How do you convert a mesh to a smooth surface?

rhinocommon
rhino5

#1

Hi. I am new on this forum and hope to learn a lot and collaborate on Rhino with everyone.

I have a pretty cool mesh object of an IL-86 (Russian airliner from the 1980s). I want to convert the mesh to a smooth surface so that I can refine it and create a mesh with smaller size panels suitable for 3D-printng.

I tried tracing curves from each point, which is manageable process. However, I began to wonder how I was going to use these curves to form surfaces? If I trace horizontal curves down the length of the fuselage, then trace vertical curves for cross sections, I wind up with a very good curved outline of the fuselage. The problem with this technique is that the vertical curves and the horizontal curves do not necessarily intersect with each other. Therefore, its not really conducive to creating surfaces. Maybe this is the proper technique, but I am doing something wrong?

Next, I downloaded T-Splines, and ran the command: “Convert Nurbs or Mesh to T-splines”. It did transform the topology to T-splines, but I am not sure what to do next. After running the command, it seemed to result in a mesh that looked exactly the same as what I started with – except now it was composed of T-Splines.

I was hoping that this command would take all points and generate a smooth curve that represented a best fit for a particular row or column of points --> then join them to result in a smooth surface.

I have heard that creating converting a mesh to a surface can never be fully automated. Which is OK with me. I was just thinking that it would give me starting point, that I could use as my base; then refine it manually.

I don’t see how I can post a picture, otherwise I would post a screenshot.

Thank you in advance for any help or insight that anyone can offer.


(David Cockey) #2

To post a picture click on the vertical arrow icon above where you type your post and upload the image…

Method for creating a smooth surface from a mesh depends in part on:

  • shape of the object
  • density of the mesh
  • accuracy and level of detail needed

Also very helpful in modeling an object is to consider how the object was initially designed and made.

If part of the fuselage has constant section then you could create a section curve and extrude it but do not try to include the fuselage to wing blend. The wing could be created using several sections with lofts between the sections, again not including the fuselage to wing blend. A blend surface between the fuselage and wing can then be created. The rear of the fueslage is likely to be a cone with a blend to the main part of the fueselage.

A common mistake of novices is to try to model the shape with a few complex shapes. Frequently a few more but simpler shapes will give better results.

Do you know enough of Rhino to create a shape similar to the airliner without starting with Rhino?


(David Cockey) #3

Looking at some online photos of the IL-86 the rear of the fuselage does not appear to be a simple cone. It could be created with several cross section curves, a top and bottom profile and Sweep2. Another possibility would be several cross section curves, top, bottom and middle profile curves and NetworkSrf.


(David Cockey) #4

How accurate and detailed is the mesh model?

How accurate and detailed do you want your model to be?


#5

Thank you. Not sure why I didn’t figure out how to upload the image this morning. Here are some pix of the base model.

Do you know enough of Rhino to create a shape similar to the airliner without starting with Rhino?
Yes, I can create a basic shape of the airliner.

How accurate and detailed is the mesh model?
I would say it is average. In its current form, it could not be printed with 3D technology…It would be too tessellated. For other purposes, such as illustration, it might be OK. Not sure.

How accurate and detailed do you want your model to be?

  • I would like to print it, but that opens a whole new topic. I would say it should be printable with a smooth surface – no tesselations to the naked eye. Some amount of sanding is OK, but not too much. Perhaps, it might be better to think of it as making a cast that would then be used for injection molding. I am still investigating that part of this project. How much would something like that cost? I’m not going to spend a lot of money on this, but I don’t want a toy either. Just want something that is detailed. A show piece to impress other people. The quality that you would expect to see on exhibit at a trade show. No landing gear. No antennas. Just a souvenir piece with a really smooth paint job and a true-to-life shape/outline/profile.

(David Cockey) #6

Given your use I would create a model of the airliner using the mesh as a guide.

There is no need to use every mesh vertex to create the model. Use as many vertices as needed.

Use PointDeviation to check how close your surfaces are to the mesh vertices. In Rhino 6 WIP you can select the mesh as the source of points to test. In Rhino 5 you will need to use ExtractPt to first extract the mesh vertices before using PointDeviation.

If the fuselage has a circular cross section a corresponding circle can be created using Circle with the 3Point option and selecting three vertices as the three points.


#7

OK. Good point. I’m not familiar with PointDeviation or ExtractPt. I will look them up, but I think I understand the general idea.

Thank you. This gives me some direction to go by.


(David Cockey) #8

Another tip - Contour and Section can be used to create polylines from a mesh. FitCrv can then be used to create a NURBS curve from the vertices of the polyline. Try to determine the simplest curve (fewest control points) which is close enough to the vertices.

Also, do you know the source of the data used to generate the mesh? Was it created from design bureau drawings or from a scan of an actual aircraft? Or was the mesh created by someone using some photos and published dimensions?


