CFD state of the art

[Remmlinger]

Uli:
Does this mean that the Miles Cornwell statement about test conditions being FW at 20C is incorrect?
Don MacPherson

I trust the file "Tank Testing Results & Report" from the Wolfson Unit at
https://drive.google.com/file/d/0B2CUun8QUAMsb1FRcHdRajdLSjQ/view?pref=2&pli=1
In the file "raw data" there is the statement "Water Temp = 18.5 deg C".
In the file "all results" the resistance values are larger, because the values are corrected to match the standard condition of 15 deg C.
In the FINE/Marine Results Submission in the report from NUMECA you can find on page 8 the condition for the simulation. It is fresh water at 15 deg C. They have obviously used the corrected drag values for the comparison.
To me this all makes sense. A value of 20 deg C is not supported by the data.
Uli

 

[Windvang]

Uli,

Can you please provide your diagram in a table/Excel format? I am also discussing the results with people from Delft.

Arthur

 

[DMacPherson]

Uli:

I wonder if the correction to 15C also considered the small change in density and draft (sinkage), or was the correction solely viscous? If one assumes that the displacement was constant (regardless of water properties) at 197 kg, it seems perhaps a better comparison will be at the tank condition (to remove any doubt about draft).

Don

 

[Remmlinger]

Uli:
I wonder if the correction to 15C also considered the small change in density and draft (sinkage), or was the correction solely viscous? If one assumes that the displacement was constant (regardless of water properties) at 197 kg, it seems perhaps a better comparison will be at the tank condition (to remove any doubt about draft).Don

Don, this is exactly what I did in my comparison with the Delft-prediction.
I used the properties of fresh-water at 18.5 deg C, displacement = 197kg and compared it to the raw data, corrected for blockage and stud-drag.

What amazes me is the fact, that the results from CFD with 64 cores and several days computing time are not better than the simple Delft-formula.
Uli

 

[DMacPherson]

Well, when the hull form very closely matches the parameters of the series it should not be surprising that the predictions are close. It is when we have a hull that doesn't match a series that a higher order method is justified.

In my (limited) experience, wide and shallow hulls often are not well modeled with flow codes (of various nature). It seems like when flow is less "waterline" and more "buttock" that the codes break down. So, this hull is actually a good test for CFD. It has quantitative benchmarks (i.e., your Delft series comparison) and it is a challenging hull.

My takeaway is that I think the claim that the results of the upright canoe body test (CB-1) are "good" is awfully optimistic. I would not be happy with a 10% under-prediction of total drag (which is a greater error of residuary or wave-making). Also, one study invalidated itself by over-simplification of the model, and this speaks to a practical concern that is too often overlooked.

Don

 

[Windvang]

Don, this is exactly what I did in my comparison with the Delft-prediction.
I used the properties of fresh-water at 18.5 deg C, displacement = 197kg and compared it to the raw data, corrected for blockage and stud-drag.

What amazes me is the fact, that the results from CFD with 64 cores and several days computing time are not better than the simple Delft-formula.
Uli

The Delft series is based on a lot of tests, that compares to an awful lot of computing time and data. As long as the yacht is within the parameters of the tested range the result is not so much of a surprise. As new as this design is, the upright parameters are not that different of older designs, so the upright resistance will be close. Note that a lot of Delft series only go up to Fn 0.60, so above that results may be unreliable. Currently there is a TP52 style hull under test, so future results may be even closer for this design.

 

[Mikko Brummer]

Myles Cornwell confirms that the temp was indeed 15 deg:

"Hi Mikko,

I noticed the debate on the forum regarding the water temp used, so I dug back through the project emails. The water temp for the analysis was agreed at 15C not 20C as some in the forums have pointed out. Sorry for the confusion, and thanks for relaying this to those interested."

Well spotted, Uli.

 

[Remmlinger]

Well, when the hull form very closely matches the parameters of the series it should not be surprising that the predictions are close.

