CFD and CAD in Ship Design

DEVELOPMENTS IN MARINE TECHNOLOGY

Vol. 1 Marine and Offshore Safety (P.A. Frieze, R.C.McG regor and I.E.Winkle, Editors)

Vol. 2 Behaviour of Offshore Structures (J.A. Battjes, Editor)Vol. 3 Steel in Marine Structures (C. Noordhoekand J. de Back, Editors)Vol. 4 Floating Structures and Offshore Operations (G. van Oortmerssen,

Editor)Vol. 5 Nonlinear Methods in Offshore Engineering (S.K. Chakrabarti,

Editor)Vol. 6 CFD and CAD in Ship Design (G. van Oortmerssen, Editor)

ELSEVIER

Developments in Marine Technology, 6

CFD and CAD inShip Design

Proceedings of the International Symposium on CFD and CAD in Ship DesignWageningen, The Netherlands, 25-26 September 1990

Edited by

G. van Oortmerssen

Maritime Research Institute Netherlands (MARIN), Wageningen, The Netherlands

Amsterdam-Oxford-NewYork- Tokyo 1990

~----~-_,....-------~

ELSEVIER SCIENCE PUBLISHERS avSara Burgerhartstraat 25P.O.Box 211,1000 AE Amsterdam, The Netherlands

Distributors for the United States and Canada:

ELSEVIER SCIENCE PUBLISHING COMPANY INC.655 Avenue of the AmericasNew York, NY. 10010, U.S.A.

/

i.J '. , ( . ,-

"

ISBN: 0444887652 (Vol. 6)ISBN 0-444-42384-2 (Series)

e Elsevier Science Publishers B.V., 1990

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted inany form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior writtenpermission of the publisher, Elsevier Science Publishers B.V'/Physical Sciences and Engineering Division,P.O. Box 1991, 1000 BZ Amsterdam, The Netherlands.

Special regulations for readers in the U.S.A. - This publication has been registered with the Copyright ClearanceCenter Inc. (CCC), Salem, Massachusetts. Information can be obtained from the CCC about conditions under whichphotocopies of parts of this publication may be made in the U.S.A. All other copyright questions, includingphotocopying outside of the U.S.A., should be referred to the copyright owner, Elsevier Science Publishers B.V.,unless otherwise specified.

No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter ofproducts liability, neg licence or otherwise, or from any use or operation of any methods, products, instructions orideas contained in the material herein.

Pp. 3-24. 25-48, 49-64, 75-86, 87-98, 233-246: copyright not transferred.

J .... {C -~'"

-I

" /, .- __ \

II I"J _,

i > .-

PREFACE

In recent years rapid advances have been made in computer applications in ship hy-drodynamics and ship design, due to an increase in the performance of computerhardware.

In the field of ship hydrodynamics, various methods have been developed for theprediction of calm water resistance and manoeuvring characteristics. These methodsrange from rather simple empirical methods to very advanced Computational FluidDynamics (CFD).

In ship design, Computer Aided Design (CAD) applications were mainly focused onthe description of the geometry of the ship and the calculation of hydrostaticproperties. Much attention has been given to drawing systems and connectivity tosystems for supporting the production process of ships (CAM).

Integration of hydrostatic calculations in ship design software offers new possi-bilities to ship designers for optimizing the design of ships.

The Maritime Research Institute Netherlands (MARIN) considered it timely to organ-ize an International Symposium on CFD and CAD in Ship Design, with the aim to pro-vide a forum for exchanging ideas and information between developers and users ofCFD and CAD systems.

The papers have been arranged in four sessions:

- CAD Developments- User Experience with CAD and CFD- CFD Developments- Future Developments

MARIN wishes to express its gratitude to all who have contributed to the successof this Symposium, either as author of a paper, as session chairman, as discusseror as participant.

Dr. Gerard van OortmerssenChairman, Organizing Committee

v

CONTENTS

KEYNOTE ADDRESS

On the Integration of CFD and CAD in Ship Design

B. Johnson (U.S. Naval Academy)

SESSION I: CAD DEVELOPMENTS

The Development of CAD-Tools at MARIN

G. Glijnis (MARIN)

Concept Exploration Models for Merchant Ships

c. Georgescu (Technical University Delft), F.J.P.H.G. Verbaas

(MARIN), and H. Boonstra (Technical University Delft)

Geometric Design of Hull Forms Using Partial Differential Equations

M.I.G. Bloor, and M.J. ~ilson (University of Leeds)

Design and Representation of Hull Form Using Rational Cubic

Bezier Curves

J.S. Kouh, and s.~. Chau (National Taiwan University)

PROSEL - Propulsion Installation Selection

J. Klein ~oud, C.C.P.J.M. Landa (Technical University Delft),

and F.J.P.H.G. Verbaas (MARIN)

SESSION II: USER EXPERIENCE VITH CAD AND CFD

Practical Experience with a Hull Form Surface Modelling

Ch. Rossier (Chantiers de l'Atlantique)

vii

3

25

49

65

75

87

101

viii

Impact of CFD on Aerodynamic Design

N. Voogt (Fokker Aircraft B.V.) 107

Practical Applications of an Integrated Design System in

Ship Hydrodynamics

I. Kuutti (Masa-Yards Inc.) 117

SESSION III: CFD DEVELOPMENTS

Automatic Panel Generation for Seakeeping and Wave Resistance

Calculations

P.S. Jensen (Danish Maritime Institute) 133

Incorporation of Seakeeping Theories in CAD

O. Faltinsen (Norwegian Institute of Technology), and

T. Svensen (MARINTEK A/S) 147

Free-Surface Potential Flow Calculations for Merchant Vessels

W. van den Berg, H.C. Raven, and H.H. Valkhof (MARIN) 165

SESSION IV: FUTURE DEVELOPMENTS

IDEAS: Interactive Design, Evaluation and Analysis System

N. Salvesen (Science Applications International Corporation) 185

Feature-Based Design Synthesis and Performance Analysis of

Underwater Bodies

D.E. Calkins, and M.A. Razzaghi (University of Washington) 197

The Development of a Design Guidance System for the Early Stages

of Design

B. Bras, W.F. Smith, and F. Mistree (University of Houston) 221

Predicting the Hydrodynamic Performance in Ship Design: Tests

or Computations?

G. van Oortmerssen (MARIN) 233

KEYNOTE ADDRESS

CFD and CAD in Ship DesignG. van Oortmerssen (editor)

Elsevier Science Publishers B.V .• 19903

ON THE INTEGRATION OF CFD AND CAD IN SHIP DESIGN

BRUCE JOHNSON

U. S. NAVAL ACADEMY

KEYNOTE ADDRESS

1. INTRODUCTION

I am deeply honored to be invited to address thisInternational Symposium on CFD and CAD in ShipDesign. I have never written a flow code or a CADpackage nor do I supervise graduate students whoperform these tasks. My modest qualifications formaking the following comments are based on1) a working knowledge of turbulence research [1]and continuum mechanics a number of years ago and2) more recent experience in using various CFD(Computational Fluid Dynamics) and CAD (Computer-Aided-Design, in my case primarily hull geometry) codesand observing the inefficiencies caused by the lack ofdesign system integration. Attempts to improve theefficiency of decision support tools used in thehydrodynamic design process for ships and vehicles havebeen a central focus of my professional life for nearlytwenty years.

