Delft Hull Series

Have you read the first sentence of the original post?.......or the rest of the thread?

 

I have, and it was only recently restarted by Uli. And I have been trying to keep my own posts on topic, which is commentary on his recent paper. I would cautiously suggest that a discussion of alternate drag strategies, or resistance in general, or pretty much anything other than Uli's paper, be moved to a new thread.

 

Maybe it was not clear in my previous post that I was suggesting an alternative sequence of actions to select dimensionless parameters.

Uli suggested in his paper that there are hull shapes that are within the range of the Delft experiments yet are effectively extrapolations due to the particular combination of shape parameters.

Said another way, my suggestion for selection of parameters is to first select those which can represent the variations within a series. Then for each series create a representative hull (NewParent). Then choose parameters that can be set to create the variations between these new parents. This two step parameter selection process might automatically avoid cross-correlation problems.

 

It took me over one year of intensive work, several iteration loops with numerous dead end streets and lots of trial and error to achieve the current result. To master such a workload one has to have a lot of confidence that the outcome is worth the effort. A creative idea without a mathematical foundation is not sufficient.

There is no hint in any textbook on regression analysis that a manipulation of the initial database will improve the results. I am curious to see the results of your regression analysis.

You can use Exploratory Factor Analysis to find out how many major influences there are that determine resistance, but for me it is impossible to find a physical explanation for the combination of geometric properties that are linked to these influences.

 

Not sure if adding the trimming moment is the best approach. With increasing speed, at least models with significant aft overhangs and rocker tend to trim bow up (a lot, up to 4 deg), opposite to the effect of adding the trimming moment. Could you not have used the recorded trim angles at each speed, for a more realistic wetted area at speed?

 

Trim and sinkage do not necessarily mean that the wetted surface area will
change in a consistent manner. If the water level around the hull sinks and
trims too, then the wetted area will be close to its static (i.e. at-rest) value.

 

I do not know, of course, what level of accuracy is desired in the calculations but I do not think this statement could be generalized for all types of hulls. You should have to determine for which shapes of boats that may be right and what % change does not significantly alter the results.

 

Wetted surface of the SYSSER 50 model at 2,88 m/s (Fn 0,65). First one in rest at but with pitch fixed to the Delft test measured pitch, second at speed with the same pitch (pitch fixed but free to heave). The sahape of the wet area is quite different, but I'm not sure it is any larger at speed.

The run is the same as in the video

 

Thanks, Mikko.
It looks like there is more wetted area forward, but less aft. That is fairly typical for Froude numbers between about 0.3 and 0.6.

 

This is only possible for the tested models, but since the regression analysis is aimed at the resistance-prediction for new models that have not been tested in the tank, these measured trim and sinkage values can not be used as INPUT-data. For new models these values are just not known.

The method that I have chosen avoids the introduction of the moment around the y-axis as an additional parameter. This moment in itself is not a good parameter, because its impact on resistance depends on the hull-form. The hight of the towing point and any additional movable trimming-weight have a significant influence on the resistance, so they have to be included some how. I do not intend to predict the water level at speed, I only calculate the force equilibrium at zero speed. This way I get comparable starting conditions, that incorporate all external moments.

 

wetted surface

Mikko, after re-reading your post it came to my mind, that you might have thought about a correlation of the actual wetted surface at a given speed to the hull parameters. In this case one would employ two predictions, one for the wave resistance and one for the wetted surface. Each of these predictions would use its own set of parameters and its own set of correlation coefficients. The viscous resistance would then be calculated based on the actual wetted surface.
In theory this might sound promising. The problem is, that the hull wave profile and hence the wetted surface is not recorded in the experiments at Delft. Sinkage and trim are not sufficient. Leo has cited Prof. Sharma in post #52 in http://www.boatdesign.net/forums/design-software/flotilla-6-2-released-50116-4.html
This is the best explanation that I came across for this phenomenon.
Thats why I ended up with my crude method
Uli

 

Thanks Uli, an interesting thread you have been hiding with Leo in the Flotilla discussion ;-).

Yes, I was thinking about correleting the actual wetted surface (WS) of the measurement at hand to the friction coeff, to get a more "correct" residuary resistance. But as you can see from the SYSSER 50 example, I'm not sure there is much difference (or in which way) between the static WS and the WS at speed, so maybe it would not change anything.

Prof. S. D. Sharma put it in his comments on another paper by Doctors
and Day:
"The reason why the use of measured trim and sinkage in calculating the
flow by Michell's theory is counterproductive may be that in model
experiments the sinkage is measured relative to the carriage, not
relative to the surrounding water surface. In reality, most of the so-
called sinkage is simply due to the mean local lowering of the free
water surface around the ship. It does not really affect the wave-
making potency of the hull."

In the same run SYSSER 50 sinks about 15 mm, while the measured value from Delft data (measured from the carriage) is -16 mm, or the model is rising 16 mm. So clearly the model is falling in its own hole, or the surrounding water is lowered. My channel is narrow and shallow, I could not find anywhere the real size of the Delft tank used. On the other hand, my system would not allow a much larger tank at this (already rather coarse) resolution.

 

I did a basic RANS simulation without free surf effects on the Sysser 26 model (I assume it is pretty similar to the Sysser 24?). At an ambient turbulence level 1%, and velocity 1,77 m/s (Fn= 0,4), the average Cf is 0,0039, very close to Uli's integral method in "Resistance prediction for the Delft Sysser", Fig 2. The shape of the Cf curve along the centerline of the boat is shown below in Excel. Rather different from that of a body of revolution, but veryb similar in average.

At an ambient turbulence of 0,5% the RANS code predicts much more laminar flow, and the average Cf is then 0,0034 (the second pic).

Correction to above: The ambient turbulence at the inlet is 1%, at the model bow it's about 0,7% and 0,35%.

 

I also looked at the penetration of the turbulence into the water under the model, with the Sysser 50 at Fn= 0,4, the free surface modeled. The illustration is a cut at the midplane - it would appear that under the bow the "bow wave induced turbulence" does penetrate at some distance under the water, at a strenght of 1% or so. At the depths of the keel the turbulence intensity would be small, yet probably more than 0,1%, which could be enough to influence laminar flow. In full size the numbers could be different.

The turbulence behind the boat, in the wake, would deny any hope of laminar flow for someone sailing there.

The simulation is that of a flume, water moving and model in place. Preserving the mass & momentum integrals does create some turbulence into the flow, as the flow speed at the inlet is regulated by the software (such a red "disturbance" just hitting the bow at the surface).

 

Great Mikko! and thank you for sharing. But as always, every answer creates new questions:

I assume you used natural transition, not the forced transition by the sandpaper strips as used in the Delft-experiment. The sand paper at the bow eliminates the laminar flow close to the keel near the bow.
Is the bow in the Excel-diagram at the right?
Is the friction coefficient shown that one for the centerline?
How would the friction coefficient look like, if it is averaged in the y-direction, like I do it?
I could imagine that you would be getting the same curve as I, if you were using b.l.-trips and averaging in y-direction.
The only difference might be the fact, that I get flow separation at the far rear end and you don't. But this does not really make a big impact on the overall resistance.