HELP. Trying to create surfaces on irregular curves

I was given a task to create surfaces about a port. The lines were made from a CMM machine and are rather irregular. I have tried multiple commands including loft, patch, surface from network of curves, and sweep 2 rails. The sweep 2 rails I have the most success in. My end goal is to have a rather smooth and precise surface that I can save and export the file to SolidEdge where I will be able to use it to cut out these ports.

Any help would be much appreciated.

Intake Port: (not able to make a closed sweep)

Exhaust Port Back:

Exhaust Port Front:

SolidEdge Final Goal: (exhaust port semi-working)

I’m not sure I understand your question. Do you want the surface to be smooth? Or ?


Yes. Very smooth and flowing. Rounded completely. The exhaust port that is shown in the pictures, I was able to do a sweep 2 rails. I selected the top “circle” then the bottom “circle”. I then selected all the lines that connected the two “circles”. The surface that it gave me was rough and seemed to be pinched at points.

Thanks for your response Bill!

Hi Dylan - there is not much anyone can do without the curves - can you post a file?


exhaust port.2.3dm (67.9 KB)

Here is one file.

Thank you Pascal!

Hm… well, clearly you’ll need to re-interpret these - do you have the original digitized(?) points? That might be easier to deal with. I don’t see how you can come close to those curves with a clean surface, but the points might give a clue.


This is all I was given to try and make.

Wow… how do you know how close you need to get? It is not too hard to make something with that general shape that is smooth but the chances of it being a viable intake port seems to be practically none, to me.


Maybe you need to show your point supplier what his/her points look like when visualized. Perhaps they will realize they didn’t do a very good job and redo them.

I can’t imagine any real exhaust port would look like that.

Pascal, I was told to be within .0150 in. I was trying to make things easy by using the points provided as these are from a head that is long gone and hard to find.

AIW, yes I understand the points and the way this was made up with the CMM is not the greatest. This was one of the first times that the guy had ever run one. He does understand what he did wrong and will learn how to do better.

The option to redo the points is not possible as the head is long gone and the option to use a CMM is very limited for this project.

Thank you all for your feedback!

It looks like your best bet would then be to use the two end curves and put the rest of the points through your common sense processor to come up with your best guess of what the original designer had in mind. Maybe even rationalize the end curves too.

I agree with AIW. Common sense should be your guide. How old is the head you are trying to duplicate? What technology was available when that head was manufactured? What is the goal of the “port”. Here is a carpenter way of approaching this problem. I took both of the end curves, Laid them on top of one another, using SETPT to move those curves into a common plane. I then “averaged” those curves by drawing lines at 0 degree, 90 degree, 45 and 270. And trimmed them using the end curves, finding the mid point of those trimmed curves i placed a point. I then drew a curve though those points. Thus “Average” of the two end curves. Move and rotate that new curve half way between the end curves. Loft end curves and new curve to get a best guess of the shape needed. Here are some views of my results. If I can figure out how to share them…

1 Like



Can you make an assumption that the CMM points may have under-read but cannot have over-read? If so you can try fitting attempts at a smooth port over the wavy model and check that they graze at the outermost ridges of the latter.

If you don’t have access to an original head, how do you know where the ports begin and end? - they have to interface with other components which will be in set positions. Your end curves are not planar but presumably ought to be to in order to mate up cleanly.

Can you find pictures of the head in technical manuals or online? Even if you can’t measure they will help you visualise.

1 Like

Is this a port from an engine, compressor or similar with a valve at one end of the port? If so the inside of the original port probably had a protrusion where the valve stem passed through the port.

Also, the input curves look like interpolated curves based on a few points for each curve. The shape of the curves may be very different then the actual shape of the port.

Yeah… - that is why I asked for the digitized points but it seems those are not available… I do not see a good way out with this as the only input…


Thank you everyone for the replies. I took a lot of everyone’s input and kind of did it a little differently. What I did was, I saved it as a STP file. I then opened my SolidEdge, and opened that file. It brought in the sketch along with a parasolid surface. I took the long route to try and match it up and created a plane every .25" along one of the curves.

The next thing I did was that i created a 3D point that was coincident with each line for each plane. Then created a curve on each plane that connected each of these planes which I think there was about 20 or so which it actually felt like 1000 planes. lol. But finally created a loft of these sketches

My supervisor was more than happy with the results and he said it was a lot more accurate than expected. I probably could have done this all within Rhino, but I am more familiar with SolidEdge. I originally was asking for help to do this all within Rhino and make one less step in the conversion process.

Again, thank you to everyone for their input. The help on these forums is above and beyond! Such rapid response time and great feedback that’s truly honest.