Quoting from a SNAME 1990 annual meeting paperwhich I co-authored, [2]

"The advantages of an integrated hullformhydrodynamic design system which fully accessesand utilizes historic data are obvious. Anexcellent example of what can be accomplished(even without full integration) is the design ofthe SL7 class of high-speed containerships forSeaLand by MARIN (Marine Research Institute,Netherlands) [3]. Despite extensive modeltesting of different bulb and stern configurations,the design goal for speed was not being met. Ina very constrained time frame MARIN developeda hull form based on its historic data which,when tested, proved significantly better than anyprevious design, and met the required designspeed of 33 knots. Note that 'extensive modeltesting' of new hull forms was used in a 'what ifmode because bulb and stern configurations for33 knot ships represented a case where littleguidance was available from the various 'firstlevel' computer codes based on regressionanalysis of historical data. Consider how much

more efficient this process could be if CFD toolssuch as those discussed in this symposium wereintegrated into CAE/CAD/CFD/CAM/DBMSsystems which would be used to solve suchproblems in the future.

"Consequently, hull geometry codes, CFD codes,model tank testing and other hull form designtools would be more useful to the ship designcommunity if they were efficiently linked to eachother and to standardized neutral format data-bases containing hull form and propellergeometry data from previous ship designs and thecorresponding hydrodynamic performancecharacteristics from both model and full scaletests. Using computer-linked combinations ofdesign tools, a naval architect could accuratelyand expeditiously optimize hydrodynamicperformance and compare his/her design withother ship hull forms with similar missionrequirements." [2]

There are large international programs attempting tostandardize the electronic transfer of manufacturing(product definition) data from preliminary designthrough production [4], [5]. These phases of theproduction cycle consume approximately 95% of thetotal costs, so making data transfers more efficient canmake for large cost savings, even when the designrequires only a few changes before entering production.It is estimated, however, that concept design, which isgenerally less than 5% of total costs, commits 40% to60% of later spending and that the cost of designchanges increases by an order of magnitude at eachmajor stage of design and production. [6]

Let us consider the historical and current developmentsin integrated vehicle design systems, especially thoseaspects of hydrodynamic analysis during hullform designwhich can make "Concurrent Engineering" cost effectivefor the ship design process.

4

2. THE DEVELOPMENT OF INTEGRATEDVEHICLE DESIGN SYSTEMS

One of the first large scale attempts to reduce time andcost and consequently the number of "bottlenecks" in thedesign process for complex systems was the IPAD(Integrated Programs for Aerospace-Vehicle Design)Project carried out from 1971-1984 as a $30Mcooperative government/industry venture under thesponsorship of NASA, and later the U. S. Navy (seeFulton [7]). The principal contractor on the projectwas the Boeing Commercial Airplane Company from1976-1984. According to Miller, [8]

''The requirements were for a general purpose,interactive computer aided engineering (CAE)system capable of supporting engineering dataassociated with the design process, includingmanufacturing interfaces. The system wouldserve management and engineering staffs at alldesign levels as well as the downstreammanufacturing processes." [8]

The goal of the project was to improve aerospaceproductivity through better CAD/CAM informationmanagement including the development of a prototypefor a future integrated CAD system.

The IPAD project envisioned a global databasemanagement system built around a common RelationalInformation Management (RIM) database manager(Blackburn [9]). Various schemes for storing thenecessary geometric data included the use of the InitialGraphics Exchange Specifications (IGES) derived fromexperience gained with the Boeing CAD/CAMIntegrated Information Network (CIIN) and GeneralElectric's Neutral Database [10] as well as otherapproaches [11]. Fulton [7] concluded that,

•...a single product definition database containingan electronic description of the designed productsthat are being constructed or manufactured is akeystone to the successful utilization ofCAD/CAM technology .... (however) a key datamanagement requirement not yet commerciallyavailable is the ability to manage unifiedCAD/CAM information distributed acrosscomputers of different manufacture ....(thus) theoriginal intent of an integrated software systemwas abandoned as being too expensive,inappropriate for NASA, and not the best use ofthe limited R&D resources." [7]

2.1 Hullform Design/Manufacturing Support DatabaseDevelopments

Today, there are numerous design/manufacturingsupport databases for many aspects of ship, aircraft, and

land vehicle design. Most frequently they are used tostore the details of the post-conceptual design processfor use in the manufacturing process: drawing numbers,weight estimates, arrangements plans, parts lists,structural details and other information needed toproduce a product or system. If one considers the stagesof the so-called basic ship design spiral (see Figure 3from Taggert [12]), existing integrated databasedevelopment efforts have largely concentrated oninterfacing the output of the preliminary design processwith the contract and detail design and manufacturingprocess, i.e. the CAD/CAM/CIM (Computer-AidedDesign/Computer-Aided Manufacturing/ComputerIntegrated Manufacturing) link. It is in this part of theship production life cycle that cost savings have beenconsidered most important.

2.2 Proprietary Hydrodynamic Hullform DesignSupport Database Developments

To support the research, analysis and developmentphases of the hullform design process, a number oforganizations have built databases to store towing tankand wind tunnel data. These databases, discussed byJohnson [2], are used by individual organizations tomake comparative performance studies of variousvehicles whose characteristics lie within the bounds oftheir databases. Except for recent discussions within theInternational Towing Tank Conference community[13] however, proposals to develop standardized neutraldata exchange formats for the comparativehydrodynamic analysis stages of conceptual design andfor the ship performance estimation phase of preliminarydesign have not been pursued. This may be the result ofnot only the proprietary nature of a particular shipdesign but also the fact that the early stage comparativeanalysis studies represent a small fraction of the lifecycle cost of a large ship and, with few exceptions, thehydrodynamic analysis comparisons are not passed alongto the next stage of ship design/production. Regardlessof the reasons, however, inter-database comparisons arepresently limited to "hardcopy" data exchanges sincenearly all proprietary hydrodynamic design systemspresently lack the ability to exchange digitalhydrodynamic data with each other.

Thus it is not surprising that one cannot yet identify aship which was designed using an integratedCAE/CAD/CFD/CAM/DBMS system which directlylinks validated computational fluid dynamics (CFD) toolsto a database containing easily retrieved information onthe geometry and performance characteristics ofpreviously tested vehicles with similar missionrequirements.

What is surprising, however, is that integratedaerodynamic design database systems are apparently not

yet operational in the aircraft industry in spite of theprogress made during the IPAD project. One mightexpect that integrated aircraft design and manufacturingcompanies would have fewer proprietary data exchangeproblems than those non-integrated industries whichhave separate design, testing and constructionorganizations. However, because designer/buildersfrequently have industrial partners and sub-contractorswith dissimilar computer environments, performancedata exchange problems are nearly universal during theconceptual design phase for all types of performanceoriented vehicles.

2.3 Proprietary Hydrodynamic Design Support Systems

2.3.1 MARIN (Marine Research Institute, Netherlands)

In 1983, MARIN began developing HOSDES, anintegrated concept design system for high speed navalships with support from the Dutch Navy. HOSDES isone of the most highly integrated ship design systemsdeveloped to date. HOSDES includes a CAD (graphics)system, CADAS (Computer-Aided Design Analysis)modules, and HYDDB (Hydrodynamic DesignDatabase) capabilities. A number of papers (Koops[14], Ooman [15], van Oossanen [16], Koops[17], van Oortmerssen [18], Glijnis [19], Koops[20]) have described the HOSDES system whichincludes the following elements:1) Support Systems including a System Management

Subsystem, a User-Interface ManagementSubsystem (UMIS), a Swedish relational DBMS"MIMER", and a Designer Guidance Subsystem(the "Assist" Expert System).

2) Application Systems including variouscomputational levels of software analysisprograms for hull geometry, powering, massdistribution, hydrostatics, motions, strength,dynamics, endurance and engine configuration.

The HOSDES system has now been expanded to includemerchant ships (MARDES, discussed in Glijnis [19]).We all look forward to learning more about this systemduring this symposium.

