New Lines Prediction Program

[Leo Lazauskas]

Originally Posted by Remmlinger

Congratulations Mikko and Quequen for the realistic hulls!

It seems that I have to add an additional parameter for the bow-shape, so that the program can handle modern bows with a vertical stem. May be I will find the time during the coming holidays.
Uli

Come on, Uli, Xmas is when you can get some real work done.
You've bombed out the free-loaders on turkey and booze and seen them off
the premises; it's 2am and the house is quiet, and the wife and kids don't
know you are using their computers to run your programs in parallel.
Nerd heaven!

All the best for the holidays,
Leo.

[Remmlinger]

Originally Posted by Leo Lazauskas

Come on, Uli, Xmas is when you can get some real work done.
You've bombed out the free-loaders on turkey and booze and seen them off
the premises; it's 2am and the house is quiet, and the wife and kids don't
know you are using their computers to run your programs in parallel.
Nerd heaven!

All the best for the holidays,
Leo.

How come you know all the details, are you married?
I'll meet you in Nerd heaven (hopefully not too soon)

Nerdy holidays for you
Uli

[Leo Lazauskas]

Originally Posted by Remmlinger

How come you know all the details, are you married?

I've got a reasonable idea of how it works as I am still with the same girl I met 43 years ago.

I haven't had time to try out your program to see how it bolts onto Michlet. Does it produce a table of offsets?

[Remmlinger]

Originally Posted by Leo Lazauskas

Does it produce a table of offsets?

Yes it does, but your splines create a slightly different hull than e.g. Rhino.

P.S.: Good to see you back, your 4 weeks absence was irritating.

[Mikko Brummer]

Originally Posted by Remmlinger

You also showed me a new usage of my program that I did not think of: Reverse engineering! But I have to point out that the hull that you created with my program still differs from the real Dehler 33. It is a specific characteristic of the Dehler 33 that the max. Tcb and the max. Bwl do not occur at the same section. This is something that I can not model with my program, where the main section has max. Tcb and max. Bwl. This seems not to be a big issue since I used the model from my program in a VPP and the results compare well with towing tank measurements.
Uli

If you want to use Ulilines to actually design complete hulls, I find it a bit problematic that there's so little control over the above waterline shape of the hull... for real designs, the Bmax at the deck is often not in the same spot as Bmax at WL. Also, the above WL shape of the hull does affect drag when the boat heels or when in waves, or at higher speeds, as shown for instance in the paper of Keuning attached. Two open 60-footers with in practise identical under water bodies have different drags at higher speeds, when water starts raising along the topsides.

The fact that max. Tcb and the max. Bwl do not occur at the same section may also be a design factor. If so, since the Delft series neglect this factor, all our VPPs could be at err at this point. Maybe top designers understood this long ago and have exploited it in rules like IMS or ORC.

[Remmlinger]

Originally Posted by Mikko Brummer

If you want to use Ulilines to actually design complete hulls, I find it a bit problematic that there's so little control over the above waterline shape of the hull... for real designs, the Bmax at the deck is often not in the same spot as Bmax at WL. Also, the above WL shape of the hull does affect drag when the boat heels or when in waves, or at higher speeds, as shown for instance in the paper of Keuning attached. Two open 60-footers with in practise identical under water bodies have different drags at higher speeds, when water starts raising along the topsides.

The fact that max. Tcb and the max. Bwl do not occur at the same section may also be a design factor. If so, since the Delft series neglect this factor, all our VPPs could be at err at this point. Maybe top designers understood this long ago and have exploited it in rules like IMS or ORC.

Of course you are right Mikko. I sat down over the holidays and tried a modification of my program that also allows for vertical stems without flare. I am almost done, the algorithms finally work well and I get faired lines at the bow. What is puzzling me is the result when I use my sections as input to Rhino. The NURBS-surfaces are wavy, even when corresponding points are lying on a strait line.

Please keep in mind, that the program is a by-product of my optimization program. I am only interested in the wetted part of the hull. In my optimization process computing speed is more important than second order effects on the resistance. Keuning shows that bow flare has an effect at large trim and on added resistance in waves. The effect of large trim would be taken care of in my VPP, because I calculate the input parameters for the Delft-regression from the hull in the heeled and trimmed attitude.
The effect of bow flare on added resistance would disappear unnoticed because Gerritsma's formula for added resistance does not care about bow flare.

If the updated version of UliLines with more control over the topsides at the bow is of any help to you, I can try to speed up my work and release the update asap.
Uli

[DCockey]

Originally Posted by Remmlinger

..... What is puzzling me is the result when I use my sections as input to Rhino. The NURBS-surfaces are wavy, even when corresponding points are lying on a strait line. .....

Interpolating with NURBS or other spline formulations. can lead to oscillations. Try a simple experiment. Create a series of lines which are parallel and planar. Move one line so that it is no longer planar with the others. Now use those lines as input for a NURBS surface. The surface will have waves.

[Mikko Brummer]

Originally Posted by Remmlinger

Of course you are right Mikko. I sat down over the holidays and tried a modification of my program that also allows for vertical stems without flare. I am almost done, the algorithms finally work well and I get faired lines at the bow. What is puzzling me is the result when I use my sections as input to Rhino. The NURBS-surfaces are wavy, even when corresponding points are lying on a strait line.

If the updated version of UliLines with more control over the topsides at the bow is of any help to you, I can try to speed up my work and release the update asap. Uli

Yes, a vertical stem that closes nicely would be of help for me, but I'm not in a hurry, please don't sweat for my sake! If you could post a DelftS_file with the modification for me, I could try how my surfacing treats it.

