Dry tank testing. Can there be such a material?

[Omeron]

For a scale model of 1:10, would it be possible to manufacture a fine powdered material which would mimic water?
I guess it is all about similarity of dimensions and densities.
For a scale model, water is probably too dense to give actual readings.
What would be the density of a hypothetical material, liquid or dry to equal water, and behave like water at this scale ratio?
Is the result such that there is no such material which is as little dense as required.

[Pericles]

http://en.wikipedia.org/wiki/Talc

Pericles

[Omeron]

Interesting reading. Any reference to the subject above?
Density is even greater than water. I would have thought it should go the other way.

[GuestR01312011]

If you wanted to try and scale the fluid you are testing in it's the viscocity you'd want to scale to get the same reynolds numbers, but you can just trip the flow to turbulent to simlate the flow situations.

[Omeron]

Hmm... Thats interesting too.. A quick search of viscosity of substances
reveals that at a scale of 1:10 the only liquid which would resemble water
is liquid nitrogen. Not a very friendly material to handle for this purpose.

[GuestR01312011]

yeah thats why its not done yhou can only satisfy the Froude scaling, with if you trip the flow on the model to turbulents at aknown length you can simulate real Rn numbers.
Alternativley you can test at full scale in a wind tunnell, they do this with keel/bulb configurations and keel/hull interactions.

[Wolczko]

Another thing you have to consider is that solids, even in a very fine powder, will not deform the way a liquid will. Any resistance encountered as an effect of hullform or hydrodynamics would be different than what one would expect to see with water. It seems to me that one would be unnecessarily reinventing the wheel by using some powdered solid to try and imitate a test fluid that could be used anyway. Why wouldn't you just use water? There is certainly a good deal of it within easy reach.

[nico]

For a correct model test (of a model travelling at the water-air interface), you need to keep the froude number and the reynolds constant. This cannot be done if water is used at model and full scale (in this case, only the froude number is kept constant).
Writing both conditions (you can try it); the following result can be found:
(kinematic viscosity)@model_scale = (kinematic viscosity)@full_scale * sqrt(scale).

water is at around 1.14e-6; so with a scale at 1/10; we are looking for a fluid at 0.3e-6. (mercury is at 0.1e-6).

[Pericles]

Sand can be made to act like a liquid by passing high pressure air through it, so that the grains are dancing. If a rubber duck were placed at the bottom of the sand bed and a cannon ball placed on the surface of the sand, the cannon ball would sink and the rubber duck would bob to the surface as the air supply is turned on. http://www.patentstorm.us/patents/5911201.html

Actually, how this helps I haven't a clue, unless talc replaced the sand.

Pericles

[Omeron]

nico, you are on to something. But having gone that far you can perhaps let us know what comes as close as 0.3e-6. Surely mercury is way out.
Are you implying that nothing comes close to this figure.

[water addict]

For a scale model of 1:10, would it be possible to manufacture a fine powdered material which would mimic water?
I guess it is all about similarity of dimensions and densities.
For a scale model, water is probably too dense to give actual readings.
What would be the density of a hypothetical material, liquid or dry to equal water, and behave like water at this scale ratio?
Is the result such that there is no such material which is as little dense as required.

That's a pretty big model scale, water is the most likely medium in which to test. You will get good results at 1:10 scale, and eliminate a HUGE host of complications trying to use another testing medium. I work at David Taylor Model Basin (navy's testing tanks). Most of the models we test are lucky to be in the 1:25 range for scale factor. 1:10 - definitely use water!

