Shape Analisys

[Jacoolo]

Wide range of discusion taking subject as a base.

Well, in fact in this moment I am interested in analysis used when evaluating surface models - what should be used and when?



some of them are easy to answer "what can be expected from them?", but there are still some I do not know what are they for?

 

[Jacoolo]

an example

I made a 3 sided boundary blend, trim it, and then using such conditions I made 4 sided boundary blend

now, this is gausian analisys for 3 sided boundary blend



I suppoosed to obtain red corner when the iso parm meet each other.

this is gausian analisys for 4 sided B.B



Can somebody tell me:

1) what are the right values, exceptable ones for gausian analisys?
2)second how can I control values in the result window - set the range?

thx in advance

 

[Jacoolo]

I supposed if a pic with naked girl would be included, the answer would become just a matter of seconds

come on guys, it just can`t be so difficult!

 

[jeff4136]

> it just can`t be so difficult!


Difficult? Not? Everyone decides for themselves what interests them.
I do know you can't bring fine surfacing down to your level. You must
climb up to it's level and that will take some 'homework' on your part.


A good reference, right out of their Help ...
[url]http://www.kxcad.net/Rhinoceros/Rhino/commands/curve-surface analysis.htm[/url]


Gaussian curvature is the product of the priciple curvatures.
As such, it's primarily an indicator of compound curvature and
'developability' (e.g. can an undistorted flat pattern be
made from the surface or vice versa).


Mean (average of priciples) or Maximum curvature are more universally
applicable for evaluating form, shape, flow. (Help mentions a 3rd
derivative analysis. I don't see it.)


Except for Gaussian Curvature, where zero is the "right" value for
developable surfaces, there are no correct values other than those
dictated by your design criteria, i.e. minimum radius. Just look at
the graduations in color. In your pic the irregular boundaries
between colors (blue to green, green to red) indicate surface
irregularities. Be aware, though, that fineness / courseness of the
analysis mesh will affect indications so if other analyses don't
correlate increase Quality. You are, of course ,just as interested
in continuity of the color regions across surface boundaries.


Each analysis type tells you ~something~ about the surface.
None of them, discreetly, tell you the whole story.


Shaded analysis range can be changed, to shift the color 'bands',
by dragging the 'pointers' on the left side of the graph or typing
values in the upper and lower 'cells'.

 

[Jacoolo]

thx jeff, I am gonna check that

btw

jeff, I am sure you understand that the"difuclult" was mentioned according to the case of shape analisys overall, not the geometry included on pics

 

[Jacoolo]

I think I have a problem that is topic related.

I made following curves for B.B with following conditions(see below)



Pro\e reports an error for tangency transition in place marked on pic above.

The red curve is made by sketch projected on surface, so it must be tangent I thought. So what the hell problem is?

I`ve made some investigation on my own and realize that if surfaces are not tangent the curve would not be tangent either. This is however strange while the tangent edge(gray) already exist between them.

So in the next step I`ve made analisys for dihedral angle, `cause as far as I know this tool can provide real information about tangency.

This is what I obtained



so please tell - if my conclusion considering dihedral angle are right - what is the minimal value for dihedral angle between tangent surfaces?

 

[Jacoolo]

I`ve made further investigation and I am now 99% sure those surfaces are not tangent in 100%. I `ve found also that if dihedral angle is less than 0.05 Pro\e do not claim with transition problems while creating Boundary Blend.

More, I find dihedral angle as a tool great for analising surface offset problems. I noticed that I could not reached with original surfaces(all merged with grey edges - tangent) thickness more than 1.5. With the same shape but surfaces made in another way and dihedral angle close to 0(0.00015) or 0, the thickness problem does not occure and 2.5 mm comes with no problem.

The last thing I am confused about is, why the hell pro\e displays such edges as tangent one if in reality there is no tangency?

 

[jeff4136]

The more you know; the more complicated it becomes, huh?


Like all things, tangency is subject to tolerances.


For shaded display of solid face two sided edges; tolerance is set by
the config option tan_angle_for_disp. I'm not sure if there's a fixed
default value or if it varies with model tolerance or some other
variable(s). Look at current_session.pro showing all options to see
what it says.


