Porposing Four Winns

[Joakim]

Rick: There is no problem at beta=0 for the Savitsky model. Only the 2006 introduced whisker spray model has a singularity, since the whisker spray angle formula used can not handle it, thus an if(beta<0.1)... is needed.

I think the Savitsky method with 2006 additions is very good. I have also added a model for hull roughness. The accuracy is much more limited by the input values than the model. E.g. LCG, aerodynamic drag coefficient, hull roughness are seldom well known. For a racing boat all these have a huge effect.

Being a CFD professional I don't think there will be much more to be gained from CFD for truly planning vessel, with a bottom fitting to Savitsky model. Semiplanning might be a different issue.

[Rick Willoughby]

Originally Posted by Joakim

...
Being a CFD professional I don't think there will be much more to be gained from CFD for truly planning vessel, with a bottom fitting to Savitsky model. Semiplanning might be a different issue.

Joakim
This paper shows some of the work I was referring to:
http://nparc.cisti-icist.nrc-cnrc.gc...5319&article=0
They have published other papers as well.

The results are very interesting if you are into planing.

I think some of the work Leo Lazauskas is doing with lift and sinkage may be relevant for the semi-planing mode.

Rick W

[Joakim]

Originally Posted by Rick Willoughby

Joakim
This paper shows some of the work I was referring to:
http://nparc.cisti-icist.nrc-cnrc.gc...5319&article=0
They have published other papers as well.

You don't happen to have a link for the CFD-results? Or to the porpoising paper?

I did some tests with my Savitsky program just ignoring the flat part and using beta=5.9 and beam to chine. I could reproduce the results from 3 to 7 m/s within +-7% in resistance and +-0.5 degrees in trim. Mostly much more accurately and all the changes due to changes in LCG or displacement were accurately reproduced. The speed for the porpoising onset was spot on.

Also the 2 and 2.5 m/s results were within +-10% and +-0.7 degrees, althogh Savitsky model is no longer valid there (bow touches water).

[Ad Hoc]

You can try here for the graph of 'limits' for porpoising (according to Savitsky)
Porpoising
post #4

[baeckmo]

You must not forget the origin of the Savitsky method. It is NOT a stringent mathematical model of a physical phenomenon. It is a set of empirical curve-fitting algorithms describing mean values of a number of test observations from a variety of test facilities using a multitude of hull shapes, sizes and proportions put into a context through application of hydromechanical principles! The bulk of these tests date back to the 50-ies.

Treating the outcome from the playstation as a de facto truth, without realizing the scatter and application limits of the ingoing data, is just nonsense. Most of the tests cover deadrise angles 10 to 20 degrees, and many hulls are not pure constant-deadrise. Going outside this range is in fact unceartain territory. And further, as Joakim correctly has observed, real world surface relative roughness is higher in the common boat sizes discussed here, than in the lab prepared models.

I have previously remarked on the lack of compensation for aspect ratio in the drag calculations as well as in the calc of COP and the resulting trim. It influences the basic drag equation (tan(trim)+surface friction), where there is a 3-d flow generating induced drag connected to the trim term. This has been neglected by most Savitsky users. It becomes obvious once you start checking the various programmes for validation. The general trend is that the Savitsky algorithms underestimate drag and overestimate trim. Introducing a simple aspect compensation à la Joukowsi is a first step that works out fairly well.

This becomes obvious in one of the basic equations that describes the difference (wetted keel length)-(chine length). Savitsky is taking this difference as (b/pi*tan(deadrise)/tan(trim)). Obviously the expression goes to zero with deadrise zero. BUT that requires an infinitely wide plate; a flat boat bottom generates a strong three-dimensional flow in the spray region of a planing, zero-deadrise flat surface. Ie there will be a difference between Lk and Lc in reality.

One manifestation of that phenomenon is a change in COP and trim as if there were a small, but finite deadrise. This is the explanation for Joakims result, setting beta=5.9 degrees, resulting in a "fake" COP and drag that is closer to empirical reality.

So, the Savitsky method is a practical tool, nice in terms of calculating time, but still just a tool with a tolerance band and application limits; to be used with sound engineering judgement!!! And it can still be improved upon!!!

[Frosty]

Reg, I guess all this is above your head as it is almost every one. This happens sometimes.

To get your boat to a drivable situation, moving stuff you can move forward will help as indeed you have said it does. You may try additional weight such as sand bags on the bow. If this cures the prob then you know how much weight you need to move.

Some trim tabs would definately help. They are small flaps, say 1 foot by 1foot fitted to the transom on the planing water line. This tricks the boat into thinking it is longer. Some are fixed and some are hydraulically adjustable.