[Omeron]

Water may be the easiest and cheapest liquid, and it is fair enough to
look no further. But in reality you need tons and tons of it alongside millions of dollars worth of equipment to make it work. Or you need to pay a high price to rent an existing facility.So at the end of the day the fact that it is free does not mean much.
What i was wondering was whether a small scale test tank could be made,
(no bigger than a domestic acquarium)where the model is kept stationary, and a controlled circulating flow imposed upon it.
I guess water is useless at this scale as you circulate and pump water it would be totally turbulent.
Sand particles mixed with pressurised air sounds interesting but i doubt it would work in practice.
Water Addict i agree with you that 1:10 is a very large scale for a naval vessel of any type. But my interest is for sailboats no bigger than 40 ft.
So at most i am talking about 3-4 ft models.
Can anybody advise whether a small test tank with a stationary model can be made with readily available materials such as a glass container, few feets of tubing, a pump etc. Or is this a futile project because of flow characteristics?
Woludnt it be nice to have this simple setup at your home, and try several things just by plugging it into mains and video results without spending thusands each time...

[GuestR01312011]

Omeron, there's no point in your setup, not to put it down, it could be usefull, if you let the water 'settle' after the pump, to get a general idea of the boats trim etc, but for accurate force and moment measurements therefore resistance estimates it would not be realistic due to the 'swirl' and associated turbulent characteristecs that would be imparted onto the hull and appendages. Americas Cup teams usuall test at 1:9 to get the general setup and then refine at 1:3 scale, yes their models are as big as some nice 'real' boats!
Also to satisfy both Reynolds and Froude scaling, needed to get completely accurate results, you would need something denser than mercury and the speed of the model would have to be very fast, which is not practicle in most tanks as they need roughly 10m to get to const speed and 10m to slow down so in say a 30m tank you only get 10m of useable data and at a large speed (for complete scaling satisfaction at 1:25 scale, model would have to be 25times fater than full scale for const. Fn, say 12knot full scale = 300 knots at model scale!!!) this is not enough at all!
Current tank testing here, at Southampton Uni, is satisfactorily accurate, Americas Cup test acchieve 0.5% accuracy compared to full scale trials! The smaller details are of more importance than a whole new regeime, such as water temperature variations and time between runs to allow currents created due to test to dissipate....
If you want a cheap(ish) estimation of resistance... go for CFD, in the right hands you can get 8-15% accuracy for fully appended yacht...

[water addict]

Water may be the easiest and cheapest liquid, and it is fair enough to
look no further. But in reality you need tons and tons of it alongside millions of dollars worth of equipment to make it work. Or you need to pay a high price to rent an existing facility.So at the end of the day the fact that it is free does not mean much.
What i was wondering was whether a small scale test tank could be made,
(no bigger than a domestic acquarium)where the model is kept stationary, and a controlled circulating flow imposed upon it.
I guess water is useless at this scale as you circulate and pump water it would be totally turbulent.
Sand particles mixed with pressurised air sounds interesting but i doubt it would work in practice.
Water Addict i agree with you that 1:10 is a very large scale for a naval vessel of any type. But my interest is for sailboats no bigger than 40 ft.
So at most i am talking about 3-4 ft models.
Can anybody advise whether a small test tank with a stationary model can be made with readily available materials such as a glass container, few feets of tubing, a pump etc. Or is this a futile project because of flow characteristics?
Woludnt it be nice to have this simple setup at your home, and try several things just by plugging it into mains and video results without spending thusands each time...

You can pump in a circulating channel and create a mostly laminar test section by shaping the water channel. This is common in wind and water tunnels for flow testing, as we have both here at Carderock. If you are looking to create your own test facility, may I be so bold as to ask why?

[Omeron]

Well, as i said before, if you spend a few hundred dollars to have a small setup at your garage, i think it would be fun to experiment with ideas, and also help friends/amateurs.
It is one thing to take your model to a professional facility and pay for it each time, and it is completely another thing to mess around with your own setup without worrying about the cost and hassle of taking it to a facility.
And İ dont think test labs are as abundant as indoor swimming pools in your neighborhood.
Apart from the cost it would take some amount of travelling to and from.
And probably you would feel uneasy going back and forth every weekend
for weeks just to see how minor changes affects your design.
As for shaping the water channel, do you mean using some sort of strainers, or tubing to streamline the flow after exiting the pump?
Would layers of honeycomb sheets do that?