In addition to being governed by a tolerance the indicated dihedral
angle is influenced by G0 (positional) tolerance. That's logical if
you assume the angle vertex is an estimated point somewhere between
mismatched edges. Sounds good, anyway. If you set model tolerance to
a finer value, you'll usually see a reduction in indicated dihedral
angle.


Once you cross the analytic to NURBS frontier everything gets a 'fuzz
factor'. All this stuff gets mixed in a pot and processed thru some
lofty mathematical magic (the stuff good geometry engines / 'kernels'
are made of) to get results that work ... most of the time.


Tangency's effect on offsets: You've stumbled on one of the great
mysteries of surface brep modeling and the subject of ongoing
mathematical magic refinement / development (this is a biggie among
the things that separate geometry engine men from boys). When
considering the problem you must consider the process involved in
offsetting and rejoining individual surfaces to create a quilt and
imagine the difficulty of creating a usable intersection curve
between two 'fuzzy' surface definitions. If the dihedral angle is
zero there is no need to extend or trim a surface normal offset to
rejoin edges as they will be coincident. If the angle is several (a
few?) degrees and the surface definitions allow extending (if
necessary; this is where 'clean' surface definitions become important,
too) there will be no particular problem creating a new intersection
curve and joining. The most difficult case for intersector functions
is a near tangent intersection of two 'fuzzy', at best when compared
to analytics, surfaces usually compounded by dihedral angles that
wander back and forth across zero and a forced, 'dirty' surface
definition. The mathematicians have to work some real wizardry to get
that all to come together. When it doesn't come together look for the
condition that's preventing it.


Again; changing model tolerance will often have an effect but it's
not always what you might expect. I've seen loosening tolerances
cause as many failures as tightening. If you use Rel Acc you have to
watch for anything, i.e. a datum point off in space, that blows your
model size out of reasonable proportion in relation to the actual
surface geometry. Aside from that, I'd try to figure out what it is
about my surface definition / creation method / input that's causing
dihedral deviations. It's not always apparent and sometimes no
fathomable, to me, rhyme or reason seems to apply. You must determine,
as well, if whatever deviation you are seeing is significant.


> The last thing I am confused about is, why the
> hell pro\e displays such edges as tangent one if
> in reality there is no tangency?


Shaded display is shaded display. Analysis tools are analysis tools.
Remember, too, analysis tools have their own set of 'tolerances' or
'fuzz factor'; analysis mesh quality, etc.

 

[Jacoolo]

Thanks a lot Jeff,

each day I am 2 miles further, however- still - 1000 miles stay ahead of me

anyway, I work with absolute accuracy, never meesed with rel, only problems come from such colaboration(rel accuracy)


I wonder - how far user can go with understanding the core of Pro\e to not accept all of its errors as - this is because of mathematics. I sound now as a noob, I know

Next question - is there a way to obtain info about surface normals(I know james run a topic this stuff related, but there is no info about that)

 

[jeff4136]

> ... noob


We all are. Some just older noobs. Beware the 'experts'.

> normals


Analysis / Geometry / Curvature: Defininition / Plot Type
(it shows up on the 'list' when a surface is the Reference Geometry)
(Notice, also, the Tool Tips that show if you hold the mouse pointer
over a menu command or tool button.)

 

[Jacoolo]

exactly this is it!

still I miss what advantage one can have from Radius, Deviation and Slope analisys?

 

[Jacoolo]

is there a way to obtain an information if already created composite curve is a chain of tangent entities?

What I mean - it happens one has to create a trajectory from two different curves for VSS, and a basic demand for VSS with Variable Section is to have tangent chain trajectory.

Such a curve can be also used in another tasks like projecting on other surface, offseting and then such constructed curve is used by VSS. But if the first one is not a tangent chain then...

So, normaly, after I create a composite curve by Copy>Paste, I run VSS and try to make sketch, then if VSS let me do this - curve is a tangent chain if not, well, it isn`t.

However I curious if such information could be obtained directly in Paste function, or while being in this function?

 

[jeff4136]

Analysis -> Measure -> Type: Angle ?

 

[Jacoolo]

well, I like simple solution and here it is - thx jeff