Edge Blend Self Intersection 1

[kbarne]

I have some fairly simple geometry. Using NX12, a cylinder is angled at 45 degrees relative to a flat plate, and I would like to put an edge blend around the intersection of the cylinder and the plate. This works at very small radius values, but as the radius increases, there appears to be a "pinch" in the edge blend, and eventually, a self intersection and the blend cannot be created. Eventually, I would like to put a conical blend around this intersection, but it is failing similarly.

Does anyone know why this happens? I try to imagine a ball physically rolling around an inclined cylindrical rod, and I believe it would maintain 2 point contact everywhere. Perhaps it has to do with how blends are calculated in 3D?

Thanks all for the help!

 

[mmauldin]

Try playing around with blend settings to see if you get an acceptable result.

 

[kbarne]

Hi mauuldin, thanks for the reply! Unfortunately these setting work only some of the time, and greatly depends on the fillet size, cylinder diameter, and cylinder angle. For my case, these settings don't help much, and leave a "cusp" in the fillet still.

In your example, I bet if you were to increase your 0.4 radius to 0.6 or so, you would see a similar problem to what I am seeing. The fillet may create, but it looks like the curvature or something isn't consistent in the fillet.

Thanks for the help though!

 

[jerry1423]

Increase your tolerance to see in that helps.

Jerry J.
UGV5-NX1884

 

[mmauldin]

I went back and created a set spheres that represented the blend and then sectioned them. At .75in the sections overlap the center line, that is the problem.

At .25" they don't, hence the smooth blend.

I've uploaded the model if you want to play around with different radii.

 

[kbarne]

Hi Jerry,

I had tried increasing the tolerance all the way up to 0.1", and it didn't seem to help.

mmauldin,

Do you know of a way around this problem?

Essentially, here are the things that are fixed:
Cylinder radius
Cylinder angle to flat plate
"Radius" size

I have "Radius" in parenthesis because I don't need it to be a true radius. Ideally, I could use a blend that approximates a conical blend and all areas in the blend are tangent. I am thinking kind of like a free form surface or something. I am hoping that having this flexibility allows for some solution.

Thanks again for all the help!

 

[cowski]

Is it a true cylindrical face or a B-surface that is cylindrical-ish? If it is a B-surface, it might be the underlying surface geometry that is causing the issue.
Can you post your file (or a simplified example file that shows the same issue)?

www.nxjournaling.com

 

[kbarne]

Hi cowski,

I sketched a circle on the flat plate, then extruded that circle at 60 degree angle relative to the flat plate to create the cylinder. I can try uploading the file later today, but it is definitely not a B-surface.

Thanks,
Kevin

 

[cowski]

Based on your description, the "boss" created would have an elliptical cross section normal to the extrusion direction. This should not result in a B-surface, but the surface isn't truly cylindrical either (info -> object would probably report "extruded surface"). The resulting blend surface may start to get weird as the blend radius gets larger than the semi-major radius of the ellipse.

www.nxjournaling.com

 

[mmauldin]

Outside of freeform surfaces, I don't know of a practical way around this issue.

I'm going to play around with a bit more to see what can be done. Maybe by then, the heavy hitters on the board will wade into fight.

 

[kbarne]

Hi cowski,

You are right, how I described the previous "cylinder" wasn't really cylindrical. I made a new model, where the circle sketch is oriented normal to the projection direction to make a true cylinder. Unfortunately, I am running into the same issue.

Maybe the blend radius being larger than a radius in the circle is the problem? Do you have a recommendation to get a smooth fillet (doesn't have to be a constant radius in the problem area)

Thanks all.

 

[mmauldin]

After some rough experimentation, it seems you will run into an issue at about blend=1.2*boss dia at 45deg.

 

[cowski]

In the edge blend settings, if you turn on "remove self intersections" and "patch areas of complex geometry", you'll still get a pinched area, but it will avoid the self intersection error.

Try a face blend with the "swept disc" section type (pick the edge at the intersection of the cylinder and plate as the spine curve). This might give better results than the edge blend for the same radius value.

www.nxjournaling.com

 

[kbarne]

Hi mmauldin,

Yes, I am seeing similar limits to the blend. Now to figure out how to change the blend such that I can get a decent size conical blend without changing the cylinder size or angle. Thanks for the help!

cowski,

Yes I am trying to avoid both the pinched area and the self intersection error--maybe it's not possible with a true conical blend but I'll have to find something that is close but maybe deviates in the pinch area. I will give the swept disc section a try. Thanks for your advice!

 

[Toost]

The geometry definition in this case is actually pretty tricky, as you say, its a rolling ball definition, but that also means that the tangent section between the two contact points will rotate about the imaginary axis that the two points define because the plane of the circle is defined by the surface normal in each point.
In the left example in the picture, i have created "iso parametric curves" ; these curves will replicate the tangent-sections of that rolling ball. Each single curve should closely replicate a circle. ( The few i selected are true circles.)
Notice how these section-planes "turn inwards towards each other" where you get that wrinkle, the tangent sections overlap .

On the right, i created a face blend instead of edge blend, this time "a swept disc", here i can control the section plane using a "spine curve", all circle sections are normal to the spine. The spine curve used in this example is the original sketch circle. any closed curve can be used for this.
Here the wrinkle doesn't occur.
you can also define a conic section instead of circular using the swept disc option.

Regards,
Tomas