Delft Series Resistance Equations - Variety

It might have significant effect for small remainder of big values. I don't understand why they keep one digit for some coefficients, it is methodologically wrong.

 

Depends on the magnitude of the coefficient and the other factor in each term, and the precision desired.

Overall uncertainty of a sum will be dominiated by term with largest uncertainty. So if a term's value is smaller in magnitude compared to the others, the number of significant digitis in that term can be less without adversely affecting uncertainty of the sum. The Delft engineers may have taken this into account when deciding how many significant figures to report for each coefficient. Or it may be an indicator of the level of confidence the values given for each coefficient.

As Uli brought up there are legitimate questions about the accuracy the ITTC 1957 curve provides in estimating viscous resistance. Delft adopted the convention of using 70% of the LWL for the Reynold's number, but then mention in some of their later papers that using 90% of LWL might be more appropriate for some of the models.

An analysis of the uncertainty of the various DSYHS based regression derived formulas for resistance would be very helpful, but would not be a casual undertaking given the complex, non-linear nature of the regression. And then there is the difficulty in finding test cases to compare to where the experimental uncertainty can be estimated.

 

One of the negatives of calculators and computers is the ease with which any number of significant digits can be calculated. Many times I've seen a formula based simple regression of scattered data used to calculate results with precision far greater than than warrented.

"Four significant figures" as a standard for calculations may have originated with slide rules. My recollection is I was generally good for about 3.5 significant figures for most calculations using a slide rule, so writing down 4 signficant digits for intermediate results would ensure that no accuracy was lost, assuming the input and equations used could provide that level of accuracy.

 

It might be rounded up for publication. I do have doubts that 1 significant digit would be enough. Coefficients from early editions of Larsson have more digits at low Fn, so there is not negative RR cases.

Of course I agree that Rn will effect the result at sloe speeds.

 

Depending upon the type of model and speeds, we would set ours nominally around 85-90%.

 

Just one sample from yesterday. I was making new spreadsheet for catamaran resistance calculations using Muller-Graf method. Girl in our office was typing coefficients and she used 0.33 instead of 1/3 everywhere. So the resistance curve looked funny and some points were definitely drop out. Once the problem was corrected, the curve became as expected - smooth with humps where they should be.

So the conclusion is: the more regressors are used in formula, the less predictable the result is if some of significant digits are 'cut'. Keep at least 4 digits in intermediate calculations and 3 digits in final is good engineering style.

 

Delft: probably the reasons why they used 70% of length for Rn is a) underwater shape of canoe body of 1980-es boat that was very pointy at front (actually I used 2 turbulence stimulators on such hulls - on stn.1 and lover part of stn.3) and b) they included appendages in earlier series.

 

It may be concidental that earlier versions had more significant figures and you haven't seen negative residual resistance with an earlier version.

 

May be. But I still believe they should give more digits so there is no doubt. Just present numbers in exponential form instead of rounding up would be proper way.

 

Alik,

Can you give an indication of how one goes about writing such a spreadsheet for Catamarans given the fact that the hull resistance calculations derived from the Delft Series data do not apply?

Where can one find some more info about the Muller-Graf method?

Thanks in advance from an amateur student

 

Just get Muller-Graf or Radojic paper, they were published by SNAME. Some notes and references are also provided in new book by Faltinsen.

 

Gee you are fast.

Thanks.

 

Alik,

I have yet to find the Sname papers but a summary of the Faltinsen book indicates that it talks about power Multihulls. While resistance due to friction and waves would probably be well covered, what about resistance due to leeway or trim? Would these be insignificant?

regards

 

For leeway effect, You might look at ship manoeuvrability hydrodynamics or general hydrodynamics of low aspect ratio bodies. There are number of research for plates and ellipsoids, but there references I know are in Russian.

 

All the different "Delft series"-based Rr formula are in the 2008 InnovSail paper, from the 1975 (Kerwin) one, down to the latest (Keuning & Katgert).

The paper can now be downloaded by anybody that registers on the "Delft series"-website: http://dsyhs.tudelft.nl/dsyhs.php