2.3.2 BMT (British Maritime Technologies)

The various CAE/CAD/CAM software packagesavailable from BSRA/BMT have been discussed in anumber of papers [21], [22], [23] concerningthe BRITSHIPS/BRITDES systems. The BRITSHIPSCAD/CAM system has a direct interface with theBRITDES system. The HYDDB STARTER databaseassociated with the BRITDES system is discussed byJohnson [2].

5

2.3.3 Other Hydrodynamic Design Support Systems

Johnson [2] also summarizes much of the publishedliterature concerning various other proprietaryhydrodynamic design support systems. Typical PersonalComputer/workstation based hull fairing programs arereviewed by Ames [24].

I look forward to hearing about several other integratedhydrodynamic design support systems being discussed byseveral authors during this symposium.

2.3.4 Integrated "Hydro-Numeric Design" Systems

Nils Salvesen of SAiC was scheduled to discuss IDEAS:Interactive Design, Evaluation and Analysis System atthis symposium. IDEAS is currently under developmentin Annapolis, MD and will be discussed by Fritts [25]during the mini-symposium on "Hydro-Numeric Design"at the 1990 SNAME Annual Meeting. IDEAS willincorporate various flow codes for resistance, seakeepingand performance assessment tied to a user-friendly hullgeometry design package "FASTSHIP" (in use byundergraduate naval architects at the U. S. NavalAcademy). IDEAS will eventually be linked to the U. S.Navy/U. K. Navy Hull Form Design Database (HDDS)[2].

Also during the 1990 Annual Meeting of SNAME, LarsLarsson [26] will discuss the SHIPFLOW designsupport system developed by SSPA/FLOWTECHInternational, AB., which consists of an extensive set ofCFD tools for calculating the potential flow with freesurface, the boundary layer and the viscous flow in thestern region of a ship. (These methods will be discussedin the next section of this paper.) The degree ofintegration with a CAD system for hull geometry is notmentioned in an early version of the paper.

3. CURRENT STATUS OF CFD IN VEHICLEDESIGN

The generally acknowledged leaders in usingcomputational fluid dynamics methods as significanttools in the design of production vehicles are the aircraftand high performance yacht design communities. CFDhas been effectively used in partnership with wind tunneland towing tank testing to optimize those aspects ofvehicle performance which depend on a detaileddescription of the flow fields. The documented successstories which are validated to some degree by full scaleperformance trials will be reviewed in this section of thepaper.

6

3.1 The Use of CFD in the Aircraft Industry

The valuable and recent survey of "CFD Application toComplex Configurations" was authored by EdwardTinoco [27] of the Boeing Commercial AirplaneCompany. According to Tinoco,

"CFD has joined the wind tunnel and flight testas a principal technology for aerodynamicdesign .... Airplanes are flying today that may nothave been deemed practical if not for the insightprovided by CFD studies. On the basis of CFDanalysis, modifications such as the addition ofnew engines, antennas, or sensor pods have beenincorporated into existing airframes with little orno wind tunnel testing before flight... ..To aid inthe design of such complex configurations, theengineering community needs computationaltools that faithfully represent the dominant flowphysics and the required geometric complexityand are capable of providing reliable solutions ina timely, affordable manner. This is a difficultand as yet unfilled order by any single method.The approach in industry has been to assemble acollection of CFD tools that meet the precedingrequirements to varying degrees." [27]

Tinoco then makes the distinction between what heterms "production tools" which are well documented,validated as to range of applicability, and can be run by"nonexpert" CFD users. Production tools include mostlinear panel methods and many full-potential codes."Expert user" codes generally require complex field gridgeneration, run on a supercomputer, and include codesbased on the Euler and/or Navier-Stokes equations.Again quoting from Tinoco,

"The choice of what level of flow physicsformulation to use on a particular problemrequires the user to have some idea of whatphysics is relevant... .. Experience has shown thatmethods based on the linear potential and full-potential formulations can be routinely convergedto a higher level of accuracy than has beenpractical with methods based on the Euler orNavier-Stokes formulation." [27]

Another recent paper by Goldhammer and Rubbert[28] of the Boeing Company makes the point thatalthough the fluid mechanics community has always beeninterested in modeling fluid flow problems as accuratelyas possible,

"The connection between physical insight and themathematical formulations is perhaps not asstrong in today's CFD researcher. The verysolution of the basic fluid mechanics equationshas been taken over by mathematicians andcomputer scientists who program supercomputersto solve the equations by numerical

approximation. But the solutions are really onlythe beginning of providing an engineering tool toimprove the technology of airframe design." [28]

This concern was also expressed by Jameson [29] inhis keynote address to the 8th AlAA CFD Conferencein 1987.

"...the disparity of scales in turbulent flow is solarge that direct simulation is not likely to befeasible without radical developments incomputer technology Progress in simulatingseparated viscous flows may now be moredependent on improvement in turbulencemodeling than it is on algorithm development."[29]

The Advisory Group for Aerospace Research &Development (AGARD) has sponsored numerous paneland working group meetings of specialists in variousareas of CFD. Summaries of two recent meetingsprovide a good state-of-the-art view of critical issues inthe use of CFD in vehicle design. AGARD AdvisoryReport No. 257 [30] is a technical evaluation reportof a 1988 meeting of the Fluid Dynamics Panelconcerning the validation of CFD codes. One of thepanel co-chairmen, R. G. Bradley, made the followingdistinction between "Validation" and "Calibration" whichis applicable to ship hydrodynamics as well,

"Four categories of experiments are required toachieve a mature CFD capability:1. Experiments to understand flow physics2. Experiments to develop physical models3. Experiments to calibrate CFD codes4. Experiments to validate CFD codesBounds of errors, for numerical CFD solutions aswell as for experimental data, have to beevaluated carefully." [30]

Although billed as a CFD validation symposium, theevaluation report by Sacher [30] observed that at least50% of the papers were commercials for the hugevariety of existing' computer codes, i.e., "Software Power".(This problem exists at many conferences on numericalmethods in both aero and hydro dynamics.) Among hisconclusions Sacher observed that,

"Databases for flow code validation have to bespecially designed, carefully performed during thewind tunnel tests, and all boundary conditionshave to be specified in addition ....(in this respect)'External flow people' seem to be betterorganized than 'internal flow engineers'." [30]

This comment was made in reference to the existence ofrelatively adequate experimental databases (formatunspecified) specially tailored for the validation ofvorticity dominated external flow codes. Sacher notedthe lack of experimental databases for validation ofinternal flows (especially at high Reynolds numbers)

where the physics is dominated by large viscous shearlayer regions and viscous wakes.

AGARD Advisory Report No. 256 [31] is a State ofthe Art Technical Status Review on Drag Prediction andAnalysis from Computational Fluid Dynamics withforward and conclusions written by J. W. Slooff of theNetherlands. A number of Slooffs conclusions areworth repeating for this symposium:

• "Accurate and consistent computation throughCFD of (absolute) drag levels for complexconfigurations is, not surprisingly, beyond reachfor a considerable time to come [because ofdifficulties with grid generation, turbulencemodeling, grid resolution, speed and economicsof computation J.

• "For separated flows, inadequate turbulencemodelling in combination with inappropriate gridclustering and refinement are problem areas evenin 2D airfoil flow where most but not all codescan now predict drag with an accuracy of withinabout 5%. (,But we don't fly airfoils!')

