Delft Hull Series

Here's cf on Sysser 50 at the centerline, botom is 0 and top 0,0051 on the scale. This is at 2,44 m/s (Fn 0,55), with the free surface. Integrated Cf is 0,0044, all turbulent. Here cf is steadily increasing until about the max canoe body depth and then slightly decreasing towards the transom edge.

 

I can not believe what I'm seeing. Why is there a Cf-value above the waterline and why are the colors dark red, the highest values? It should be zero above the WL.
To me it looks as if the scales are flipped.
In the diagram for Cf at the centerline it also looks as if the positive direction is downwards. I have attached the flipped version of your diagram. This version would make sense.
It starts with the stagnation point at the bow (highest Cf), then there is a short part of laminar flow (low Cf), the transition to turbulent flow (a steep rise in Cf) and a gradual decrease of Cf like on the flat plate. It tends to go to zero at the end (as in my diagram), but then comes the stern wave (which I don't have in my simulation) and causes a rather high Cf on the transom overhang. At the transom itself it looks like separation with Cf=0.

 

Uli, you cannot always believe what you see ;-). The curve is not flipped, if you compare with the surface plot they correspond to each other. You should start reading the curve at about the bow knuckle, the part before that is in the bow above the air. The colours on the topsides above WL are in the air, they are nonsense, the free surface part is still a work in progress. The surface plot is shaded by (the approximate) surface level so what you see is really under water. There can be no laminar parts in the flow as it is all turbulent.

Here's still another plot, with cf integrated (custom field1) over points in the bottom, and shown as a graf. The shape is pretty similar to the centerline shape. Here, the blue topsides also show the "real", dry topsides. The holes in the curve are probably due to no points at those locations.

What is your velocity distribution along the equivalent body of revolution like?

 

Here comes the velocity for Sysser 50 (attached).

Mikko, even at the risk of making you angry, I still think that there is an error in your software. At least in your post-processing.

You use LES with a wall function. In other words you do not resolve the velocities within the boundary-layer, vertical to the wall on a dense gid, instead you assume a velocity distribution, typical for turbulent flow. This makes sense because it significantly reduces the computing time.

The classic similarity rules for such a distribution are the law of the wall and the law of the wake. If the shear stress (or cf) increases along the wall, then according to these rules and v. Karman's momentum equation this is only possible if the boundary-layer becomes thinner. A reduction of the b.l.-thickness along the wall in flow-direction is against all laws of physics and contradictory to all textbooks on fluid dynamics.
The only region where the thickness of the b.l. decreases is in the immediate vicinity of a 3D-stagnation point, where the girth-length of the body has an extremely sharp increase when moving in flow direction, because then the b.l. is "stretched" sideways.

 

No way you will make me angry, Uli! I do agree with you, this is odd, the Sysser 50 LES sim would have the cf values flipped also compared to the Sysser 26 RANSE plot (which again has that waterline shaped oddity at the back). I do think it's a postprocessing problem only, as the results for Sysser 50 compare so well with experiment, even more favorably than the validations presented in the Chalmers Sailing Yacht Transom Sterns-paper, on Shipflow. And there's obviously plenty of validated aerodynamics cases where cf is a crucial factor. But something to look into for sure, I will try to look at some other case. I have not looked at cf earlier, not much of a factor for sails.

Your velocity distribution does look rather similar to mine (mine is for 2,88 m/s, at the centerline).

 

Is this a volume-of-fluid approach, or a multiphase Eulerian solver? I'm hearing from people very experienced in marine CFD that the VOF approach is prone to streaks of low skin friction, due to air entrained in the water that passes under the hull. This is a VOF problem, not a physical problem, and it arises when the cells at the hull/surface intersection have less than 100% water in them. The lower density water then flows under the hull and the air can't escape.

The reason I ask is your skin friction plots look a bit streaky and I wonder if this could be a reason.

 

The RANS sim is one phase only, with symmetry at the "water surface", so the streakiness cannot be due to "numerical air entrainment". It is a well known problem with VOF, with different correction approaches with different solvers.

