knit is a bit of a hassle.

                                  • Re: To Knit or not to Knit?
                                    Charles Culp

                                    Same model (almost) verification with rebuild off:

                                     

                                    VerificationOnRebuildOff.png