• "Navier-Stokes codes typically do not (yet)involve drag prediction except for 2D airfoilflows. Even then they do not do a better jobthan zonal methods involving potential flow orEuler schemes coupled with boundarylayers ....For (Reynolds-averaged) Navier-Stokesmethods the identification and quantification ofthe viscous, induced and wave drag componentsis as yet unclear and might even be impossiblewithout introducing certain assumptions withrespect to the asymptotic structure of the flowfield." [31]

Charles Boppe [32] of Grumman Corporation agreeswith Slooff that consistent and accurate CFD predictionsof absolute drag level for aircraft configurations arecurrently beyond reach. In other papers concerningcomputational aerodynamic design [33], Boppe citesthe successful use of CFD in solving design challengesfor the X-29 forward swept wing demonstrator aircraftand the evolutions in the Gulfstream series corporate jet.Judging from the tentative program, we will hear othersuccess stories concerning the integration of CFD intothe vehicle design process.

3.2 The Use of CFD in Yacht Design

The dramatic loss of the America's Cup to theAustralian challengers in 1983 proved to be a "milestoneevent" in the history of the use of CFD in marine design.The integration of CFD methods in combination withtank testing as discussed by van Oossanen [34] andSlooff [35] radically and permanently changed themethodology for high performance yacht design. Dutchnaval architects and aerodynamicists developed the

7

complimentary use of CFD tools and tank testing todesign the famous ''winglets'' on Australia II's keel.Slooff and van Oossanen became "media personalities"for a short time and were sporting enough to share theirnew methods [34, 35] with the Americans, who promptlywent to work on their own versions of this Dutchmethodology and proceeded to win the Cup back in 1987[36], [37], [38], [39]. This integrated useof CFD, tank testing and performance analysis codes isnow an essential part of high performance yacht design.(See, for example, Greeley [40].)

3.3 The Use of CFD in Ship Design

The state-of-the-art (as of 1987) concerning the role ofCFD in Ship Design was summarized by Morgan andLin [41] during the Group Discussion on the Impactof Numerical Techniques in Tankery during the 18thI'Tl'C, At that time Morgan and Lin concluded,

"In general, one can say that most solutions todaywhich have practical application are solutionsderived from linear theories where viscosity isneglected .....Although more work needs to bedone for blunt bows and transom sterns, thewave resistance of fine form ships can often bepredicted with a degree of accuracy that iscomparable to that measured experimentally."...In all viscous flow calculations at anysignificant Reynolds number it is presentlynecessary to resort to Reynolds averagedequations rather than the full Navier-Stokesequations. As a result, some sort of artifice,generally referred to as turbulence modeling,must be used to approximate the turbulence inthe flow."...The computational tools now available alwaysgive better results to problems where viscouseffects are small. Examples are the prediction ofship motions, wave resistance, and propelleraction." [41]

Various ship motions programs, generally based onlinear strip theory, are regularly used during the shipdesign process. The effect on ship motions of significantchanges in hull form geometry is predicted reasonablywell for moderate sea conditions. One must be careful,however, about how a particular version of a codespecifies the directional characteristics of the seaconditions. The severe rolling problems caused by theaddition of blisters to the U. S. S. Midway (Ricketts[42]) illustrates the need to number and catalog thecapabilities and limitations of various versions of allCFD codes. According to Ricketts,

"It should be noted, however, that theseassessments assumed a theoretical short-crestedseaway with a cosine-squared energy spread anddid not account for frequent real-world situations

8

where a low-frequency ground swell is present incombination with a local wind-generated sea, thetwo coming from very different directions." [42]

None of the existing programs appear to be sensitive tosmall changes in hull geometry, but seakeeping modeltests are not all that sensitive to small geometry changes,either. Time domain predictions of large amplitudemotions in non-breaking seas as discussed by Lin andYue [43] show promising initial results, but are so farlimited to regular wave examples for relatively simplehull forms. A major development in this method is anautomatic geometry repanelizer to account for changingunderwater geometry as a result of the time dependentfree surface.

The prediction of the motions of small craft in breakingextreme seas is a long way from being solved. Althoughthe use of potential flow theory to describe steep gravitywaves is the subject of validation studies (Dommermuth[44]), the more sophisticated types of ship motionanalysis codes are not adequately validated andtherefore contribute little to the hullform design process.Part of the validation problem is the lack of adequatefull scale ship motions data in measured directional seas(Kjeldsen [45]). I look forward to the futureintegration of directional sea data (with and withoutextreme waves present) and time domain ship motionscodes for the prediction of ship motions in realisticseaways. At the present time we can only speculatewhether or not this problem will be computable (seesection 4.1) at some future date.

On the other hand, wave resistance codes appear to besensitive to small changes in hull form geometry such asbow bulb configurations for high speed ships (Hoyle[46]). These potential-flow codes are gaining acceptanceby the design community, especially as an "order ofmerit" ranking technique for hull form candidates inorder to reduce the scope of model testing. They havebeen adapted to transom sterned ships (Cheng [47]),yacht testing (see previous section) and in partiallyintegrated systems (see FLOWTECH discussion below).Slender-ship codes show promise (Noblesse [48]) inthe prediction of far-field wakes, but presently lack theability to correctly predict sinkage and trim, as near-fieldpotential-flow codes can do, especially if they areiterated using the correct free surface boundarycondition.

As to the impact of CFD on propeller design, Morganand Lin [41]' stated,

"The use of propeller design methodology basedon lifting-line and lifting-surface theories hasbecome routine for conventional propellers.These techniques have led to the practicaldevelopment of the highly-skewed propeller for

reducing vibration and cavitation. The design ofa propeller operating in uniform flow using CFDis also a matter of routine. Nevertheless, asthese theoretical techniques still depend on thelift-curve slope being independent of Reynoldsnumber, difficulty remains in using thesetechniques to predict the performance of anexisting propeller with arbitrary camber lines."

Morgan and Lin also discuss the difficulties in predictingthe effects of cavitation and propeller-induced vibrationssince these phenomena depend on viscous flow effects,hull-propeller and wake interactions.

As to the future of CFD developments, Morgan and Linconclude,

"It will be many years before the full Navier-Stokes equations can be solved on a computer atthe ship-scale Reynolds numbers to the detailnecessary ......the progress toward the solution ofthe Navier-Stokes equations for complexgeometries at high Reynolds numbers has beendisappointingly slow......(however) if the CFDprediction does not compare well with theexperiment, there is always the question ofwhether the problem is with the theory or theexperiment, or both." [41]

As an alternative to solving the viscous equations ofmotion around the entire ship hull, Larsson and his co-workers at SSPAjFLOWTECH [26], [49] haveproposed dividing the flow into three zones, a procedurealso used by Hoekstra [50] of MARIN and similar tomethods used by the aircraft industry as discussed at the1987 AlAA CFD Conference [29]. Larsson'sSHIPFLOW method defines the following separateregions which have defined interactions within thevarious codes:Zone 1- the potential flow (including free surface)outside the boundary layer,Zone 2- the boundary layer over that portion of the hulllength where boundary layer integral methods may beapplied, i.e., the layer is thin relative to the local radiusof curvature and no separation is present, andZone 3- the stern and wake regions where the flow fieldmust be calculated using a form of the Navier-Stokesequations.

This integrated CFD system has been used in variousforms for more than 30 projects since 1983. SHIPFLOW [51]does not currently predict the correct inflow into thepropeller of ships with significant bilge vortices.However, the iterative methods used to calculate thefree surface wave greatly improve the prediction ofsinkage and trim. The various references toSHIPFLOW do not discuss the CAD interface in anydetail.

