Michell v CFD and EFD: Round 2007

I've seen some two-phase RANS results that are pretty spectacular. I don't think the CFD state of the art in hydrodynamics is behind the aeronautical state of the art. However, the use of CFD in actual design engineering may lag in the naval architecture field compared to typical aeronautical practice.

CFD is still more expensive than experiment when it comes to generating parametric studies, and it's not yet so turn-the-crank that non-specialists can be assured of getting reliable results. CFD has always had great promise, but the traditional design methods are the tradition for good reason.

I'm reminded of the story of two women, best of friends, who were getting married at about the same time. One was marrying a program manager, and the other a CFD specialist. After their honeymoons, they got together over lunch to compare notes.

"How did your honeymoon go?" asked the CFD wife. "Well, it was a bit strange at first," the program manager's wife replied. "When we got to the hotel room, he pulled out a stack of viewgraphs and outlined how he expected the week to go. But after that, things went more or less according to plan, and everything worked out just fine."

"How did your honeymoon go?" asked the program manager's wife. "Not so well as yours," replied the CFD specialist's wife. "All he did was sit on the edge of the bed and tell me how great it was going to be."

 

A rough try on the bulb idea did yeild a reduction in wave drag ~1%. Unfortunately the improvement was negated by the additional surface area.

 

How would you join the thin aluminium sheets together?

Gareth

 

Ooops! I just realized that I've hijacked Leo's thread about Michlet, and its comparison to experiment and other CFD methods. My apologies'

Leo, could you mention a bit more about the new 20 parameter hull definition?

Also, did you ever get any tow tank data from those guys who built your Godzilla kayak and did it match up well with the Michlet results?

 

Gareth
With flat sheet development there would only be a single joint in the middle. If you did not open the previous pdf file of the plate I have made it easier with the attached picture.

If you wanted to make the boat I can take you through the detail. There are a few simple things that make it very easy.

To join the plates I would use 3mm x 40mm flat aluminium as the connecting strip. Screw the bow and stern sections to this using a flat foam rubber sealing strip and stainless screws (al rivets if you are worried about corrosion). Fast, strong and simple.

Photo shows a slender hull made from a single 1.2 x 3.6m sheet of 1mm thick aluminium. It took me about 14 hours to make and I have better methods since then. No time spent waiting for epoxy or resin to cure. No need for surface finish.

An interesting aside is that the water does not experience any difficulty negotiating the abrupt change in angle in the water plane.

Rick W

 

I used to allow great freedom in offsets, but I cut that capability out because it was a bit messy. I've started to re-introduce more freedom using other methods but I haven't had time to do much with Michlet lately.

It is possible to get very unusual hulls that have very low wave drag. I've attached a Michlet input file for Ward's Optimum Symmetric Ship which was designed by Ward in 1965 to have very low wave drag.

Unfortunately these sort of wobbly hulls arise whenever too much freedom is given to the offsets. Of course, constraints can be placed on the hull slopes in order to prevent boundary layer separation and other weird viscous effects.

Have fun,
Leo.

 

Me too. And the colours were extremely impressive.

What I want to see is some spectacular results for very simple cases like the prediction of turbulent skin-friction on a flat plate at high Reynolds numbers. Unfortunately, no CFD methods give reliable results, often differing from experiments by as much as 10%. Until that gap is narrowed, I have very little faith in predictions for complex geometries. Tweaking meshes in order to get better agreement with experiments is a complete crock.

All the best,
Leo.

 

Leo
That is an interesting shape. I would like to see a longer version.

Do you have a method of generating the hull for zero wave resistance for a given speed and displacement? I tried scaling it but it does not get better than Godzilla produces as far as I can determine.

Rick W.

 

I looked at it too and there is that nice minimum in wave resistance at
speed = 1.223 * SQRT( new_length/1.524 )

That of course is when Fn = 1.223 / SQRT(9.8*1.524) = 0.3165

 

Is there a report on Ward's work that I could download?

