I would like to know if it would be possible to increase the scope of a law under catia if so, how can I do it ?
Let me explain, I'm modeling a vault, it is defined by a hyperbolic cosine equation.
To obtain the shape of my vault I created a straight line (on my image it is straight 2) and then the law allowing me to obtain my shape (called the intrados equation). Then I use the parallel curve function. I select my line 2 and then apply my law to it.
And I get the following curve:
And we see that my curve is set to [0,1] and I'd like to have an integer curve, which is to have its symmetry on the left-hand side. Giving a definition of the type [-1,1]. And I specify that I do not want to use the " mirror " function. I would like to get my curve directly to have a " single block " part.
For your information, my law is as follows :
z=((-1)*A *(cosh((y*span)/A ))+(A +height ))/1mm
With z and y as the real type parameter.
I wanted to add the following equation using the operators : -1<= y <= 1. But Catia tells me each time a syntax error.
If you have a solution to offer me, I'm all for it.
I want to obtain it so that I can then make a mechanical analysis of it.
But via the way I make a symmetry, I found that I could not apply any link conditions or make a mesh, or anything else. It's as if symmetry is dead.
I would like to have a second vault of one room so that I can compare the results of the 2 analyses to see if I get the same thing or not. (Because my vault being symmetrical, can be a modeled part is enough to define everything).
I think Franck Ceroux's idea is to use the "extrapolation" function of the surface fabric to extend a surface from an edge and the curvature of the existing part.
As for whether half of the model is enough, there is a fairly simple criterion: all the elements of the model must be symmetrical (boundary conditions, forces, etc.). If this is not the case (e.g. pivot on one side and plane support on the other), then everything must be modelled.
Otherwise, if I understand correctly, you first extruded the curve and then did the symmetry. Have you tried to do the opposite? You make a symmetrical of your curve, you assemble these 2 curves and then you extrude. This way, the entire surface is created in 1 block, right?
Hello sorry late answer but no internet at my place at the moment PB line phone.
Yes "Chamade" is the objective
Extract the "sketch" curve in the surface workshop, extend the curve hoping that the deviation is zero.
Otherwise another solution (DIY) but it allows to have a curve without the break (fictitious stop) due to the mirror function.
Symmetrize the curve / assemble the curve + symmetry / point-on-curve repetition function (create enough points) / select the points in the graph from the first to the last (shift) / then curve function passing through these points.
We obtain a curve without discontinuity, i.e. a surface with a single tile.
Of course, there is a small difference between the curve evaluated by a law and the one drawn on the points of this curve (but very small, especially if we extend the curve to these two front ends.
Thank you for your answers. Franck.ceroux I'm going to try your idea right away.
Pounding
Post 1: normally my structure being a vault all the elements will be symmetrical. But what I have a problem with is in the case where we have a point load placed on only one side of our vault, I would like to see its impact on the entire vault. Even with this case, do you think that just half a model is enough or not?
Post 2: Well before coming I tried the 2 methods.
First by extruding the curve and then applying a symmetry.
And creation of the complete coubre by repeating the operation of creating a curve (axis creation, law application etc...) 2 times. Then 2 extrusions.
And I come back each time to this problem of edge in the middle designating me "2blocs"
Thank you very much for your invaluable help and for your time.
Franck I tried your technique first, unfortunately I can't get a curve according to my law. I can only have rights. I'm probably doing it wrong.
I tried your second technique only halfway. Let me explain: the assembler function allows me to obtain a continuous curve with my two curves and without having to create a multitude of points. I wanted to see if I could directly do an extrusion and yes it is possible. So my question is the "assemble" function allows me to obtain a continuous curve, what is the point of creating a multitude of points to make a curve pass? (Model more accurate than if I go a curve assembly? Necessary for the software? Other?)
The assemble function creates a single curve of the curve + symmetry, but at the point of symmetry, even if we assemble with continuity in curvature, it sometimes remains as a connection point that generates this famous fictitious edge.
I explain the phenomenon by the tolerance of the software, the starting and ending points of the curve are created on an axis (X, and Y), a coordinate is therefore at zero (+ or - the tol of the software) so I imagine that in reality the point is not exactly on the axis but can be above or below right or left.
So by manually creating a single continuous curve (passing through the points created coincides with the assembly) we do not have a discontinuity as observed on the assembly.
I just finished applying your second 2 method. Indeed, it allows me to obtain a magnificent vault without this fictitious edge.
I don't have a significant difference in dimension, 2.7 mm for a vault of 3000 mm wide, I think it's correct for an application in civil engineering (it works to the centimeter anyway).
Last question is after I stop (for this topic ;) ) I applied a repetition of 1000 stitches, it seems a lot right? How to define the appropriate number of points to draw.
By this I mean that there must be a point limit allowing us to obtain a curve with correct precision and this without "superfluous" points that will only weigh down the model. It is obvious that the "precision" of the curve depends on the number of points constituting it. (This would amount to a minimization of the number of points used to obtain the most accurate curve).
Example: I could add 1 billion points, my curve would be more than well defined, but what would be the added value of these points? There must be a lower number of stitches to achieve the same precision/fineness.
To reduce the number of points I do this empirically.
On each point (not possible with hundreds) I put a tangent line to the assembly of the curve + the symmetry).
Then after creating my curve by the points I define at each point a tangiance (line created previously.
So I can redefine these lines (angle passing through a point) and thus adjust the tangiance directions in order to obtain an approximate result with a reduced number of points (it's approximate and you have to put the same values on the symmetric side.
Just to come back to the question of calculation, as I said, if there is a load that is only on one side, then the problem is no longer symmetrical, and so you need the complete model.
It is easy to imagine in this case that if we put, for example, a 45° load oriented towards the center on one side only, then this side will be pushed towards the center while the opposite side will be pushed outwards. So symmetry reduction doesn't work.