4. MATHEMATICAL MODELING IN CLASSICALMECHANICS

4.1 The Limits of Computing

Recently I read a fascinating book entitled TheEmperor's New Mind [52] by Roger Penrose, RouseBall Professor of Mathematics at Oxford University. InPenrose's view,

"...it is our present lack of understanding of thefundamental laws of physics that prevents usfrom coming to grips with the concept of 'mind'in physical or logical terms." (page 4)

The book reviews the present status of our knowledge ofthe existing laws of physics and discusses thecorresponding limits placed on our knowledge of thephysical world and on the future possibilities for artificialintelligence.

In reviewing the basis for recursive and non-recursivemathematics, Penrose makes an important distinctionbetween the deterministic nature of many of the existinglaws of classical (Newtonian) physics and whether or notvarious physical phenomena are algorithmic in a usefulway, i.e.,

"Computability is a different question fromdeterminism." (page 170)"...'determinism' means that initial data at oneparticular time completely fix the behaviour at allother times." (page 214)"There is a sense, however, in which this world is'non-computable' in practice. This arises fromthe fact that the accuracy with which the initialdata can be known is always limited." (page 173)"....no matter how accurately we know the initialstate of a system (within some reasonable limits),the uncertainties will tend to grow in time andour initial information may become almostuseless. Classical mechanics is, in this kind ofsense, essentially unpredictable. (Recall theconcept of' chaos' considered above.)" (page 183)"The future behavior (of chaotic phenomena)would still be determined, right from the big bang,even though we would be unable to compute it."(page 432) [52]

Penrose then makes the case that true intelligencecannot be present unless accompanied by consciousnesswhich includes essentially non-algorithmic ingredients.Thus true intelligence (including judgement-forming)cannot be properly simulated by algorithmic means. Infact,

"...the decision as to the validity of an algorithmis not itself an algorithmic process!" (page 414)

This conclusion has important implications for thediscussion on future directions in "intelligent" ship designsystems (see Section 6 below).

9

4.2 The Equations of Motion for CFD

The major focus for solving viscous flow problems inship hydrodynamics appears to center around attemptsto use some version of the Navier-Stokes equationswithout any real effort to ask the question, "which formof the equations of motion contains a reasonable modelof the fluid flow conditions present in the problem athand"? Itwould appear that most investigations regardthe Navier-Stokes equations as the basic equations ofmotion rather than a very special form for which thenormal and shear stresses are linear functions of thedeformation rate tensor. Let us review the mathematicalbasis for various formulations of Newton's Second Lawas applied to continuum mechanics in general and fluidmechanics in particular. Excellent treatments of thesubject are found in Rutherford Aris' book Vector.s,Tensors, and the Basic Equations of Fluid Mechanics[53] and James Serrin's "Mathematical Principles ofClassical Fluid Mechanics" in Handbuch der Physik, Vol.VIIIjl. [54] Figure 1, which I developed whileteaching a high level undergraduate fluid mechanicscourse over 20 years ago, illustrates the relationshipsbetween the various "equations of motion". Note thatdecades of time are involved in working from left to rightin Figure 1.

4.2.1 Descriptions of "Fluid Particle" Motions

The purpose of solving the various differential equationsof fluid motion is to calculate the flow fields associatedwith the problem at hand. (Most integral methods use anassumed flow field to calculate gross properties of theflow such as lift and drag of a hydrofoil, etc.) In orderto set up the equations of motion of a "fluid particle" werecall from the concept of continuum mechanics (seeShapiro [55]) that a fluid particle represents thesmallest "material" volume liVm which contains so manymolecules as to make statistical averages meaningful, i.e,a density equal to mass per unit volume can be assignedto each particle in a flow field.

To describe the motion of a fluid particle using"material" or "convected" coordinates (a Lagrangiandescription), a given particle of fixed mass (a closedsystem in thermodynamics) is tracked as it deforms whileconvected through space with time (Aris [53], p 83). A"field" description attempts to describe the motionrelative to a set of spatial coordinates in which differentfluid particles occupy a given field point from instant toinstant (a control volume approach in thermodynamics).Newton's laws of motion (1682) are formulated in"material" coordinates, but strictly apply only to themotion of the center of mass. Euler (1759) developedwhat is now called the Euler Equations to describe themotion of continuous media in the absence of shearstresses (inviscid flow of "perfect fluids"). In 1823,

10

Cauchy introduced the concept of the stress tensor whichled to the development of what is now called Cauchy'sEquation oj Motion (1828) which is valid for anycontinuous fluid no matter bow the stress tensor T isrelated to the fluid kinematics. ([53], p 102, [54], P 135)

The basic postulate concerning the linear momentumequation for a material volume (a fixed mass for whichNewton's Second Law is valid) is as follows,

''There exists a frame of reference (an inertial frame)for which, at any instant, the rate of change of linearmomentum of a material volume Vm is equal to theresultant force acting on the mass in that volume."i.e.

IF = F.utface + Fbody = D/Dt {JpV dVm}

Vmwhere

F.utface = f Ton dA = Sum of the surface forces onthe material volume

Fbod = Jpf dVm = Sum of the body forces on they material volume

T = Stress tensor acting on the material volume

f = Body force/unit mass vector acting on thematerial volume (fj are the components)

V = Velocity vector (Vjare the components)

D/Dt = a/at + (Vov) = Material derivative

but because of the equation of continuity forincompressible fluids, divV = 0 [53] so

D/Dt JpV av, = JpDVjDt ev,and by use of the divergence theorem

f Ton dA = I divT ev,Thus the use of the divergence (Green's) theorem allowsthe collection of all the terms into a single integrand asfollows

J {divT + pr- pDV/Dt} dv; = 0

Since the material volume Vm is arbitrary, the integrandmust vanish everywhere in the flow region and oneobtains Cauchy's Equation of Motion

divT + pf - pDV/Dt = 0 (1)

or rearranged in cartesian tensor notation

av./at + v», . = f. + T· J») 1 ).1 ) ').1(2)

Cauchy's Equation of Motion contain 15 dependentvariables plus 4 basic variables. It is also non-linearbecause ofthe convective acceleration terms. Obviously,many additional equations to reduce the number ofdependent variables would be required to solve theseequations directly. As a first step, "constitutiveequations" relating the stress tensor, Tjj, to some functionof the velocity gradient tensor, Vj, j, and thethermodynamic pressure, p, are used. ([53J, p. 107) It iscustomary to separate the velocity gradient tensor intoits symmetric and anti-symmetric components, i.e.

v.. = (v.. + v· .)/2 + (v . - v, .)/2l,t J.t t,) J.t I,)

orVj.i = Di2 + Oi/2

where D.. is defined as the symmetric deformation ratetensor arid or is defined as the anti-symmetric spin orrotation tensor. (Notation is not consistent in thesedefinitions and the factor 1/2 mayor may not beincluded. Sr is frequently used to represent the rate ofstrain (defoimation rate) tensor rather than Dl!" Thisproblem is beginning to be addressed by the Symbolsand Terminology Group of the ITTC. [56]) Notethat the deformation rate tensor, D, is invariant to allobservers (for Galilean transformations), while the spintensor, 0, is not invariant.