How does Michlet account for the boudary layer? Does it calculate the boundary layer thickness and then add that to the hull offset?

 

They don't call CFD "The science of pretty pictures" for nothing!

I agree. The older inviscid method + boundary layer codes seem to be just as accurate for drag prediction as the RANS CFD codes.

From an engineering perspective, at least one can get answers from a CFD code for complex configurations, even if the results are off somewhat. And you know why the total number is what it is. But I'd sure like to see things get a lot more reliable, too.

 

I don't know of any reports online.

Yes, Michlet adds the BL displacement thickness to the offsets. However, in the Michlet input for Ward's hull, I used ntheta = -512. The negative value is a flag to the code to turn off BL effects. It can be shown that the minimum wave resistance (in linear thin-ship theory) occurs when hulls are fore-aft symmetric. BL effects, of course, destroy that symmetry.

Cheers,
Leo.

 

The wave resistance is never exactly zero.

Yes, I could generate hulls with very small wave resistance. Some hulls were shown in a report I wrote called "Godzilla versus the Sea Urchins".

At higher Froude numbers the optimal hulls have all manner of weird bumps and hollows. Some have large bulbs (as in Ward's hull), others have large bulges amidships.

There are some recent published works by L.J. Doctors and Alexander (Sandy) Day that describe similar hulls to Ward's but for catamarans. Ward's hull is only one of many found by several hydrodynamic luminaries e.g. G. Weinblum, John Wehausen, William Webster, and C.C. Hsiung.

The early works are not online AFAIK, but can be found in the proceedings of the first Numerical Ship Hydrodynamic conferences, the Journal of Ship Research, and RINA Transactions. Sorry, but if you are really, really keen, you will have to do the leg work to find them in Uni libraries.

Anther hull type with vanishing wave resistance is called "Krein's Caravan". In Michlet, press "n", then "k" and enter 3 for the number of satellite hulls. The resultant hull will have very low wave drag for Froude numbers below about 0.32. You will need an input file with a lot of stations to capture the oscillatory hull form. An input file for a Wigley is Ok. (The actual hull shape in the file is just a dummy and will be replaced by Krein's). Try ntheta = -512. (The negative value turns off boundary layer effects).

Krein's original caravan is a mathematical curiosity with infinite satellite hulls. For Michlet, I allow a finite number of hulls. I also transformed the original formulation so that the ship length does not increase when adding extra satellite hulls.

Have fun,
Leo.


Leo.

 

I'm pretty pessimistic about CFD's capabilities at very high Reynolds numbers.
The numerical difficulties that CFD models face at high Reynolds numbers are mostly concerned with the very fine grids that are necessary to secure accurate converged solutions. In principle the difficulty should not be insurmountable given sufficiently robust algorithms but, as V. Patel notes, ``solutions may prove prohibitively expensive, if not elusive".

Patel also highlights that a more worrying difficulty for CFD at high Reynolds numbers is the inadequacy of current physical models of turbulence and assumptions about flow very close to solid boundaries.

IMO, CFD does too much for certain problems. Sure, it gives reasonable answers for some complex configurations, but the same methods are far too elaborate for some simpler geometries (e.g. hydraulically smooth thin hulls).

Leo.

 

Hijack away, John. I don't claim ownership of any thread here.

I've put up a version of Michlet that uses Series 20 offsets at:
http://www.cyberiad.net/michlet_betas.htm

To run the example after downloading unzipping etc, just run "godzilla.bat"
This will load up an example using series20. There are some other examples using the same series in the directory. These show the effect of using narrow or wide ranges for the parameters.

There is an explanation of the series in the michoffsets.htm file.
The Godzilla example is described in godzexam.htm

I've no idea how the kayak went against the clock. It was a NASA/Uni of Somewhere, Calif. student project. I think building it as a team was the main objective.

Have fun,
Leo.