Read without ads and support Scribd by becoming a Scribd Premium Reader.
See Premium Plans
 
The Ten Commandmentsof Reliable Speed Prediction
Donald M. MacPherson
VP Technical Director HydroComp, Inc.
INTRODUCTION
As with all technical development, the techniquesused to predict vessel speed have evolved from years of accumulated trial and experience. One of the moresuccessful speed prediction techniques utilizes the
dimensional analysis
of empirical data (from tests or trials). This approach is the principal method used byvirtually all small craft speed/power applicationsoftware.Every designer and naval architect is likely to befamiliar with the more popular dimensional analysisdata sets - Savitsky, Holtrop, Series 60, Series 64. Notalways, however, have these methods been usedsuccessfully. In light of their poor results, some peopleargue that these methods are inadequate to do realspeed prediction. Since it is not possible to fund amodel test program for every design - particularly for small craft where budgets are tight - many designerscontinue to plod along trying to find that successfulcombination of “fudge factors” which generatesreasonable answers.This paper offers up the
ten commandments of reliable speed prediction
. It will illustrate some of themore common shortcomings found in speed predictionmethods and in software using the methods. It will also point out strategies that can be applied to virtually allsmall craft speed prediction.
COMMANDMENT 1 - USE TECHNIQUES THATMIRROR PHYSICAL REALITY
This sounds obvious, doesn’t it? One would think so, but in practice not all techniques reflect a real physical system.
Hull-propulsor-engine equilibrium
The fundamental objective of a speed prediction isto predict the speed of the vessel for a given engine power. To derive engine power from resistance, onetraditional prediction path has been to estimate the OPC by some typical value.The problem with this approach is that theestimated OPC may, or may not, bear any resemblanceto actual figures in practice. A single representativeOPC (such as the ever popular 0.55) cannot possiblyaccount for the many differences in propulsor  performance or in contemporary trends that may effect performance. (Current OPC figures are often in excessof 0.65. This means nearly 15% less engine is reallyneeded!)Increasing engine powers, higher cavitation levels,more cupped propellers, deeper gear ratios and higher  pitches all contribute to changing trends in propulsor  performance and OPC. It is therefore necessary to usean equilibrium performance model where the propulsor is the central figure.
Propulsor 
 
Hull
 
Engine
 In this way, an RPM can be found which satisfiesthe physical thrust equilibrium requirement between thehull and propeller.
 Real 
propeller efficiency, OPC andengine power are then derived from this RPM.
Planing hull trim
Take the case of planing hull trim for the nextexample. Most planing hull resistance methods (e.g.,Savitsky, 1964), include vessel trim as one of the initial(independent) variables. In other words, drag isaffected by trim, and anything that affects trim affectsdrag. In the real example below, the bare hull resistancewas over 5% in error, simply because an equilibriumtrim correction was not applied.Many computer-based speed prediction programsdo not allow for the determination of equilibrium trim.The propeller or waterjet thrust line, appendages, wind,and trim tabs all greatly influence the hull’s trim anddrag. An equilibrium trim technique should beemployed for all planing hull predictions.
Copyright © 1996 HydroComp, Inc. All rights reserved. Presented at the Small Craft Marine Engineering Resistance and PropulsionSymposium, University of Michigan, May 1996.
1
 
0.100.110.120.130.140.150.162 2.5 3 3.5FvR/W4
 
 Figure 1 – Drag with and without trim correction
COMMANDMENT 2 - USE CONTEMPORARYTECHNIQUES
Up-to-date techniques represent the currentconsensus from the various international researchinstitutions. For example, three-dimensional analyseshave replaced two-dimensional, with new friction linesand more accurate model-ship scale corrections.
0.00200.00250.00300.00350.00400.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28FnCt
 