For a perfect (ideal) fluid, i.e. the viscosity equals 0, theshear stresses vanish in equation 1, and

divT = -grad p

The resulting Euler Equa:tions of Motion weredeveloped by L. Euler in 1755-56, over 65 years beforethe stress principle of Cauchy (Serrin [54]) and are givenin vector form

pDV/Dt = pf - grad p (3)

The notion of viscous fluidity was developed 90 yearslater by G.G. Stokes in 1845. Stokes stated,

"...that the difference between the pressure on aplane in a given direction passing through anypoint P of a fluid in motion and the pressurewhich would exist in all directions about P if thefluid in its neighborhood were in a state ofrelative equilibrium depends only on the relativemotion of the fluid immediately about P and thatthe relative motion due to any motion of rotationmay be eliminated without affecting thedifference of the pressures above mentioned."(bold mine) [57]

The statement in bold font indicates that Stokespostulated that "material invariance" was an essential

concept when defining material properties of viscousfluids long before Noll [58] formalized the "axiom ofobjectivity" for constitutive equations used in continuummechanics. Serrin [54] defines a large class of real"Stokesian Fluids" as those which have the followingcharacteristics,

1.The stress tensor Tijis a continuous function ofthe deformation tensor Dij and the localthermodynamic state, but independent of otherkinematical quantities.2. The fluid is homogeneous.3. The fluid is isotropic.4. When there is no deformation, the stress ishydrostatic, i.e. Tlj= -poij .

A Newtonian fluid is defined as a linear Stokesian fluid,which for incompressible flow has the followingconstitutive equation (compressible flows are onlyslightly more complex).

Tij= -poij + p.Dij (4)

where p. is the dynamic viscosity.

Combining equations (1) and (4) one obtains theNavier-Stokes Equations for incompressible flow,

pDV/Dt = pf - grad p + div(p.D) (5)

orav./at + vv- . = f. - grad pip + (vDi) i (6)) 1),1) ,

since the kinematic viscosity v = p./p

For laminar flow of an incompressible fluid withconstant kinematic viscosity, v can be factored outsidethe derivative with respect to Xi

pDV/Dt = pf - grad p + p. divD

A limited set of laminar flows in simple geometries haveexact solutions of Eqn (7). For inviscid flow, Eqn (7)reduces to the Euler Equation (3).

4.2.2 Turbulence Modeling

To discuss the "Reynolds-Averaged Navier-Stokesequations" (RANS) using ordinary text mode on a wordprocessor, the overscore on time-averaged quantities(avaiJable only as a macro or in graphics mode forequations) will be replaced by the superscript A (arecent proposal of M. Schmiechen [56]) added to thenotation for Reynolds stress, Rit_used by Lumley [59]in his 1970 paper, "Toward a Turbulent ConstitutiveRelation". Thus we define the time varying velocityvector components as

II

v.(t) = UA + u·) ) )(8)

whereut = time averaged local velocity,

uj = fluctuating velocity, and

RA = -(pU.U.)A = -pU-:-U-:-I) I) I ) (9)

so that

RA = time averaged Reynolds stress tensor whosecomponents are given by Eqn (9)

(The above definition may not outlast this symposiumbecause of possible confusion with the correlationcoefficient RA,Bdefined by Hinze [60] and others, butsomething needs to be done in an age when authors usetheir own personal computers to write papers.)

The derivation of the RANS equations may be found inHinze [60], Tennekes and Lumley [61] and othertexts and will not be repeated here. Both Hinze andTennekes and Lumley begin their derivations withCauchy's equation of motion (Eqn (1), above) and usethe rules for evaluating the effects of time averaging onthe substitution of Eqn (8) in Eqn (6). The resultingReynolds stress term, R/, which. comes from theacceleration terms in Eqn (6), is combined with Eqn (4)to obtain the following constitutive equation

T. = _pAo.. + IIDA + RAI) I)'" I) I)

(10)

or

(7)

This equation in combination with Cauchy's equation ofmotion for the mean flow yields the Reynolds-AveragedNavier-Stokes (RANS) equation without any assumptionsabout eddy viscosity.

pDUA/Dt = prA _ grad pA + div(vDA + RA) (11)

Many papers concerning CFD for viscous flows use theso called "Boussinesq assumption" that the Reynoldsstress can be linearly related to the average deformationrate by using an "apparent" or turbulent eddy viscosity,Vt' i.e.

(12)

so that the implied constitutive equation for turbulentflow having a linear relationship between mean stressand mean deformation rate' becomes

(13)

6

3.1 The Use of CFD in the Aircraft Industry

The valuable and recent survey of "CFD Application toComplex Configurations" was authored by EdwardTinoco [27] of the Boeing Commercial AirplaneCompany. According to Tinoco,

"CFD has joined the wind tunnel and flight testas a principal technology for aerodynamicdesign .... Airplanes are flying today that may nothave been deemed practical if not for the insightprovided by CFD studies. On the basis of CFDanalysis, modifications such as the addition ofnew engines, antennas, or sensor pods have beenincorporated into existing airframes with little orno wind tunnel testing before flight... ..To aid inthe design of such complex configurations, theengineering community needs computationaltools that faithfully represent the dominant flowphysics and the required geometric complexityand are capable of providing reliable solutions ina timely, affordable manner. This is a difficultand as yet unfilled order by any single method.The approach in industry has been to assemble acollection of CFD tools that meet the precedingrequirements to varying degrees." [27]

Tinoco then makes the distinction between what heterms "production tools" which are well documented,validated as to range of applicability, and can be run by"nonexpert" CFD users. Production tools include mostlinear panel methods and many full-potential codes."Expert user" codes generally require complex field gridgeneration, run on a supercomputer, and include codesbased on the Euler and/or Navier-Stokes equations.Again quoting from Tinoco,

'The choice of what level of flow physicsformulation to use on a particular problemrequires the user to have some idea of whatphysics is relevant.. ... Experience has shown thatmethods based on the linear potential and full-potential formulations can be routinely convergedto a higher level of accuracy than has beenpractical with methods based on the Euler orNavier-Stokes formulation." [27]

Another recent paper by Goldhammer and Rubbert[28] of the Boeing Company makes the point thatalthough the fluid mechanics community has always beeninterested in modeling fluid flow problems as accuratelyas possible,

"The connection between physical insight and themathematical formulations is perhaps not asstrong in today's CFD researcher. The verysolution of the basic fluid mechanics equationshas been taken over by mathematicians andcomputer scientists who program supercomputersto solve the equations by numerical

approximation. But the solutions are really onlythe beginning of providing an engineering tool toimprove the technology of airframe design." [28]

This concern was also expressed by Jameson [29] inhis keynote address to the 8th AIAA CFD Conferencein 1987. .

"...the disparity of scales in turbulent flow is solarge that direct simulation is not likely to befeasible without radical developments incomputer technology Progress in simulatingseparated viscous flows may now be moredependent on improvement in turbulencemodeling than it is on algorithm development."[29]

The Advisory Group for Aerospace Research &Development (AGARD) has sponsored numerous paneland working group meetings of specialists in variousareas of CFD. Summaries of two recent meetingsprovide a good state-of-the-art view of critical issues inthe use of CFD in vehicle design. AGARD AdvisoryReport No. 257 [30] is a technical evaluation reportof a 1988 meeting of the Fluid Dynamics Panelconcerning the validation of CFD codes. One of thepanel co-chairmen, R. G. Bradley, made the followingdistinction between "Validation" and "Calibration" whichis applicable to ship hydrodynamics as well,

"Four categories of experiments are required toachieve a mature CFD capability:1. Experiments to understand flow physics2. Experiments to develop physical models3. Experiments to calibrate CFD codes4. Experiments to validate CFD codesBounds of errors, for numerical CFD solutions aswell as for experimental data, have to beevaluated carefully." [30]

Although billed as a CFD validation symposium, theevaluation report by Sacher [30] observed that at least50% of the papers were commercials for the hugevariety of existing'computer codes, i.e., "Software Power".(This problem exists at many conferences on numericalmethods in both aero and hydro dynamics.) Among hisconclusions Sacher observed that,

"Databases for flow code validation have to bespecially designed, carefully performed during thewind tunnel tests, and all boundary conditionshave to be specified in addition ....(in this respect)'External flow people' seem to be betterorganized than 'internal flow engineers'." [30]

This comment was made in reference to the existence ofrelatively adequate experimental databases (formatunspecified) specially tailored for the validation ofvorticity dominated external flow codes. Sacher notedthe lack of experimental databases for validation ofinternal flows (especially at high Reynolds numbers)

where the physics is dominated by large viscous shearlayer r~gions and viscous wakes.

