# Tool for 3d Fairing

We have the FAIR comment that will deal with a curve in 2d. Is there any tool that will fair in 3d.

Let’s say I have a set of points taken at fixed X and Z values (with varying Y values). For Each X value and for each Z value I have curve. Each point then has an curve through it horizontally and vertically.

Is there then a tool that will adjust the Y values only such that the curves will get rid of the curve squiglies but will keep the curves intersecting?

In other words, one that fair two sets of curves simultaneously.

Smooth command works for you?

Else-wise you can write simple laplacian function to project neighbour vertices to their average point

@bigjimslade I assume you are describing fairing a set of boat/ship “lines” starting with a table of offsets.

My approach is to first find the “typos”. I import the offset points and then create the curves using CurveThroughPt on the points for each waterline, buttock, sheer, rabbet, etc. I carefully look at the curves for the points which are obviously off, look at the corresponding offset, and usually can figure out that 1-7-6 should have been 1-6-6 or that 2-3-1 should have been 2-3-7 or similar. Frequently this will resolve most of the problems with the lines.

Next I create a “virtual half model” of the surface using Patch. But, important, I don’t just use Patch on the offset points. Instead I look at the shape and decide how to divide it into a set of surfaces. Then I create a set of boundary curves for the those areas using the relevant offsets. Sometimes the boundary curves will extend off of the surface. Then I create a set of starting surfaces for each area, typically using EdgeSrf or Sweep2. Then I use Patch on each starting surface with the interior offset points. I may rebuild the surface with fewer control points and apply Patch again. PointDeviation is used to check for the fit of the surface to the offset points. Next I fine tuning the surface by moving individual control points.

Faired lines are generated using DupEdge and Contour.

1 Like

The data I have is very fair. There were a few clunkers but not many.

Here is the problem I am trying to solve. With the data rounded to 1/16" there is the zig-zag problem.

Something like:

8-8-1+
8-8-1
8-8-1
8-8-1
8-8-1
8-8-0+

Is supposed to be smooth but does not come out when rounded.

I did a faired hull from the data, going back and forth between the X and Z directions. The problem I now have is that internal members are no lining up. I have parts that are supposed to have a specific location on the shell but they are often noticeably off because of fairing.

What I was hoping for was a one-shot fairing that would do both directions such that the points remain within 1/16" of the tabular data but would have smooth real values.

I doubt you will find it.

What is the overall size of the vessel? Where the offsets taken from drawings, full sized lofting, or an actual hull. The accuracy of the offsets may be less than the precision.

In what order was the actual vessel built? How closely did the physical vessel match the drawings? Perhaps the position and or shape of the parts was altered during the physical build until they lined up with each other?

Molded length is 887’ 3".
The figures I am using for all the parts are from the full sized mold loft tables. Al the parts I am trying to match up come from same set of mold loft tables.

What happens is the fairing gets the sharpest points. A part will match up well until it hits a sharper turn in the hull then there will be a visible difference.

What’s the beam - 100’ or so? If it’s 100 feet then 1/16" is about 1 part in 20,000 or 0.00005 of the the beam. I strongly doubt that accuracy was held in building the vessel.

Depending on how the frames are built and the hull is plated some curves may not be “fair” everywhere.

For example start with a straight line and an arc joined such they are tangent. Pick a finite number of points along the joined curve. Now try to create a degree 3 (or higher) curve through those points. Wiggles are guaranteed. The solution is to use multiple curves, not a single curve.

The big challenge is managing all the data sets. I have tried to separate out all the shell data sets.

If I just use the table of offsets to draw the frames, I find (eg) that longitudinals are outside the shell. This is mostly a problem near the end of a frame where the TOO data tends to create curves that cut the corner.

Then these other data sets have points that get so close to the TOO points that they totally screw up the frames, even when they are as close as possible to the correct smooth location within 7-7-7 precision.

In addition to the numbers in the table of offsets I look closely at the line drawing and construction drawings and try to figure out how the curves, particularly the station/frame curves, were drawn and how the hull was to be constructed. For example if the frame was designed to be a series of straight sections and arcs then using a single degree 3 NURBS curve will cause problems.

Its interesting that the plans indicate straight lines at certain points on the hull. Yet, the table of offsets clearly do not show straight lines at these points.