#9

Look up t-splines retopo snap on YouTube:


#10

Do you know the source of the data used to generate the mesh?
A friend of mine at work gave it to me. I saw him working on it and thought it was interesting, so I asked him if I could play around with it as a way to learn Rhino. I will have to ask him when I get into work tomorrow.

Contour and Section can be used to create polylines from a mesh
?
Are these commands? Where do I start with this? (I think I was thinking along the same lines this morning…I though "Why can’t I just generate cross sections of the fuselage - maybe 1 ft apart, make sure the points used to generate the horizontal lines are the same reference point on each cross section (i.e.; don’t pick the points at random – be consistent in which points I pick.) However, I am not sure if that technique, alone, will work on the nose or the tail. Also, the area where the cockpit glass is flat [no curvature] will be tricky. – especially in going from compound curve to a flat area.

If you want to see what I am talking about, look at the 737 cockpit glass. It is 100% flat. However, the curvature of the fuselage blends into this area…Its actually pretty interesting if you study it (for instance, there is a crease in the air frame just above the cockpit window that sort of gradually fades into a compound curve – pretty neat, but hard to reproduce). Interestingly, I think the 787 was designed with cockpit windows that allowed for compound curves and thus able to conform with the curvature of the surrounding air frame.

Also, 1 foot apart might be overkill? However, I know Boeing defines station numbers for each section of the fuselage just like this (1 foot per section I think). I was able to see some blueprints one time of the 737…they had a lot of stations running vertically down the fuselage. I think there is something about station 42 that is particularly complex, because that was the context that the forum thread was referring to. It was a while ago. I think I read about it on airliners.net.

Look up t-splines retopo snap on YouTube:
I will watch it tonight… Thank you.


(David Cockey) #11

Contour and Section are standard Rhino commands.

[quote=“dvanauken, post:10, topic:43171”]
Are these commands? Where do I start with this? (I think I was thinking along the same lines this morning…I though "Why can’t I just generate cross sections of the fuselage - maybe 1 ft apart, make sure the points used to generate the horizontal lines are the same reference point on each cross section (i.e.; don’t pick the points at random – be consistent in which points I pick.) [/quote] A very common approach to create a surface from a mesh. Create a set of section curves and use Loft for a surface. Sometimes it works well if the shape is sufficiently smooth such as a sailboat hull. But for a shape like an aircraft fuselage waves are likely to occur. Also if lots of sections are used the resulting surface will be “heavy” with a tremendous number of control points and difficult to impossible to smooth.

However, I am not sure if that technique, alone, will work on the nose or the tail. Also, the area where the cockpit glass is flat [no curvature] will be tricky. – especially in going from compound curve to a flat area.

[quote=“dvanauken, post:10, topic:43171”]
However, I am not sure if that technique, alone, will work on the nose or the tail. Also, the area where the cockpit glass is flat [no curvature] will be tricky. – especially in going from compound curve to a flat area.

If you want to see what I am talking about, look at the 737 cockpit glass. It is 100% flat. However, the curvature of the fuselage blends into this area…Its actually pretty interesting if you study it (for instance, there is a crease in the air frame just above the cockpit window that sort of gradually fades into a compound curve – pretty neat, but hard to reproduce). [/quote]
The flat glass is best modeled as separate flat surfaces. Consider how the shape was originally designed. Then use that as a guide to how to construct the surface model of the shape. [quote=“dvanauken, post:10, topic:43171”]
Also, 1 foot apart might be overkill? However, I know Boeing defines station numbers for each section of the fuselage just like this (1 foot per section I think). I was able to see some blueprints one time of the 737…they had a lot of stations running vertically down the fuselage. I think there is something about station 42 that is particularly complex, because that was the context that the forum thread was referring to. It was a while ago. I think I read about it on airliners.net.[/quote] Yes, 1 foot spacing would be overkill and result in a surface which is about impossible to edit without rebuilding as a simpler surface. Why model the constant section portion of the fuselage as anything other than a simple constant section surface? As far as the area with the transition between the fuselage and wing goes it is best modeled in the same manner as it was originally designed. A constant section fuselage, a tapered portion of the wing with straight leading and trailing edges which extends into the fuselage, and then a transition surface or surfaces between the fuselage and wing.


#12

Hi. I followed your advice…

  1. I cleaned up the base model a little bit more. For example, I removed the wing - body mating structure…Basically, it is just a long tube with compound curves in the front and back.
  2. I traced horizontal and vertical profile curves
  3. I did a FitCrv on 6 of the curves. 3 vertical and 3 horizontal.
  4. I used the DIR command to ensure that the vertical curves ran one direction and the horizontal curves ran in the opposite direction.
  5. I ran the NetworkSrf command, chose the 6 curves, then followed the prompts to select the first, second, third curve in the vertical direction and did corresponding action for horizontal curves.

The result was a message from Rhino that said:
Unable to use this curve network.

I am attaching two screen shots and the my .3dm file.

The curves seem to oriented correctly

Thank you in advance and I really do appreciate the help given so far.

I chose three lines, just to keep it simple.