AGARD Advisory Report No. 256 [31] is a State ofthe Art Technical Status Review on Drag Prediction andAnalysis from Computational Fluid Dynamics withforward and conclusions written by J. W. Slooff of theNetherlands. A number of Slooffs conclusions areworth repeating for this symposium:

• "Accurate and consistent computation throughCFD of (absolute) drag levels for complexconfigurations is, not surprisingly, beyond reachfor a considerable time to come [because ofdifficulties with grid generation, turbulencemodeling, grid resolution, speed and economicsof computation].

• "For separated flows, inadequate turbulencemodelling in combination with inappropriate gridclustering and refinement are problem areas evenin 2D airfoil flow where most but not all codescan now predict drag with an accuracy of withinabout 5%. (,But we don't fly airfoils!')

• "Navier-Stokes codes typically do not (yet)involve drag prediction except for 2D airfoilflows. Even then they do not do a better jobthan zonal methods involving potential flow orEuler schemes coupled with boundarylayers....For (Reynolds-averaged) Navier-Stokesmethods the identification and quantification ofthe viscous, induced and wave drag componentsis as yet unclear and might even be impossiblewithout introducing certain assumptions withrespect to the asymptotic structure of the flowfield." [31]

Charles Boppe [32] of Grumman Corporation agreeswith Slooff that consistent and accurate CFD predictionsof absolute drag level for aircraft configurations arecurrently beyond reach. In other papers concerningcomputational aerodynamic design [33], Boppe citesthe successful use of CFD in solving design challengesfor the X-29 forward swept wing demonstrator aircraftand the evolutions in the Gulfstream series corporate jet.Judging from the tentative program, we will hear othersuccess stories concerning the integration of CFD intothe vehicle design process.

3.2 The Use of CFn in Yacht Design

The dramatic loss of the America's Cup to theAustralian challengers in 1983 proved to be a "milestoneevent" in the history of the use of CFD in marine design.The integration of CFD methods in combination withtank testing as discussed by van Oossanen [34] andSlooff [35] radically and permanently changed themethodology for high performance yacht design. Dutchnaval architects and aerodynamicists developed the

7

complimentary use of CFD tools and tank testing todesign the famous "winglets" on Australia II's keel.Slooff and van Oossanen became "media personalities"for a short time and were sporting enough to share theirnew methods [34, 35] with the Americans, who promptlywent to work on their own versions of this Dutchmethodology and proceeded to win the Cup back in 1987[36], [37], [38], [39]. This integrated useof CFD, tank testing and performance analysis codes isnow an essential part of high performance yacht design.(See, for example, Greeley [40].)

3.3 The Use of CFD in Ship Design

The state-of-the-art (as of 1987) concerning the role ofCFD in Ship Design was summarized by Morgan andLin [41] during the Group Discussion on the Impactof Numerical Techniques in Tankery during the 18thITIC. At that time Morgan and Lin concluded,

"In general, one can say that most solutions todaywhich have practical application are solutionsderived from linear theories where viscosity isneglected .....Although more work needs to bedone for blunt bows and transom sterns, thewave resistance of fine form ships can often bepredicted with a degree of accuracy that iscomparable to that measured experimentally."...In all viscous flow calculations at anysignificant Reynolds number it is presentlynecessary to resort to Reynolds averagedequations rather than the full Navier-Stokesequations. As a result, some sort of artifice,generally referred to as turbulence modeling,must be used to approximate the turbulence inthe flow."...The computational tools now available alwaysgive better results to problems where viscouseffects are small. Examples are the prediction ofship motions, wave resistance, and propelleraction." [41]

Various ship motions programs, generally based onlinear strip theory, are regularly used during the shipdesign process. The effect on ship motions of significantchanges in hull form geometry is predicted reasonablywell for moderate sea conditions. One must be careful,however, about how a particular version of a codespecifies the directional characteristics of the seaconditions. The severe rolling problems caused by theaddition of blisters to the U. S. S. Midway (Ricketts[42]) illustrates the need to number and catalog thecapabilities and limitations of various versions of allCFD codes. According to Ricketts,

"It should be noted, however, that theseassessments assumed a theoretical short-crestedseaway with a cosine-squared energy spread anddid not account for frequent real-world situations

where a low-frequency ground swell is present incombination with a local wind-generated sea, thetwo coming from very different directions." [42]

None of the existing programs appear to be sensitive tosmall changes in hull geometry, but seakeeping modeltests are not all that sensitive to small geometry changes,either. Time domain predictions of large amplitudemotions in non-breaking seas as discussed by Lin andYue [43] show promising initial results, but are so farlimited to regular wave examples for relatively simplehull forms. A major development in this method is anautomatic geometry repanelizer to account for changingunderwater geometry as a result of the time dependentfree surface.

The prediction of the motions of small craft in breakingextreme seas is a long way from being solved. Althoughthe use of potential flow theory to describe steep gravitywaves is the subject of validation studies (Dommermuth[44]), the more sophisticated types of ship motionanalysis codes are not adequately validated andtherefore contribute little to the hullform design process.Part of the validation problem is the lack of adequatefull scale ship motions data in measured directional seas(Kjeldsen [45]). I look forward to the futureintegration of directional sea data (with and withoutextreme waves present) and time domain ship motionscodes for the prediction of ship motions in realisticseaways. At the present time we can only speculatewhether or not this problem will be computable (seesection 4.1) at some future date.

On the other hand, wave resistance codes appear to besensitive to small changes in hull form geometry such asbow bulb configurations for high speed ships (Hoyle[46]). These potential-flow codes are gaining acceptanceby the design community, especially as an "order ofmerit" ranking technique for hull form candidates inorder to reduce the scope of model testing. They havebeen adapted to transom stemed ships (Cheng [47]),yacht testing (see previous section) and in partiallyintegrated systems (see FLO WIECH discussion below).Slender-ship codes show promise (Noblesse [48]) inthe prediction of far-field wakes, but presently lack theability to correctly predict sinkage and trim, as near-fieldpotential-flow codes can do, especially if they areiterated using the correct free surface boundarycondition.

As to the impact of CFD on propeller design, Morganand lin [41]:stated,

"The use of propeller design methodology basedon lifting-line and lifting-surface theories hasbecome routine for conventional propellers.These techniques have led to the practicaldevelopment of the highly-skewed propeller for

reducing vibration and cavitation. The design ofa propeller operating in uniform flow using CFDis also a matter of routine. Nevertheless, asthese theoretical techniques still depend on thelift-curve slope being independent of Reynoldsnumber, difficulty remains in using thesetechniques to predict the performance of anexisting propeller with arbitrary camber lines."

Morgan and Lin also discuss the difficulties in predictingthe effects of cavitation and propeller-induced vibrationssince these phenomena depend on viscous flow effects,hull-propeller and wake interactions.

