driving area with trajpar

[design-engine]

I've been thinking about trajpar variables and running sweeps with equations a lot lately.

For an example: Say a rectangle with rounds on one end to a square with rounds on the other and over the distance I want the area cross section to double.


Edited by: design-engine

 

[puppet]

I once did this for a micro fluidic's application in the attempt to minimise dead space


off the top of my head the key was in getting pro-e to solve the simulations equations

area = 100
perimeter = 50
SOLVE
d1*d2 = area
2*(d1 + d2) = perimeter
FOR d1 d2

 

[jeff4136]

> I've been thinking about trajpar variables and
> running sweeps with equations a lot lately.


That's a worthwhile thing to think about.


> For an example: Say a rectangle with rounds on one
> end to a square with rounds on the other and over
> the distance I want the area cross section to double.


In my experience the biggest impediment to getting the
desired results from a VSS is my own inability to
adequately describe the shape.


2008-03-19_082731_vss-area_driven.igs.zip

 

[Jacoolo]

In my experience the biggest impediment to getting the
desired results from a VSS is my own inability to
adequately describe the shape.

I suffer the same

 

[design-engine]

Thats why I have always in the past bust out the geometry with curves and boundary surfaces...... so I can hit all the roadblocks to form. And let the managers say no to the form 20 times. Once I get the basic form... then recreate that with a vss.


At least thats what I've done a couple times in the past.

 

[csusie]

why dont you use a swept blend where you can control the cross sectional area as you sweep.

 

[jeff4136]

> why dont you use a swept blend where you can
> control the cross sectional area as you sweep


A Swept Blend gives you less 'control' than a VSS.


Think of it as the difference between an interpolated curve
(a spline you create by 'picking' N points) versus a curve
created by equation. With the interpolated curve you have
no idea what the value is between points. With the equation
driven curve you know precisely where every point is because
it's defined as a function of t. (Interp curve is the Swept
Blend, Equation curve is the VSS.)


The problem we have here (as I envisioned it):
An arbitrary rectangle with rounded corners --> a square with
rounded corners (which can include a square with side equal
corner round * 2) and start section area * X.


Once you decide how you will define intermediate sections for
the Swept blend you've also figured out how to define the section
for the VSS.

 

[csusie]

however it sounds like he has somesection in mind at the beginning and end of the sweep. And the only requirement is that the section area doubles from start to finish. I didnt hear any other variance requirements in that statement.


If you do not define tangency conditions at either end, you will get a linear increase from the start to the end. It will be predictable.


I understand if you use cross-sectional control and add areas inbetween the end sections it interpolates what happens between, but that doesnt sound like what he described in the first post.

 

[jeff4136]

"For an example" sounds like a hypothetical proposition to me.


"Driving Area" indicates a linear section transition is probably not the desired solution.
Edited by: jeff4136

 

[design-engine]

yea... I just want to be able to take one shape into another over a distance and drive the area... like over 100 percent of the distance the cross sectional area grows 1.1 or 1.5 over that distance.

 

[jeff4136]

It's do-able.
You just have to decide what your driving parameter is and how to describe it.


A rectangle to square could be something like:


/* linear dimension transition ...
dim_base = start_base + (end_base - start_base) * trajpar
dim_height = start_height + (end_height - start_height) * trajpar


/* ... or linear area transition ...
dim_base = start_base + (end_base - start_base) * trajpar * optional_modifier
dim_height = (start_area + (end_area - start_area) * trajpar) / dim_base


Adding corner rounds (constant? variable; linear or nonlinear transition?)
simply (snicker) requires adding a description of the 'lost' area,
i.e. [ (2*r)^2 - pi*r^2 ], to the expressions.