[water addict]

Omeron,
Have a look at this link:

http://www.rollinghillsresearch.com/Water_Tunnels/Model_0710.html

[GuestR01312011]

Yeah thats fine for non critical flow regimes like for prop.s or ships since you are not looking to minimise resistances, look a wetted areas, trim angles and wave formation etc, for an accurate yacht test you need to know exactly the speed of the flow, which is quite difficult to control due to boundry layers etc in the case of moving the water, air is fine obviously though.
If you were trying to create a circulation tank you would need different types of flow straightners in the form of vanes/fins etc. to eliminate 'swirl' caused by the bend in the tunnel and the pump depending on what type youre using, you may also have a problem with free surface effects waves...
A simple enought panel coded (potential theory) CFD package, (quite cheap - PALUSIPAN) would be a good investment and will prob be more accurate than the circulation tank in terms of resistances, but the absence of free-surface modelling is a drawback.

[water addict]

Let's face it.
The reality is you ain't gonna get an accurate home built test tank to predict resistance. It's a bit nutty to think about it. So if what you want it a toy to mess around with, the link I put above is a good example.

[Wolczko]

I agree with Water Addict. Especially with something up to 4' long... You'd have to use something a good bit bigger than the rolling hills research flow channel posted above, and even then, there are significant problems with the use of flow channels for resistance testing. Aside from the problem of turbulent flow in the channel, as well as a flow velocity gradient across the entire channel (which may or may not be easily measured), there will be problems arising with eddies in the flow which will change pretty much every component of resistance in such a way that would be very difficult if not impossible to account for accurately.

There could be some possibility of non-scalable comparative testing in a flow channel, supposing (probably - I'm not sure) that the hulls tested are similar and that there are a great deal of test runs conducted on each hull. But, like Water Addict was saying, I think that trying to get any resistance data which could be applied to a full scale boat from flow channel testing is probably futile. Besides, if you've got a swimming pool handy, you can always just do some old school testing in that. ;) I seem to remember Ben Franklin rigging up some sort of weight-driven towing apparatus in his attic which you could at least have a good time designing and building. And again, comparative testing could probably achieve some level of relevance in that setting, so long as you remember to turn off your circulation pumps and buy a nice stopwatch.

:) Wolczko

[FarmerColin]

Given all the difficulties of scaling - gravitational wave considerations need the flow speed to be reduced by the square root of the scaling factor while Re number considerations need it to be increased by the scaling factor!! - I can't help feeling you would get more accurate indicators of the relative performance of two hull-forms by numerical methods

[fastwave]

If you want a cheap(ish) estimation of resistance... go for CFD, in the right hands you can get 8-15% accuracy for fully appended yacht...

In the right hands you will get accuracies in the order of 1%, but it is not cheap.

[nico]

nico, you are on to something. But having gone that far you can perhaps let us know what comes as close as 0.3e-6. Surely mercury is way out.
Are you implying that nothing comes close to this figure.Yes, i dont know if a fluid with these characteristics can be found.

go for CFDCFD is for sure the way to go, it's getting more accurate every day (computers are getting faster and bigger). You get far more insight into the physics. It's cheaper. And it's already more accurate that towing tank test; america's cup team are already moving to only cfd.

[fastwave]

Just to put some numbers into context, because there are a lot of people saying CFD is cheap.

To do free surface CFD with a RANS code and expect good results these are the following costs.

16 CPU workstation 5,000euro
commercial CFD licenses 3000euro+/month
CFD specialist 5000+euro/month

so total 13,000 euro for probably 30 points.


so 400euro per point.


Panel codes can be 50euro per point or cheaper if you do it yourself.

[Guest625101138]

I expect you are considering drag of efficient displacement hulls. In this case Michlet will give results within 5% and it is free.

Rick W.

[fastwave]

I belive that since this is posted under the sailboat section that most real boats do not fall under the L/B of michlet and also sail are not symmetric while sailing.

Also, what about lift??