Please keep in mind, that the program is a by-product of my optimization program. I am only interested in the wetted part of the hull. In my optimization process computing speed is more important than second order effects on the resistance. Keuning shows that bow flare has an effect at large trim and on added resistance in waves. The effect of large trim would be taken care of in my VPP, because I calculate the input parameters for the Delft-regression from the hull in the heeled and trimmed attitude.
The effect of bow flare on added resistance would disappear unnoticed because Gerritsma's formula for added resistance does not care about bow flare.

Yes, I guess I was trying to say that the static wetted part is different from the dynamic one at speed, and hence you cannot entirely ignore the above WL part of the hull.

Interesting to hear that you use coefficients derived from the (static) heeled & trimmed hull as Delft regression input (in the upright formulas?)... do all modern VPPs work that way? As I recall, my (ancient) VPP uses always the upright coefficients, and corrects residuary resistance with an added drag factor due to heel.

[Remmlinger]

Originally Posted by Mikko Brummer

If you could post a DelftS_file with the modification for me, I could try how my surfacing treats it.

I will do that........soon.

Originally Posted by Mikko Brummer

Interesting to hear that you use coefficients derived from the (static) heeled & trimmed hull as Delft regression input (in the upright formulas?)... do all modern VPPs work that way? As I recall, my (ancient) VPP uses always the upright coefficients, and corrects residuary resistance with an added drag factor due to heel.

My VPP calculates the regression parameters in the heeled and trimmed attitude that includes a rough estimate for the wave profile along the hull. The parameters are then used as the input to the upright Delft formula. The results compare better with towing tank measurements than the original Delft-procedure.
Uli

[Remmlinger]

Originally Posted by DCockey

Interpolating with NURBS or other spline formulations can lead to oscillations. Try a simple experiment. Create a series of lines which are parallel and planar. Move one line so that it is no longer planar with the others. Now use those lines as input for a NURBS surface. The surface will have waves.

This is correct for ordinary (non-rational) cubic splines. The big advantage of rational splines is the usage of weights that can force the spline to stay convex without waviness. Unfortunately in Rhino there is no global control over the calculation of the weights. One would have to pick each individual control point and change the weight individually. On large grids this is not feasible. There are programs that chose the weights automatically in a manner that preserves convexity. I do that in UliLines but this seems to be beyond the capabilities of Rhino.
Uli

[DCockey]

Originally Posted by Remmlinger

This is correct for ordinary (non-rational) cubic splines. The big advantage of rational splines is the usage of weights that can force the spline to stay convex without waviness. Unfortunately in Rhino there is no global control over the calculation of the weights. One would have to pick each individual control point and change the weight individually. On large grids this is not feasible. There are programs that chose the weights automatically in a manner that preserves convexity. I do that in UliLines but this seems to be beyond the capabilities of Rhino.
Uli

Any suggestions for sources of information about the methods for chosing the weights of rational interpolating splines to enforce convexity and maintain the splines passing through the designated points? I'm always interested to learn more about splines. My usual reference for computational geometry is Computation Geometry for Design and Manufacture by Faux and Pratt but it does not cover rational B-splines.

[fcfc]

Originally Posted by Mikko Brummer

Interesting to hear that you use coefficients derived from the (static) heeled & trimmed hull as Delft regression input (in the upright formulas?)... do all modern VPPs work that way? As I recall, my (ancient) VPP uses always the upright coefficients, and corrects residuary resistance with an added drag factor due to heel.

Seems current trend:

See www.orc.org/minutes/ITC%202012.pdf , I a waiting for ORC VPP2013 documentation to be published to learn more.

Seems they no longer use delft series upright with a bunch of parameters. Only a dynamic waterline length, LVR (DLR the other way), and BTR (beam/dratft ratio). And they use heeled hull parameters, not uprigh hull parameters with heel correction.

BTW, does anyone know how LSM (second moment length) are computed ?
There are some clues on page 21 chapter 4.2.2.4. http://www.orc.org/rules/ORC%20VPP%2...ion%202012.pdf
What I am missing is where is zero x stem or stern, and what unit (meter, feet).
And how is LSM0 is computed. Is it a recursive computation, since depth correction for section does include LSM0 term ?

Thanks.

[Remmlinger]

Originally Posted by DCockey

Any suggestions for sources of information about the methods for chosing the weights of rational interpolating splines to enforce convexity and maintain the splines passing through the designated points? I'm always interested to learn more about splines. My usual reference for computational geometry is Computation Geometry for Design and Manufacture by Faux and Pratt but it does not cover rational B-splines.

The paper:
John C. Clements "Ship Lines Using Convexity-Preserving Rational Cubic Interpolation", J. of Ship Research, Vol. 35, No. 1, March 1991, pp. 28-31
presents an algorithm for the determination of the weights (called Tau in the paper). The paper also contains a Fortran-program that can be used.

[DCockey]

Remmlinger, thanks for the reference. I'll have to obtain a copy and read it.

Another approach is to use ordinary/non-rational cubic B-splines and select the number of intervals and parametarization based on preserving convexity and/or minimizing/eliminating waviness. This is in contrast to the usual approach of basing the number of intervals and parametarization directly on the interpoating points, and then determining the splines (non-rational or rational) using the previously determined intervals and parametarization. I don't know but wouldn't be surprised if someone has pursued this or a similar concept and published something. I was part way there about twenty five years ago but then dropped it when it was no longer needed. Unfortunately I can't find my notes from that time.

[Ad Hoc]

Originally Posted by DCockey

I'm always interested to learn more about splines.

You may find these up your alley then:

"Application of a real-coded genetic algorithm for the fitting of a ship hull surface through a single non-uniform B-spline surface", Le T-H., Kim D-J.
JMST Vol 16, No.2 2011.
and
"Surface intersection in geometric modeling of ship hulls", Ventura M., Guedes Soares C.,
JMST, Vol 17, No.1 2012.