Here is the 3DM file

IL-86-BASE-10.3dm (446.2 KB)

(David Cockey) #13

Your curve network does not meet a requirements for NetworkSrf.
All curves in one direction must cross all curves in the other direction and cannot cross each other.
http://docs.mcneel.com/rhino/5/help/en-us/commands/networksrf.htm’’

Your contours are polylines with kinks, not smooth degree 2 or higher curves.

You may be trying to work more precisely than the resolution of the mesh you are starting with.

It appears that you are trying to develop a single surface for the nose which includes the flat windows and associated creases. That is almost certain to not work.


#14

Try taking three of four sections in both directions and building the skin that way. Ignore all the windows and smaller details at first, just try to get the overall shape you want. You’ll find that fewer curves for sweeping or network curve surfaces will return a better result.

You’ll be able to then work from that surface to get closer to the original mesh. After that you can add all the apertures and details

Have a look at some hull modelling tutorials and use the same process.

Regarding the 3D print, you’d have to print in fine or ultra fine detail in order to remove any steps, most additive processes have some form of layer visible at around 100 microns. Based on a 2-3mm wall thickness it depends on how big the fuselage is regarding the cost.

Injection moulding tooling would be machined and you’d be looking at thousands of pounds/dollars for that alone, before you knocked out the first part. That’s only really for the low quantities of thousands and above…

Look at the strong and flexible range here: https://www.shapeways.com/ you can easily use a decent filler primer with little sanding to get a good part as a show piece. The bounding box of the machine will give you an idea of what size you can make it, it has all the wall thicknesses and embossed detail sizes etc on the materials page.

To roughly cost it make an ellipsoid model at the size you want, shell it to 3mm - mesh it and export it as a.obj to shapeways and it’ll give you a ball park idea of cost. It’s priced on volume so I’m guessing the slight deviation in shape and with apertures etc the cost won’t move much.

Andy


#15

I figured it would be too expensive for injection molding…I have been using shapeways.com for experiments…the result isn’t totally smooth, but as you said, I can use something like Mr. Surfacer and sanding to smooth it. Hopefully, the 3D printing technology will continue to get better with each successive year. Shapeways seems to be the best value for the money. I have looked at other services, but they seem to be more expensive than Shapeways, and they don’t offer as many options for material. I have been using the strong and flexible range for test prints, as you mentioned. Once I get it the parts finalized I will probably switch to High Definition Acrylate for things that require detail, like engine fan blades for instance.


#16

Hi, I know it has been over a month now, so I hope it is Ok to keep using this thread. Anyway, I have played around with Rhino now and have to say I really like it. I have also tried other things like Fusion, Inventor, and some others that I can’t recall right now. Rhino3D, at least for me, seems like it is the easiest; most intuitive; and it seams to have the most features. I especially like the comand line.

In one of the replies, it was suggested that I use T-Splines for reverse engineering. I watched the video, and then downloaded the trial plugin. The reverse engineering capabilities were awesome! I proceeded to click on the “Buy” link. To my disappointment, it seems that it cannot be purchased anymore, so when my trial period ended – that was it. (That is why I tried Fusion and Inventor; but I can live without T-Spline’s reverse engineering – It was a nice-to-have feature – but I still prefer Rhino.)

Anyway, I have some more questions…I am focused right now on the cockpit windshield section. I am trying to reproduce a crease; I am trying to figure out how to make lines come together as straight sections rather than rounded curves. I took some screen shots of some problem I am trying to sort out…

In this picture, I created 4 curves that outline the cockpit windshield (on one side of the fuselage). The bottom curve is composed of two straight lines (then joined). The top curve is a smooth curve. I used NETWORKSRF to generate the surface pictured here:

As you can see, the resulting surface has issues where I outline a red ellipse. The surface is rounded at intersections, but the intersection needs to meet at straight lines.

Next, I used the SWEEP2 command, using the same set of curves to generate this surface:

This technique seems to generate the best result. The crease formed by the bottom edges does fade away as it transitions from bottom to top, but not totally.

Question: How do I make it fade away completely? (I think the crease should be totally gone as it passes no more than 2/3 of the way up the surface.)

Next, I tried the TWEENCURVES command. I was thinking that I could use PATCH or LOFT in conjunction with TWEENCURVES. However, as pictured here, the result of TWEEN curves doesn’t generate curves that I can use for PATCH or LOFT:

(The cockpit windshield surface is in CYAN. There are three tweened curves between the top and bottom – also in CYAN.)

Last, but not least, I tried this technique: I divided the cockpit area into several sections and then used the NETWORKSRF COMMAND to create multiple surfaces. I used a different color for each section, as pictured here:

I am not sure if this is the best way to do this? Any comments Yea or Nea?

Anyway, I know this was a long reply…any help, guidance, comments, suggestions are appreciated. Thank you for your help so far.

Doug.


#17

T-Splines if long gone now. Have a look at Clayoo 2 - that should help you out. it’s another steep learning curve if you’re not au fait with sub d modelling but well worth looking into.

Andy