[GuestR01312011]

No Cup teams used purely CFD for anything, and none did a complete analysis of fully appended boat, appendage designers use it but none do or ever will use purely CFD, you will not get 1% accuracy ever, even with an infanitley detailed mesh which will cost infinite computing! Towing tanks will always be around, Its very cheap compared to tank testing considering a 1:3 scale model of cup boat will cost roughly 10,000 euro and a testing programe up to 30,000 depending on test matrix and tank used...

[fastwave]

I am afraid that is not true. Many team analized fully appended boats. In addition the accuracy that can be achieved is very good. Within the experimental error or the tank, which inherently has scaling errors.

Computational resources are not an issue. Caclulations can takes a couple of days on 40+ CPUs but that is not an issue.

CFD can be cheap depending on how far you want to push it otherwise it gets expensive.

[Guest625101138]

What L/B is being considered? Michlet will give useful results at least down to 5.

As far as lift goes I would expect leeway to be slight if it is an efficient design with deep keel. As a first approximation you can add the induced drag of the hull for the determined angle of leeway.

I have not tested the pitch or roll settings in Michlet but it would be worth trying as it gives very good results in level trim. I have adjusted trim on hulls but imported the trimmed hull into Michlet. I did not use the internal settings in Michlet. You could ask Leo about the limitations. He has a number of comparisons against CFD and the real world results usually compare better with Michlet than the CFD.

Rick W.

[nico]

No Cup teams used purely CFD for anything, and none did a complete analysis of fully appended boat,
Ok, thanks for that; you seem to be well informed. :)


so total 13,000 euro for probably 30 points. make it 150 points; 80euros/point. And compare to tank test.


There are also a few comments about Michlet, it's a great tool and probably a good introduction to CFD; but it has a very very limited application to yacht design.

[GuestR01312011]

I was at the RNLI lecture at Uni Southampton given by Andy Claughton, former design co-ordinator of Emirates team NZ and now for Team Origin who said that no teams did fully apended CFD analysis, they did for just keel and bulb and also for keel hull joint.
I dont know what Michlet is, but CFX or Fluent are your best bet for accurate(comparitave)results.
Quoting from Wolfson Unit, yacht experimental techniques lectures for MSc given by Ian Campbell former cheif experimentalist for Luna Rossa:
CFD
Potential flow or RANS or LES:

•“Numerical experiment”
•Qualitatively correct
•Numbers NOT generally correct
•Not enough cells/discretisation
•Good for understanding what is
happening in flow

Ian carried out a tank testing regieme for Luna Rossa of 57 scale models over their campain for the 32nd cup and achieved a 0.5% accuracy compared to 10-15% accuracy of CFD results...

[Guest625101138]

........
I dont know what Michlet is........

Michlet uses Michell's thin ship theory for determining wave drag, has a choice of ITTC1957 or Grigson for viscous drag and also has a component for calculating transom drag.

I have found it gives reliable results for efficient hulls. The drag on such hulls tends to be dominated by viscous drag anyhow. Modern sailing craft are using more slender hulls such as trimarans, catamarans and canting keel yachts. These boats are designed to stay upright as well so heel is not a key concern.

If the hull is designed to plane then Savitsky gives reasonable accuracy. I was a little surprised that Michlet and Savitsky are not too different at the bottom end of the planing range.

There is a post here that compares Michlet with CFD and actual data:
http://boatdesign.net/forums/showthread.php?t=20497
Curves attached.

[nico]

A nice article by Jason Ker. http://valenciasailing.blogspot.com/2006/10/inside-americas-cup-team-boat-designer.html

[fastwave]

Nico,
to make it 150 point you need a lot more computing power than 16 processors for the runs I am talking about. Although the comptures might be cheap the licenses are no that cheap. so 80euro a point is very very difficult to get.

on 16 processors you will be lucky to get a point/24hours. You will only achive this if you have a very good and developed sustem in place.


With regards to people still saying that full appended runs are not possible, I only have to say the following:

1. It is done.
2. Quoting Ian Campell and Andy Claughton, is not representative. Both work, worked for the wolfson unit which is a big time fan of tank testing and not a great supporter of CFD.

people seem to know a lot about AC teams on this forum, but I wonder how many actually have a good insight of what is going on.