 Figure 2 – Effect on C 
 R
prediction of wrong C 
 F 
and C 
 A
 
Improperly using a new friction line or correlationallowance, however, can lead to less reliable results.Too often, incompatible routines are incorrectly mixedwith each other. In an attempt to use contemporarymethods, C
R 
predictions based on the older ATTCfriction line (Schoenherr, 1932) are routinely married tothe newer ITTC friction line (ITTC, 1957). Add to thisthe use of the ATTC-based traditional C
A
of 0.0004 andsignificantly incorrect predictions can result.The proper approach is to numerically recreate theoriginal model-scale results using the known modellength and friction line. Then full-scale results - freefrom incompatibility - can be built with the designer’s preference to use traditional values or to fully exploitcontemporary practices. Remember to watch out for  pre-set routines that mix old data with new.
COMMANDMENT 3 - USE THE RIGHT KINDOF ALGORITHM
All of the methods described here are based onsome statistical manipulation of model test and trialdata. Hull
 parameters
(e.g., C
B
, C
P
, L/B) are used as theinput side of the statistical analysis. How the test data isgrouped and the parameters analyzed will determinewhat role the method should and should not play in areliable speed prediction.
Systematic series vs. random data
To keep the accuracy of a parametric analysis ashigh as possible, collections of test data are oftengrouped into a
 series
. In a typical small craft series -such as the YP series (Compton, 1986) - a parent hull iscreated that has certain design features, such assectional area curve, turn of bilge, transom immersionor entrance angle. A matrix of models is then built thatvary the principal shape parameters (e.g., L/B, B/T, C
P
)in a systematic fashion to determine the effect of each parameter on resistance.The weakness of a systematic series is apparentwhen the design is of a form that is different from theseries. For example, it would be inappropriate to usethe semi-displacement YP series for a heavy tug.
 Random data
methods, on the other hand, are based on many different hull forms and test results. Thevarious components of resistance (e.g., wave-making,transom immersion, bulb effects, viscous components)are broken down into as many pieces as possible andeach numerically evaluated independent of the others.While not always able to discretely evaluate the effectof a single parameter on resistance like a series can,these methods frequently offer better prediction of the
magnitude
of resistance.
Design or analysis
The choice of a prediction method often dependson the purpose of the task. In a design or feasibilitystudy, the objective might be to determine suitableoverall parameters. A sea trial forensic analysis mightrequire a more detailed review of the entire system performance.
 Design
methods sacrifice precision for the sake of fewer parameters and well-behaved “average” results.Almost all systematic series fall into this category,along with a few random data methods. Choose adesign method for early-stage design studies.
2
 
Analysis
routines are built with sufficientlydetailed components to allow for a finely-tunedanalysis of performance. If a hull closely matches aseries parent, then a series-based method can also beused for an analysis. Select an analysis method for later-stage design using mature data, or for theevaluation of a trial or model test.
COMMANDMENT 4 - USE A METHOD THATCONTAINS A SUITABLE DATA SET
Perhaps the most significant contributor to poor  prediction reliability is the inappropriate selection of the basic prediction formula. In a setting where only afew methods are available and designers apply thesemethods to a broad range of hull types, it is easy to seehow errors can occur.
Hull type
The selected prediction method should be builtfrom hulls that share the same basic character as thevessel under review. You cannot rely on results from amethod derived from a fundamentally different hulltype. Referring to body plans and profiles of theincluded hull forms is the first step to filtering outunsuitable algorithms.Some designers use the well-known Taylor data(Taylor, 1943) for virtually every vessel under consideration. It should be easy to see how using a plumb bow, cruiser stern, 1940’s hull form could leadto trouble predicting drag for new hull types.
Range of parameters
After hull type, the method’s range of data set parameters must be considered. Extrapolating beyondthe scope of the data set is risky.The most critical parameter to maintain is that of speed (typically Froude number). When it is absolutelynecessary to extrapolate speed, a speed-dependentalgorithm (i.e., that where speed is a variable in a singlerelationship) will tend to be better behaved.The obvious way to avoid difficulty is to havemany different methods for all of the hull types of interest and to have feedback regarding the parameter ranges while selecting a method. Certainly the mostcost-effective means to acquiring a large library of routines is through a commercial software developer.Over 50 different bare-hull drag methods have beenidentified and reviewed by the author, so suitablemethods can be obtained for virtually every type of monohull and catamaran vessel.
COMMANDMENT 5 - ANCHOR THEPREDICTION TO TEST DATA
Predictions based on the described numericalmethods should adhere to the extents of the originaldata or significant error is risked. As noted above,robust commercial software packages have numerousmethods so that a suitable method can generally beapplied.Sometimes a designer encounters a vessel that doesnot neatly fit into one of the available parametricmodels. In these cases, techniques exist to improve the prediction by “anchoring” the prediction to known performance.These techniques, often called
aligned prediction
 or 
model correlation
, compare the calculated numericalresults to a model test or sea trial of a similar vessel.The objective is to first test the numerical algorithmagainst known performance and develop a multiplier which is applied to the prediction of the new design. Non-dimensional relationships, such as Froude number and resistance-displacement ratio, are used for thecorrelation.
0.0000.0020.0040.0060.0080.0100.0120.0140.016Actual Holtrop OortCr 
 
 Figure 3a – Supply boat numerical prediction of C 
 R
 
This can be a very powerful technique. The initial performance results are first built from the selectedmethod for the vessel’s parameters. The results are thenrefined to reflect the local features of the tested modelor ship. Of course, the model and subject hull must besimilar in form. It would be inappropriate to correlate atug to a fast ferry. Likewise, the chosen algorithmshould be built from the same basic hull type.The figures compare the prediction accuracy of asupply boat using the Holtrop method (Holtrop, 1984)with all parameters in range and the Oortmerssenmethod (Oortmerssen, 1971) with C
P
and L/B outsidethe range. In addition, Oortmerssen did not have anysupply boat hulls in the data set, while the Holtrop did.
3
Search
Search History:
Searching...
Result 00 of 00
00 results for result for
  • p.
    • Notes
    Load more