Michlet

[pavel915]

I have some simple questions regarding Michlet,

For what range of L/B ratio, B/T ratio or for what speed range Michlet is applicable? I know the process of using Michlet but i want to be clear about for what kind of vessels i can use it with confidence?

[Leo Lazauskas]

Originally Posted by pavel915

I have some simple questions regarding Michlet,

For what range of L/B ratio, B/T ratio or for what speed range Michlet is applicable? I know the process of using Michlet but i want be clear about for what kind of vessels i can use it with confidence?

Michell's thin-ship theory assumes that the longitudinal slope of the hull is small.

I would not use Michlet for L/B < 6, or B/T > 6.

Best results will be for 0.3 < Fr < 1.0
For high Froude numbers remember to specify a large value of Ntheta in the input file.

For multihulls use a larger Ntheta because the wave integrals are more oscillatory.

If you want to see whether Ntheta is adequate, check the free wave spectrum plot to see if you are capturing the diverging wave component accurately, especially for super-critical Froude numbers in finite depth.


Good luck!
Leo.

[pavel915]

thanks Leo,

[Leo Lazauskas]

Originally Posted by pavel915

thanks Leo,

Pavel et al,
I have attached a first draft of a short note comparing Michlet predictions with experiments with a DTMB 5415 destroyer hull. This hull is a fairly severe test for linear codes - the hull is not thin in the waterplane (L/B is about 7.7), it has a small transom, and a sonar bulb. I think it is at the limit of what Michlet can handle.

I'll leave you to judge whether the predictions are reasonable. You should get better estimates with thinner hulls.

You might also like to ask yourself...

Which experiments do I believe?
Do I believe in form factors?
Do I believe in the so-called "transom stern hydrostatic resistance"?
Santa (Y/N)

All the best,
Leo.

[pavel915]

Thanks leo for attaching that paper,

I have a plan to varify that ,whether the Cargo vessels operating in our country are compatible with michlet or not where L/B is not that much higher arround 5.5. Because i have heard some naval architects that they found michlet good for L/B>4(bellow 6),,,

[Leo Lazauskas]

Originally Posted by pavel915

Thanks leo for attaching that paper,

I have a plan to varify that ,whether the Cargo vessels operating in our country are compatible with michlet or not where L/B is not that much higher arround 5.5. Because i have heard some naval architects that they found michlet good for L/B>4(bellow 6),,,

That is a very useful exercise, Pavel.

If you are looking for some "standard" test cases where Michlet should do poorly, try:
KRISO Tanker (KVLCC2)
KRISO container ship (KCS)
Series 60
SL7

The first two vessels were used as test cases at the Gothenburg 2000 Workshop on Numerical Ship Hydrodynamics.

I might have the offsets for the KVLCC2 somewhere. If I find them I will post them here. I also have the predictions of 13 CFD programs for this vessel so you will be able to see how well (or badly) Michlet predictions compare.

Have fun!
Leo.

[daiquiri]

Hello,
I have stumbled across a web page with FormSys' Hullspeed manual, Appendix C in particular:
http://www.fdsfiles.com/webmanuals/h...tm#Wigley_hull
As you can see, they have included some validation graphs for Wigley hull in various configurations (mono, cat, tri) and for a NPL-series hull, with experimental data vs. Hullspeed and Michlet prediction.

It appears to be a complete defeat for both Hullspeed and Michlet!
In case of Wigley hull, up to 50% error in 0.4+ froude region is observable (and 100% and more for Fr<0.3) when confronted to experimental curve. But it is the NPL's wet transom that apparently puts Michlet in crisis, below Fr=0.3 and above Fr=0.45 (last 2 graphs).
Can you please comment on this validation result, and if you think they have commited some error while using Michlet, could you please point out where the error might be?

Thank you in advance.

Slavi

[Guest625101138]

You will find they have an error in the hull offsets used in the Michlet in.mlt file. I have seen the problem before when Freeship or Delftship is used to generate the in.mlt file.

They will have attached flow up a steep transom rather than flow detaching at the cut-off transom. It means the transverse wave is much larger than it should be until the hull gets above hull speed.

Mat Marsh alerted me to this problem with Delftship export. It is easy enough to correct in the in.mlt file. It was not something I had tested because all my hulls use canoe sterns.

Leo may not be aware of the problem because he may not have produced an in.mlt file from Delftship.

I expect you will find that other errors between the experimental data and Michlet are largely due to experimental error or incorrect input files. I test at large scale and get very good validation. There are slight nuances with sinkage and trim changes that I can detect. Michlet does not adjust for these but Flotilla does.