Don and Arthur, you both seem to be right.
I am currently comparing the USSAIL tank-results:
http://www.sailyachtresearch.org/19-library/57-us-sailing-nine-model-series
to the predictions of the Delft-method. For the narrow hull#7 the Delft-method is more than 20% off. It might just be good luck, that the wide-light-hull is in the upright condition close to the geometry of the DSYHS.

 

[Mikko Brummer]

With ref. to an earlier "definition of planing" thread, from the SYRF results graph included below, the model would appear to

- sail in displacement mode until about Fn 0,40 when VCG is at its deepest

- start planing at about Fn 0,53, when VCG has risen back to its static level and higher

- be fully planing at about Fn 0,70, when the trim begins to decrease

 

[Remmlinger]

Arthur reminded me of Lionel Huetz' prediction method, that was especially developed for "wide-light" boats:
http://www.sailyachtresearch.org/images/library/20150723202925.pdf
It does not compare well with the tank data, see attached diagram.
While I was working on it, I added also some CFD-results to the diagram.
All CFD results are assumed to be valid for 15 deg C and are converted to the tank condition of fresh-water at 18.5 deg C.
The raw data from the tank were only corrected for blockage effect and pin drag.
I think I learned the following:
1.) A boundary-layer calculation, that assumes the still water level and neglects waves, seriously underpredicts the viscous drag for such a flat soap-dish.
2.) If properly corrected, the CFD-results are reasonably accurate.
3.) The good prediction of the Delft-method might just be a lucky coincidence.
4.) Lionel Huetz' method is not applicable for this boat. The friction drag (that does not even include a form factor or the negative pressure at the transom) is heavily overpredicted.
Uli

 

[HJS]

With ref. to an earlier "definition of planing" thread, from the SYRF results graph included below, the model would appear to

- sail in displacement mode until about Fn 0,40 when VCG is at its deepest

- start planing at about Fn 0,53, when VCG has risen back to its static level and higher

- be fully planing at about Fn 0,70, when the trim begins to decrease

I call Fn 0.7 semi-planing where the bow wave and stern wave interact in a positive way.
Full planing lies around the Fn 1.0. and that is something completely different.
Or?

js

 

[Mikko Brummer]

I call Fn 0.7 semi-planing where the bow wave and stern wave interact in a positive way.
Full planing lies around the Fn 1.0. and that is something completely different.
Or?

js

Maybe so... I was merely quoting a method by DMacPherson in the planing thread.

If you look at the attached fotos from the tank in sequence (from Fn 0,60 to Fn 0,75), the bow first goes up and then starts to go down from 0,70 to 0,75. Whether that is a sign of planing or not, a matter of definition I guess, a definition that was never agreed on in that thread.

 

[CT249]

That's a very interesting point, Mikko; thanks. Andy Dovell uses the same definition of planing and it's interesting to see it satisfied at such low speeds.

On the other hand, on a very small "sinker" windsurfer (or a waterski) the C of G can start rising from the static position due to hydrodynamic lift at a time when the whole board is still 2' underneath the water!

 

[Mikko Brummer]

I think I learned the following:
1.) A boundary-layer calculation, that assumes the still water level and neglects waves, seriously underpredicts the viscous drag for such a flat soap-dish.
2.) If properly corrected, the CFD-results are reasonably accurate.
3.) The good prediction of the Delft-method might just be a lucky coincidence.
4.) Lionel Huetz' method is not applicable for this boat. The friction drag (that does not even include a form factor or the negative pressure at the transom) is heavily overpredicted.
Uli

Surprising indeed that the Huez prediction is so much off, after all the hull is quite a bit like an Open 60. The friction drag he uses seems to be so much off. Are you sure you have gotten it right?

Strange also that Star-CCM predicts a negative pressure drag at the slowest speed... Your bl. calc possibly underestimates the wet surface area, increasing quickly with speed with the flat, wide afterbody?

 

[DMacPherson]

Regarding indication of planing, in my "inference" that a hull is planing (as "definition" is too rigorous a term) the vessel first must pass the test that it's dynamic VCG is higher than static. (See post #2 in the thread.) That is not the case here. Heave is negative throughout the speed range, indicating downward dynamic lift. I would not consider this planing.

Also, remember that this is a really simple model...

Don