This alone may cure the probs.

I had a Black shadow at Windermere, it too was silly. It had a fuel tank up forward and if I did'nt put 40 galls in it it was a pig. It should have been in a circus.

[Joakim]

Yes, I do know that the Savitsky model is based on experimental data as are almost any other model. There is no pure physics based model for any turbulent 3D flow. And the base measurements are old, just like the base measurements for pipe flows and wing shapes.

The basic model for lift and drag is developed for flat bottom, thus I don't think beta of 0-10 degrees could be unknown territory for the Savitsky model.

Could you be more specific about the Joukowski compensation? If it is just induced drag, why it would affect COP and trim as well?

The surface roughness is not a problem for lab model vs. real boat. They do not need to have the same roughness. The problem is that you need to have a model for friction that takes into account the correct roughness for the real boat. The ITTC model + standard roughness allowance is typically not enough for boat size with clearly lower Reynolds number than ships have. Of course the friction model originally used for towing tank model must be correct as well, but I don't think that is a problem.

The results from my Savitsky program are very close to each other with both 5.9 and 5.1 beta. 5.9 is the real angle of the hull surface and 5.1 is corrected for the flat part in the keel. For this example the predicted trim angles were mostly lower than the measured ones. The predicted total resistance was lower (3-6%) than the measured at 3-4.5 m/s, but at 5-6.5 m/s it was very accurate. Under 3 m/s some parts of the model are not valid, thus I don't compare those.

The trim angle is spot on at 3 m/s, 3.5-5.5 m/s the predicted trim is too low (0.2-0.4 degrees) and then 6-7 m/s spot on.

The comparison above is for LCG 0.53 m, 29.6 kg from http://nparc.cisti-icist.nrc-cnrc.gc...5319&article=0

The surface roughness and the accurate location of the towing point were not reported. These can have some effect. Also the aerodynamic drag can have a minor effect. I used a surface roughness of 10 um and set f=Epsilon=0. The aerodynamic drag in my program was 0.3 N at 3 m/s and 1.5 N at 7 m/s. There was no whisker spray drag (only accounted for, if positive).

[baeckmo]

Joakim, we should not hijack this thread with a deeper diskussion on Savitsky's algorithms. You are welcome to mail me directly, and you may also find some of the litterature in the thread "State of art of planing hulls" interesting. The intention of my comments above was to remind some of the readers here not to stretch their conclusions too far into the region of fanatic playstation-ism. Arguing on fractional degrees of trim is just on the edge, and obscures one's perspective on real world engineering, as Frosty aptly noted.

Just a short note on your quests: For a flat planing surface of finite aspect ratio, the lift induced drag coefficient is 2/pi*Cl^2/A in its basic form, where A=(span)^2/area. It is to be added to the wave-making and friction coefficients.

The influence upon trim is seen if you imagine the spray-root (~stagnation line) of your zero-deadrise surface. With reducing A from infinity, the spray-root changes from a straight line into an arch-shape. The center of pressure and the pressure integrated over the surface will change correspondingly. Trace this influence through the chain of algorithms in your program, and you will see the difference. In his 1964 paper, Savitsky is referring to the 2D case when dealing with drag prediction (Fig 13 ibid). I cannot see why he didn't include finite-span correction to the drag issue, when he spent the rest of the paper on the effects of aspect ratio on lift???? Just bloody tired of the s--t, maybe?

"I can lead the horse to water, but he has to do the drinking himself...

[Rick Willoughby]

Originally Posted by Joakim

You don't happen to have a link for the CFD-results? Or to the porpoising paper?

I did some tests with my Savitsky program just ignoring the flat part and using beta=5.9 and beam to chine. I could reproduce the results from 3 to 7 m/s within +-7% in resistance and +-0.5 degrees in trim. Mostly much more accurately and all the changes due to changes in LCG or displacement were accurately reproduced. The speed for the porpoising onset was spot on.

Also the 2 and 2.5 m/s results were within +-10% and +-0.7 degrees, althogh Savitsky model is no longer valid there (bow touches water).

This is Thornhill CFD data against empirical methods:
http://books.nap.edu/openbook.php?re...10834&page=640

There is also another paper here showing good correlation between CFD and empirical:
http://www.icmrt07.unina.it/Proceedings/Papers/B/14.pdf

Porpoising will get into a more complex regime involving time domain modelling. It is another step in complexity over stable planing.

I have time domain modeling for electronic control systems, mechanical instability, machine modeling and, more recently, some biomechanical modeling so I can lead you into the analysis if you wanted to.

Rick W