# CONNECT - Best Practice w/Parametric Modeling

Does anybody have any ideas yet on best practices when it comes to parametric modeling and constraints?  Specifically what is least taxing on a system or has the least potential of becoming corrupt.  Specifically for fully constrained things, I fully understand not everything has to be constrained, have constraints or variables.

I can create a 3D concentric constraint between a center line of a cylinder or the surface, both seem to have the same result.  My gut tells me that a line has less information to process then a surface thus the line is a better selection for the constraint.  Moreover, I can create the same 2D shape with more variables or with more constraints does one favor faster processing then the other once our models get "big" ?

Do some constraint types process easier then others if i want to make a perfect hexagon should i set all the sides to be = then use 120 deg constraint between (3) of the vertex them and put (3) dimensions in for the flat to flats and set them parallel... or should I add a construction element of a circle make all the edges tangent to the circle and force them to be parallel and only 3 of them to be equal then have only (1) 120 Deg. constraint and (1) flat to flat dimension.  I trend to favor the least number of dimensions as necessary and use construction elements because long term it's easier to review the history of a part if it has less variables.

When adding a feature to a solid should I make the new feature a new solid and union them together with constraints.... or should i make the new feature an imprinted shape on the surface then modify the surface to give it thickness?  The protrusion tool seems to only work between solids i anticipated this to work like the protrusion feature in the old V8i tools but it doesn't give the option to protrude blind, I actually don't understand how that tool is intended to be useful to be honest.

I'm willing to except that there are lots of ways to do things and that strictly speaking they are all right, but there are always ways that are better then others.

• There isn’t really a best practice that covers all models and type of modelling. You should try to have a profile well constrained (DOF=0) whenever possible as this eliminates ambiguity in resolving the changes in parameters. We are aware that currently in some case (when a profile contains arcs) it may be not be possible to obtain DOF=0, this an issue we are aware of and are working to address.

Using as few constraints and variables as needed to obtain a well constrained profile is preferable. Computational time increases with the size of the system we are trying to solve (constraints, parameters) but, again, the system should be well constrained to minimize the risk of unpredictable behaviour. Our solver will identify over-constrained systems and won’t let you add redundant constraints on profiles or solids.

The number of variables and expressions used for parameterizing a system (profiles or solids) depends on the applications and on what someone is trying to achieve:

i.e. is it required to have everything driven by one single variable or to have several independent parameters that can be changed?  What is the degree of freedom of the system we want to obtain?

• Good questions

Kirk