Rick W

[daiquiri]

Rick,
I can accept that there is an error in transom modeling in case of NPL-series hull. That was a very first thing I have taught about. But, say that you manage to simulate the clean flow detachment at the transom. That's how Hullspeed appears to modelate the wet transom. Then you would obtain nearly the same results between Michlet and Hullspeed, because they are based on the same Mitchell's theory.
But you would still be left with a difference of about 30-50% between experimental data and the software output, which is really very hard to justify in terms of experimental error. Experimental data with such a huge error would never be published by any serious author, let alone be used for software validation purposes.
Then there might be an arror in input files due to some Deftship/Freeship procedure error, as you said. But I truly doubt that Formsys would use Delftship or Freeship to create .mlt files... They are a software house which produces Maxsurf, which supports all stages of design from draft to analysis to production, so what makes you think they would ever have to use Freeship to create offsets?

[Leo Lazauskas]

Originally Posted by daiquiri

Hello,
I have stumbled across a web page with FormSys' Hullspeed manual, Appendix C in particular:
http://www.fdsfiles.com/webmanuals/h...tm#Wigley_hull
As you can see, they have included some validation graphs for Wigley hull in various configurations (mono, cat, tri) and for a NPL-series hull, with experimental data vs. Hullspeed and Michlet prediction.

It appears to be a complete defeat for both Hullspeed and Michlet!
In case of Wigley hull, up to 50% error in 0.4+ froude region is observable (and 100% and more for Fr<0.3) when confronted to experimental curve. But it is the NPL's wet transom that apparently puts Michlet in crisis, below Fr=0.3 and above Fr=0.45 (last 2 graphs).
Can you please comment on this validation result, and if you think they have commited some error while using Michlet, could you please point out where the error might be?

Thank you in advance.

Slavi

Thanks for the question, Slavi.

1. I disagree with Rick - I don't think that the problem with the NPL series is necessarily related to bad offsets generated by other programs such as Delftship. I think it is much more fundamental than that.

1a. First and foremost, the wave resistance results for Michlet for the NPL catamaran are misrepresented in the Hullform manual. The Michlet curve on that page is actually the wave resistance plus the transom hydrostatic resistance. To be fairer, they should have shown those components separately, so maybe they didn't know how to run Michlet properly and/or how to interpret the results. If they had also shown the total resistance curve there would possibly be less of a difference between Hullform and Michlet at low Froude numbers.

1b. Secondly, the NPL hulls have a large transom stern, and that violates Michell's small longitudinal slope assumption.

A lot also depends on how the flow behind the transom is represented. Hullform uses a virtual appendage behind the transom because that gives better correlation with experiments with the NPL series. However, I don't believe that is a consistent improvement for all transom stern hulls: in many papers I have seen, sometimes the correlation is better, sometimes it's worse.

And I'll repeat my mantra until someone knocks it down...
A hollow in the free-surface that is ventilated to atmospheric pressure cannot sustain a pressure, and hence cannot generate waves.

Therefore a virtual appendage is not a physics-based formulation. It is just a convenient "hack" and one that I have resorted to, so I'm not trying to appear as pure as the driven snow here.

1c. Thirdly, the experimental results are a bit suspect.
The NPL hulls are quite small, with L=1.6m, and that introduces a variety of problems.

Others here may disagree, but results for such small models are usually far more scattered than for larger models. An example of this is the scatter in experiments coming out of the on-going "Facility Bias World-wide Campaign" currently being conducted by the ITTC.

If you look at my short note in an earlier post, you will see that there is greater scatter in the results for the smaller (3.048m) hull than for the 5.72m model. Of course, that might not hold up when more results come in from the other 30 or so towing tanks. Unfortunately someone lost some of the latest experiments (Oy Vey!) so it might take time to conclude what I regard as the best chance so far of compiling experimental data with which to validate CFD and other mathematical models.

1d. Fourthly, according to the experimenters, the form factor for the NPL4a catamaran (with w/L=0.2) is k=1.43. That is very high, and an excellent example of how poorly some of us (ship hydro boffins) actually understand free-surface flows around surface-piercing vessels.

It's also one reason I included a checklist of "beliefs" to consider in my previous post:
* Which experiments do I believe?
* Do I believe in form factors?
* Do I believe in the so-called "transom stern hydrostatic resistance"?
and I could add,
* Do I believe in virtual appendage models?

2. Form factors are still controversial and many towing tanks and ship operators do not use them. Or should I say "believe" in them?

For example, using the ITTC line for small models with high L/B often requires a form factor that is less than one, a clearly nonsensical idea. Using a physics-based friction line (such as Grigson's line) can sometimes overcome that difficulty, but it is still not a satisfactory state of affairs.

Even for the Wigley monohull, which should be understood quite well, there are some strange things going on. It is not clear that using a form factor consistently improves agreement of experiments with predictions. There is a bit on this in Chapter 7 of the thesis at:
http://www.cyberiad.net/library/pdf/llphd.pdf
There too it can be seen that results for larger models (L=3.048m) are more consistent than for smaller models (L=1.8m).

Use of a form factor (about k=1.12l) in the results for the DTMB5415 in my previous short note will cause an over-prediction of the total resistance. That's not just a problem for Michlet and Flotilla - the results shown for several CFD codes in my note indicates that they too would over-predict CT if form drag was added.

