Increasing performance on a complex 3D part

I have been modeling a part that is a cylinder ~500mm dia x 1100 that has a complex surface geometry. It has a bunch of helical slots approx 120mm long that are repeated as a polar array then are reversed and off set half the polar array angle. The result is like a herring bone pattern of overlapping slots in the surface of the cylinder. There then need to be 360 angled groove across the width to form teeth.

My approach was to:
1 - model a piece of the cylinder 120mm wide, draw a helix as path and rectangle as profile and sweep it to form an intersecting solid where the slot is required, use fillet to round the ends of the solid.
2 - Polar array the swept solid to give 24 similar solids around the circumference.
3 - 3D mirror both the cylinder and array to give the next section of cylinder with the twist of the sweeps reversed. The mirrored body is offset less than width of the cylinder so the mirrored solids intersect about 10mm. I then 3d rotate the mirrored solid by half the angle of the array. This process is repeated until I have built up the full width of the required cylinder.
4 - I then union the 120mm section of cylinder for form a single body, Then subtract all the helical sweeps to cut the slots in the surface.
5 - I model the negative form of the teeth, constrain it to the cylinder surface, again polar array to give 360 solids around the circumference intersecting with the surface.
6 - I start subtracting these from the cylinder to 'cut' the profile into the surface.

The problem I have, is the dwg is now over 14MB, takes ~5 minutes to save, I can't SU the teeth forms in one go Brics refuses to do that. So I am doing them 20~30 at a time works but the SU is taking 10~15 minutes to process. My question is is there a bitter way to do this? I have still got a reasonable amount of features to add and the processing time is killing my productivity.

Brics is running on a reasonable spec PC Dell workstation with 3Ghz 6 core xeon, 16Gb memory, graphics could be more powerful being a NVIDIA Quadro FX 580. Brics is using 1 gig of memory & 100% of one core.

PNG image attached as the DWG is too big!!

I am hoping either a better approach to modeling or some system setting can be suggested to speed thing up.

Thanks
Adrian

Comments

  • I would create 1/24 segment, put that solid in a block, array the block and create a block from the assembly.

  • Please can someone back me up or refute me on this: For BricsCAD if found that I no longer needed the Quadro type GPU. One of the reasons I got BricsCAD was because with my GTX1060 It could handle my most insane model of a wall of tyres and chains that would freeze other software. So my suggestion (at your own risk) is get a GTX1080 GPU.

    Other suggestion is, if it is possible / practical, swap to wireframe visual style. It is generally a lot less taxing on your computer, but more taxing on you.

  • Actually 14Mb seems too big even for that model.
    Try selecting it manually with click drag left and copying it to a clean blank new drawing.
    I remember sometimes I'd get a drawing that for some reason was like 20Mb and tried to purge and delete and all sorts of stuff but just couldn't get the junk out. My only option was to copy the visible stuff out into a new drawing.
    Worked perfectly.

  • akendall1966
    edited June 2018

    Hi DFLY,

    Thanks for your input. While I wouldn't be opposed to investing in a better graphics card if it gave a boost. I don't think my issue is GPU related I have been watching the CPU and GPU performance monitors and during the period when I'm waiting CPU is 10~12% (i.e. 1/12 of a hex core xeon being dynamically overclocked) and GPU 0%. So to me it number crunching the geometry that's taking the time.

    When the the screen is being re-rendered you can see the GPU working hard but there little latency whatever the visual style in use.

    Roys useful input here on modeling a part of the component then using an assembly to create the whole from a number of identical instances of the same geometry, and in another related thread pointing out I could achieve the same result with less splines by taking a different modeling approach. Reducing the amount of geometry to process decreases wait time.

    Regards
    Adrian

    Was writing the above when your follow on post appeared, will try your suggestion thanks again.

This discussion has been closed.

Howdy, Stranger!

It looks like you're new here. Click one of the buttons on the top bar to get involved!