 

With the LES/free surface solver, I am getting a bit different "streaks" under the transom... It would be exiting if these were really physical, but I think they are numerical (or graphical), maybe simply due to fairly coarse resolution and projecting to a surface close to horisontal. Since this is particle based, Lattice Bolzmann, not RANS, the source of the streaking is probably not the same as in Numeca's case.

 

But there is definitely an issue with the Cf plotting as Uli mentions. Thank you for pointing this out, Uli, I will discuss it with the developer. Cf seems to be tricky, when I run the RANS sim for Sysser 26 again with a higher resolution, the Cf plot again looks different :-(. In one phase in the air the Cf plot seems to make sence, like in this sail simulation.

 

Here's a presentation a made last spring around the Dehler 33, with keel & rudder included. It's in Finnish, but I think you can read the plots - they are for a full sized boat. It you want explanations about a particular slide, I will be happy to translate.

 

Thanks for the figures, Mikko. The streaking is something I've been told about, but hadn't seen. It's of considerable interest, since I hope to be making calculations like these myself in the not-to-distant-future.

 

Wavemaking length

I added the influence of the transom overhang to the regression and discovered a phenomenon that I can not explain.

The Froude number is defined as FN = V / SQRT(g*L) where "L" is the wavemaking length.
Normally the length of the waterline at rest is taken as an estimate for "L". One would assume that the actual wetted length, that includes the influence of the stern- and bow-wave at the given speed, would give a better approximation of the wavemaking length. This is not the case as figures 3 and 4 in the attachment indicate.
My first idea was, that it could be caused by the hollow behind the submerged transom, which would create a fictitious increased waterline. So I extrapolated the section area curve beyond the transom to calculate the wavemaking length that includes the hollow behind the transom. The resulting correlation was even worse.
It seems that the waterline length at rest is the best guess for the wavemaking length.

Any idea for an explanation?

 

I'm not that surprised by the worse approximation.

(I'm not quite sure how you estimated the hollow length).

I still maintain that the hollow behind the transom cannot sustain
a pressure and therefore it cannot be a wave-maker.

There are some similarities of the flow to the hollow behind a
cavitating, truncated strut but that doesn't necessarily mean
that the wave resistance can be predicted using that type of flow.

Many researchers (e.g. Patrick Couser, Prof. L.J. Doctors and others)
have shown better correlations with experiments using fictitious
hollows behind fully-ventilated transoms, but most admit it is just
an "engineering" solution. I have found that the method sometimes
improves predictions, and sometimes it doesn't help at all.

The thesis by Robards (a student of L.J. Doctors) used the technique
on a large number of transom stern vessels. Reasonable agreement with
experiments was demonstrated, but the method required separate form
factors for the wave resistance and viscous resistance. Furthermore,
the length of the hollow was difficult to measure, and in many cases
it wasn't possible to accurately gauge where the hollow ended.

Doctors still uses the technique in his work, but the hollow length
in his papers is either not defined precisely, or the coefficients
in the formulas vary from paper to paper and from hull series to
hull series.
Couser et al used a hollow length that varied with Fr, so it is not
surprising they got better agreement, but their method for NPL hulls
does not necessarily apply to other transom stern shapes.

 

Leo, I am puzzled by the following:

Froude number is the parameter that creates hydrodynamic similarity. All curves should be similar, if plotted against Froude number.

1.) FN based on still water length - good similarity (figure 3)
2.) FN based on actual wetted length of hull - similarity lost (figure 4)
3.) FN based on hull length + hollow behind transom - worse than figure 4

In case 1.) when the waterline at rest is used, the transom is always out of the water.
Case 2.) which includes the transom overhang at higher speeds is the more "realistic" case.
Case 1.) gives a better similarity than case 2.), this is the surprise.

 

This is only true if the hulls are close to being geometrically similar. The closer the hulls are to having identical shape and differing only in scale, the closer the wavemaking resistance vs Froude number curves will be. The three hulls with variations in rear overhang have different shapes in contact with the water when underway. So it is not a surprise that the wavemaking resistance vs Froude number curves have different shapes.