It should also be noted that the CFD codes shown in my note did not try to predict squat. They just put the the hull in the attitude determined from experiments and then estimated the resistance for the hull in that attitude. If they also tried to predict squat, some would take months to run, and some might not be able to handle it all. It is hard enough to mesh a static hull in some CFD codes; imagine what it would be like having to do that at each iteration until a hull is in equilibrium with dynamic forces and moments.

I'm not sure how the good chaps at Formsys ran Michlet for the NPLa catamaran, or even which version they used. Maybe they used the experimental values of squat for each speed, maybe not. They don't mention it on their page.

All the best,
Leo.

[Leo Lazauskas]

Originally Posted by daiquiri

Rick,
I can accept that there is an error in transom modeling in case of NPL-series hull. That was a very first thing I have taught about. But, say that you manage to simulate the clean flow detachment at the transom. That's how Hullspeed appears to modelate the wet transom. Then you would obtain nearly the same results between Michlet and Hullspeed, because they are based on the same Mitchell's theory.

No, there are still differences between the codes.
For example, have a look at Figure 8 of the paper at Formsys:

http://www225.pair.com/magic/ftp/Max...ody_Couser.pdf

There you can see that they choose the length of the virtual appendage as a function of Froude number in order to give the best agreement with the NPL series experiments. Maybe the same variation of length with Fn works for other hulls, maybe not. As I said - it's just a hack with no real physics backing.

A similar method has been used to model the behaviour of the boundary layer just behind a (pointed) stern. Ship hydro titans like Havelock used a cusped virtual appendage in their modelling and and sometimes they got better agreement with wave resistance experiments.

However, that's not how viscous layers behave! They spread out and diffuse, they don't form neat little cusped appendages.

Originally Posted by daiquiri

Rick,
Experimental data with such a huge error would never be published by any serious author, let alone be used for software validation purposes.

Don't start judging hydro experiments by your fancyman aero standards!

Leo.

[daiquiri]

Ok, I took note of that. I will re-read my papers about Mitchell's method, who knows if (with a help of a couple of glasses of red wine) some useful idea comes to my mind. Don't take it as an expression of uppityness or arrogance, it's just that I really like red wine.

Originally Posted by Leo Lazauskas

Don't start judging hydro experiments by your fancyman aero standards!

Yep... Where are those good days when 10% error in experimental data was 5% too much?
As I said few months ago, when I first approached the boat design, I was appalled by the physical and mathematical complexity of ship aero/hydrodynamics. From a mathematical point of view, I understand that it is in a way like passing through a door which separates a 2D from a 3D world. So I know what you mean. But still, I can hardly believe that such a (presumed) flawed measurements (30-50% presumed error!) would be ever officially published.
By the way, The second thing that left me numb was the amount of approximation commonly adopted in the design process, but somehow I find it also a pretty charming part of the job. The problems arise when you have to explain that state of the art to your client...

[Leo Lazauskas]

Originally Posted by daiquiri

...I know what you mean. But still, I can hardly believe that such a (presumed) flawed measurements (30-50% presumed error!) would be ever officially published.

Have you ever looked closely at the drag coefficients of some airfoils at low Reynolds numbers? I've seen 50%-100% differences for the same airfoil shape.

Originally Posted by daiquiri

By the way, The second thing that left me numb was the amount of approximation commonly adopted in the deign process, but somehow I find it also a pretty charming part of the job. The problems arise when you have to explain that state of the art to your client...

Maybe we need to add another component to the (joke) stereotypical naval architect's personality...
Not tough enough to be an engineer, not gay enough to be an interior designer, but better than a used car salesman at selling their opinions

Good luck,
Leo.

[Guest625101138]

Originally Posted by daiquiri

Rick,
I can accept that there is .... But I truly doubt that Formsys would use Delftship or Freeship to create .mlt files... They are a software house which produces Maxsurf, which supports all stages of design from draft to analysis to production, so what makes you think they would ever have to use Freeship to create offsets?

You might ask them how they created the in.mlt file. It looks very similar to the error I have seen using Delftship.

I have attached some screen dumps of Michlet comparing the correct lines and the faulty lines. There is also a direct comparison of the transverse wave only so you can see how large the error is if the transom is not correct.

Dicking around with little models is prone to error. Michlet will give you better results than playing with models and it is much faster.

Rick W

[daiquiri]

Originally Posted by Rick Willoughby

You might ask them how they created the in.mlt file.

Already done...

From the qualitative point of view, that hump at the low Fr's (your last graph) is very similar to the one present at the Formsys' validation page. There is only one way to verify it, and I'll do it... It would be an interesting surprise to discover that, with all the brain-power in their house they've had to use a third-party software to create hull offsets.
In the meanwhile remains the problem of the big discrepancy with experimental curve for both the Wigley hull and for NPL at high Fr's, which reveals a generalized problem with the alghorithm concerning wet transom. But Leo has been very generous and detailed at explaining the reason and his point of view about it so, while I'm meditating over it, a big thanks to him (forgot to do it before).