How can poor performance be avoided when mating an assembly? 2

[bloodclot]

I found this question and answer on the SW website and thought it was good mate information. Kind of like "Mates Explained" or "Mates for Dummies" summary. Most of the time I feel like I fall into the "Dummies" category anyway. I hope someone will find this useful other than me.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
The processing of the mate group follows a sequence:
Parse the mate group to identify the order in which the mates need to be processed taking account of the interdependencies between the components / sub-assemblies.
Process the mate group in the order derived by the parse process.
Verify the validity of the processed mates.
If errors are found process the mates again in the order derived from the parse operation.
Verify a good result and if not raise warnings / errors.

Note as a change may be introduced for many reasons the mate group is processed each time a component is added to the assembly or moved in addition to the process of adding a mate.

General comments on mates:
The time to solve a mate group has a squared rather than linear relationship with the number of mates to be processed.

The parsing process must take account of the possibility of an over-defining mating situation. In the simplest case consider 2 rectangular blocks. With one block fixed in space we can add three co-incident face relationships to position the second block next to the fixed block. Without these mates the second block can move with 6 degrees of freedom. Position in space – x,y an z plus a rotation about these axes – Roll, pitch and yaw. Adding a single coincident face to face mate removes three degrees of freedom, one positional degree and two rotations. Adding the required three coincident face to face relationships would remove a total of 9 degrees of freedom and hence the mating scheme is over-defined. The mating parse will recognize this as a valid over-defined situation in simple cases. With more complicated mating schemes where the mate sequence relates through a number of components, recognition of this particular situation can become even more complicated if a mate is referenced to a patterned component. The mating scheme must position the source component(s). These would then be patterned and a new mating scheme derived for the post patterning operation. The number of possible solutions will increase in this situation.

Often performance will fall dramatically if a true over-defining case occurs. In this situation there is no valid solution to the mate parsing process and as such the program will select the least bad solution. In order to derive the least bad solution all possible mating schemes will be considered. In large mating schemes the parsing process will take a considerable time.

Credit to SW for this explanation of mates!

Bloodclot

What do you see when the Pillsbury Dough Boy bends over?

Doughnuts

 

[ongybill]

Some of it made sense. Some didn't, so I tried his example.

Drew a block. Inserted 2 copies of the block into an assembly. 1 fixed, 1 not.

Added 3 mates to totally constrain the unfixed block next to the fixed block. No problems. Nothing overdefined.

The overdefined example just doesn't hold up as it only removes 6 axis of freedom, not the 9 stated.

I find SW often thinks things to be overdefined that aren't. In fact often deleting the 'overdefined mates' and recreating identical mates produces zero mate errors.
Or deleting the last mate created & suppressing the 'overdefined mate' then recreating the mate I just deleted and unsuppressing the suppressed mate causes the mates to work.

Kind of an unpredictable part of SW. Some might even call it a bug.<G>

 

[dgowans]

Adding the required three coincident face to face relationships would remove a total of 9 degrees of freedom and hence the mating scheme is over-defined. The mating parse will recognize this as a valid over-defined situation in simple cases.

ongybill,

This is in fact overdefined (3 face-to-face mates, each removing 3 degrees of freedom), but SWX must consider this a simple case. When learning another software I was taught the 3-2-1 method of mating components: the first mate should remove 3 degrees of freedom, the second removes 2, and the third removes the remaining degree of freedom. When I learned SWX, I asked my VAR about this and was told that it was essentially impossible to remove a degree of freedom that had previously been removed. A better answer should have been that in "simple cases" it didn't matter. I've since learned that "simple cases" often become complex and then SWX barfs on mates it thought were fine 5 minutes before.

Maybe I'll go back to the 3-2-1 method.....

 

[handleman]

FWIW, here's a list of common mate types and the DOFs they remove. "C" is for Coincident, "P" for parallel. Of course, concentric cylinders is essentially a Line-Line Coincident, etc.

Of course, this list is not complete, but it covers most of the mates I use.

Mate Type:	Rot	Trn	Tot

Line-Line C	2	2	4
Plane-Plane C	2	1	3
Point-Point C	0	3	3
Line-Plane C	1	1	2
Line-Point C	0	2	2
Plane-Point C	0	1	1

Line-Line P	2	0	2
Plane-Plane P	2	0	2
Line-Plane P	1	0	1

 

[ongybill]

When it suddenly barfs on mates it used to like, I use windows task manager to force the program to quit.

This loses any work since the first save, but the mates work again when I reopen the assembly. Saves lots of time recreating good mates.

Guess I don't understand why SW overdefines a mate if it restricts more then the 6 degrees of motion. Reality is it doesn't restrict 9 degrees of freedom it restricts 6, some more then once, but in the exact same location.

In my jig & fixture design training a jig wasn't considered overdesigned if parts of it restricted the same degree of freedom more then once. The key was to ensure all 6 were restricted. Though the 3-2-1 rule was taught as the optimum as it kept things as simple as possible. The fact that constraining 3 faces of the part redundantly restricts some degrees of freedom shouldn't cause the software to barf.

I've found I have a lot less mating errors if I use face mates versus line or vertex mates, so I avoid them. I think I'd have many more mate errors if I did the 3-2-1 rule then I do now (ie. using more line & vertex mates), but maybe I'm wrong.