Boat hulls with smoothly curve stems - how to model

This is also posted on the Rhino forum Boat hulls with smoothly curve stems - how to model https://discourse.mcneel.com/t/boat-hulls-with-smoothly-curve-stems-how-to-model/96507

Round bottom boat hulls frequently have stems which curve smoothly into the keel. A NURBS surface with four sides is not a simple fit to such a shape.


I know of three methods to model such a hull side using a single NURBS surface on each side. Each has advantages and disadvantages. A Rhino .3dm file with an example of each method is attached. The surfaces are very close to each other in shape. Note that the orange curves are isolines but do not indicate the number of control points used for each surface.

Perhaps the most obvious method is to use a three sided surface (actually a four sided surface but on side has zero length). This allows direct control over the shape of the stem. The shape near the upper corner where the isocurves converge can be difficult to adjust and maintain fairness. The surface cannot be simply extended at the upper corner. This method can be a quick way to create a surface for hydrostatic calculations and similar.

A second method is to split the stem curve to create an additional side. This also allows direct control of the shape of the stem but care is needed to ensure the stem curve has the desired continuity across the split. This method eliminates the converging isolines in the upper corner. The tradeoff is near the stem curve split the u and v isocurves become almost parallel, and are parallel at the split. This make adjustment and fairing of the surface shape in this area difficult. The surface may have problems extending at the split in the stem. This method can work well for creating a surface from a set of lines which will not be modified.


The third method is to extend the surface past the stem and maintain the four sides. then trim the surface to the stem. This trimming can frequently be done with a plane on or parallel to the centerline of of the boat. This method does not provide direct control over the stem shape like the previous two methods. The stem shape is determined by the shape of the untrimmed surface. The resulting surface can be easier to adjust and fair than surfaces using the other two methods. Also while the trimmed surface usually can be extended without any problems and then trimmed as desired.

 

I generally try to put any irregular facets in a really flat area. What works for me is to have a ribbon of fairly narrow facets running along the center line, and then scab any of the three methods to that ribbon. This gives me a manageable layer to find my hydrostatic targets by tweaking the entire center line, and lets me hang rudders and skegs and keels from this. I can create a family of these ribbons with different appendages and save them as parts. However you do it, look carefully at how the actual facet is distorted from the control point grid. Changing the direction of principle curvature within a facet doesn't work well. So the split method wants its split in an area of least principle curve. Or you need to gradually warp the grid to produce smaller facets there. There is nothing really special about the sort of facet in the split type. It, like all others, will render best if there is very little torsion on opposite sides. That is what is tricky in the bow. The same thing happens at the keel joint in a sailboat, but it is easier to find a nice flat spot to land the oddball facet.

As a practical approach, you can put the oddball facet where the hawsepipe or bow thruster goes and then convince yourself it isn't terribly important Sometimes, Click and Clack's advice to place a small piece of black electric tape over the offending visual is the best way to proceed.

 

philSweet - I was having trouble understanding your use of the term "facet" but then I saw that it looks like you use FreeShip. It's been a while since I tried FreeShip and my recollection is it uses a type of sub-division. I use Rhino which directly uses NURBS. It sounds like similar issues arise with both types of surface generation.

When splitting the stem I usually place the split where the curvature of the stem is maximum. This limits the area where the u and v directions are close to parallel. But I usually only split the stem when I need a quick surface and the stem has a distinct "bend". Those type of shapes typically don't work well with the 3-sides approach.

 

In this case, subdivision surfaces vs nurbs is a distinction without a difference. Since they are both set up to render circular cylinders correctly, they suffer from the same sorts of rendering issues. They can be set up differently, but as employed here, they ought work the same.

 

A quick look into Freeship/Delftship showed that they are not general sub-division modelers which allow arbitrary arrangement of input points. Freeship/Delftship use rectangular arrays of points similar to NURBS. So I agree there will be the same or similar issues when modeling a bow with a smoothly curving stem in a NURBS modeler or in Freeship/Delftship. My understanding is Freeship/Delftship do use sub-division type algorithms to generate surfaces from the input points. This differs from NURBS modelers which use algebraic equations to generate surface from the input points. The end results can be equivalent.

As an aside there are more general subdivision modelers such as Blender which allows arbitrary arrangements (other than rows and columns) of input points. These types of modelers should not have the same issues with modeling curved bows.

What is the relationship between rendering circular cylinders and designing/modeling the bow of a boat?

Rendering is not the primary factor which influences how I arrange surfaces in designing or modeling a boat. Other, usually more important factors can be ease of creating the initial surface(s), number of control points required (fewer are better), ability to efficiently modify and fair the surface(s), and ability to extend the surface(s).

 

Another method is to split the sheer to create four sides for the surface. This allows direct control of the shape of the stem. I do not have any experience with this method. Updated file with an example of this method is attached.

 

I think this issue not depends from software you use, but from surface geometry.

 

The issue occurs due to the the combination of the surface geometry and the use of surfaces which inherently have four sides although one or more sides may be zero length, such as parametric surfaces including NURBS surfaces. It does not depend on the details of the software.

Some types of mesh modeling and sub-division modeling are completely free form and would not have this issue.

 

Four boundaries surfaces usually had some flatness around the corners. To avoid it you have to play not only with inner control points of the surface, but also with tangency of boundary curves in the corner.

There is some more information, if it interesting - https://www.shmexpert.com/post/the-devil-is-in-the-details

 

All Bezier and NURBS surfaces have four sides although some of the sides may be zero length. Are you saying all Bezier and NURBS surfaces have "some flatness around the corners"?

Or do you mean that a particular method to create surfaces using four edge curves can result in ?some flatness around the corners"?

Or something else?

 

Bezier or NURBS surfaces has some flatness in the corners. Corners of this surface patches are little bit more special. Surface patch comes exactly into the corner and surface normal vector in corner point is a boundary’s tangent vectors multiplication. So you can modify any control in your surface patch, but local plane around the corner will be any way defined by tangent vectors of boundary’s in this corner. If one of the boundary became a point, it is a singular point. In this point you can’t for ex. calculate normal vector. Some of the programs do not like it or try to avoid it somehow. All this is not depends of method of surface patch definition. It is just mathematically nature of this type of surfaces.

 

My favorite method to design the stem is - don't! Instead of creating a shape for the stem, define the shape of the hull up to the forward perpendicular. The extrapolate the topsides forward, past the center plane. Take the intersection of the port and starboard topsides, and you have your stem. The stem becomes a byproduct of the hull shape instead of a driver of the hull shape.

The advantage of this method is it guarantees the stem will be fair and there won't be any weird contortions of the shape as it gets to the ends of the stem.

 

This worries me a little. I've had some spectacular failures trying to extrapolate surfaces. Can you show us some examples of this approach?

 

I also don't design the stem shape explicitly. I create a target shape for the stem, then design the hull using method C above with the hull surface going past the stem. The target stem shape is used as a reference but the shape and fairness of the hull surface takes priority. The final stem shape is created by trimming the hull surface at the centerplane. This method provides explicit control of the hull shape forward of the forward perpendicular while keeping the hull surface fair and without contortions to fit a pre-defined stem shape.

 

But you still have the problem of how to model the surfaces.