As to the future of CFD developments, Morgan and Linconclude,

"It will be many years before the full Navier-Stokes equations can be solved on a computer atthe ship-scale Reynolds numbers to the detailnecessary the progress toward the solution ofthe Navier-Stokes equations for complexgeometries at high Reynolds numbers has beendisappointingly slow (however) if the CFDprediction does not compare well with theexperiment, there is always the question ofwhether the problem is with the theory or theexperiment, or both." [41]

As an alternative to solving the viscous equations ofmotion around the entire ship hull, Larsson and his co-workers at SSPA/FLOWIECH [26], [49] haveproposed dividing the flow into three zones, a procedurealso used by Hoekstra [50] of MARIN and similar tomethods used by the aircraft industry as discussed at the1987 AIAA CFD Conference [29). Larsson'sSHIPFLOW method defines the following separateregions which have defined interactions within thevarious codes:Zone 1- the potential flow (including free surface)outside the boundary layer,Zone 2- the boundary layer over that portion of the hulllength where boundary layer integral methods may beapplied, i.e., the layer is thin relative to the local radiusof curvature and no separation is present, andZone 3- the stern and wake regions where the flow fieldmust be calculated using a form of the Navier-Stokesequations.

This integrated CFD system has been used in variousforms for more than 30 projects since 1983. SHIPFLOW [51]does not currently predict the correct inflow into thepropeller of ships with significant bilge vortices.However, the iterative methods used to calculate thefree surface wave greatly improve the prediction ofsinkage and trim. The various references toSHIPFLOW do not discuss the CAD interface in anydetail.

4. MATHEMATICAL MODELING IN CLASSICALMECHANICS

4.1 The Limits of Computing

Recently I read a fascinating book entitled TheEmperor's New Mind [52] by Roger Penrose, RouseBall Professor of Mathematics at Oxford University. InPenrose's view,

"...it is our present lack of understanding of thefundamental laws of physics that prevents usfrom coming to grips with the concept of 'mind'in physical or logical terms." (page 4)

The book reviews the present status of our knowledge ofthe existing laws of physics and discusses thecorresponding limits placed on our knowledge of thephysical world and on the future possibilities for artificialintelligence.

In reviewing the basis for recursive and non-recursivemathematics, Penrose makes an important distinctionbetween the deterministic nature of many of the existinglaws of classical (Newtonian) physics and whether or notvarious physical phenomena are algorithmic in a usefulway, i.e.,

"Computability is a different question fromdeterminism." (page 170)"...'determinism' means that initial data at oneparticular time completely fix the behaviour at allother times." (page 214)"There is a sense, however, in which this world is'non-computable' in practice. This arises fromthe fact that the accuracy with which the initialdata can be known is always limited." (page 173)"....no matter how accurately we know the initialstate of a system (within some reasonable limits),the uncertainties will tend to grow in time andour initial information may become almostuseless. Classical mechanics is, in this kind ofsense, essentially unpredictable. (Recall theconcept of 'chaos' considered above.)" (page 183)"The future behavior (of chaotic phenomena)would still be determined, right from the big bang,even though we would be unable to compute it."(page 432) [52]

Penrose then makes the case that true intelligencecannot be present unless accompanied by consciousnesswhich includes essentially non-algorithmic ingredients.Thus true intelligence (including judgement-fonning)cannot be properly simulated by algorithmic means. Infact,

"...the decision as to the validity of an algorithmis not itself an algorithmic process!" (page 414)

This conclusion has important implications for thediscussion on future directions in "intelligent" ship designsystems (see Section 6 below).

9

4.2 The Equations of Motion for CFD

The major focus for solving viscous flow problems inship hydrodynamics appears to center around attemptsto use some version of the Navier-Stokes equationswithout any real effort to ask the question, "which formof the equations of motion contains a reasonable modelof the fluid flow conditions present in the problem athand"? It would appear that most investigations regardthe Navier-Stokes equations as the basic equations ofmotion rather than a very special form for which thenormal and shear stresses are linear functions of thedeformation rate tensor. Let us review the mathematicalbasis for various formulations of Newton's Second Lawas applied to continuum mechanics in general and fluidmechanics in particular. Excellent treatments of thesubject are found in Rutherford Aris' book Vectors,Tensors, and the Basic Equations of Fluid Mechanics[53] and James Serrin's "Mathematical Principles ofClassical Fluid Mechanics" in Handbuch der Physik, Vol.VIII/I. [54] Figure 1, which I developed whileteaching a high level undergraduate fluid mechanicscourse over 20 years ago, illustrates the relationshipsbetween the various "equations of motion". Note thatdecades of time are involved in working from left to rightin Figure 1.

4.2.1 Descriptions of "Fluid Particle" Motions

The purpose of solving the various differential equationsof fluid motion is to calculate the flow fields associatedwith the problem at hand. (Most integral methods use anassumed flow field to calculate gross properties of theflow such as lift and drag of a hydrofoil, etc.) In orderto set up the equations of motion of a "fluid particle" werecall from the concept of continuum mechanics (seeShapiro [55]) that a fluid particle represents thesmallest "material" volume 6Vm which contains so manymolecules as to make statistical averages meaningful, i.e.a density equal to mass per unit volume can be assignedto each particle in a flow field.

To describe the motion of a fluid particle using"material" or "convected" coordinates (a Lagrangiandescription), a given particle of fixed mass (a closedsystem in thermodynamics) is tracked as it deforms whileconvected through space with time (Aris [53], p 83). A"field" description attempts to describe the motionrelative to a set of spatial coordinates in which differentfluid particles occupy a given field point from instant toinstant (a control volume approach in thermodynamics).Newton's laws of motion (1682) are formulated in"material" coordinates, but strictly apply only to themotion of the center of mass .. Euler (1759) developedwhat is now called the Euler Equations to describe themotion of continuous media in the absence of shearstresses (inviscid flow of "perfect fluids"). In 1823,

10

Cauchy introduced the concept of the stress tensor whichled to the development of what is now called Cauchy'sEquation of Motion (1828) which is valid for anycontinuous fluid no matter how the stress tensor T isrelated to the fluid kinematics. ([53], p 102, [54], P 135)

The basic postulate concerning the linear momentumequation for a material volume (a fixed mass for whichNewton's Second Law is valid) is as follows,

"There exists a frame of reference (an inertial frame)for which, at any instant, the rate of change of linearmomentum of a material volume Vm is equal to theresultant force acting on the mass in that volume."i.e.

where

F.urface = f Ten dA = Sum of the surface forces onthe material volume

Fbod = Ipf dVrn = Sum of the body forces on theY material volume

T = Stress tensor acting on the material volume

f = Body force/unit mass vector acting on thematerial volume (~ are the components)

V = Velocity vector (Vj are the components)

D/Dt = a/at + (VeV) = Material derivative

but because of the equation of continuity forincompressible fluids, divV = 0 [53] so

D/Dt IpV av, = IpDV/Dt av,and by use of the divergence theorem

f Ten dA = I divT ev,Thus the use of the divergence (Green's) theorem allowsthe collection of all the terms into a single integrand asfollows

I {divT + pf- pDV/Dt} ev, = 0

Since the material volume Vm is arbitrary, the integrandmust vanish everywhere in the flow region and oneobtains Cauchy's Equation of Motion

divT + pf - pDV/Dt = 0 (1)

or rearranged in cartesian tensor notation

av./at + Vy . = f. + T· ./pJ I J,l J 1J,1(2)

Cauchy's Equation of Motion contain 15 dependentvariables plus 4 basic variables. It is also non-linearbecause of the convective acceleration terms. Obviously,many additional equations to reduce the number ofdependent variables would be required to solve theseequations directly. As a first step, "constitutiveequations" relating the stress tensor, Tij, to some functionof the velocity gradient tensor, Vj, i' and thethermodynamic pressure, p, are used. ([53], p. 107) It iscustomary to separate the velocity gradient tensor intoits symmetric and anti-symmetric components, i.e.

v· . = (v.. + v· .)/2 + (v.. - v· .)/2J,I J,I 1,1 ),1 1,1

orVj, i = Di/2 + nij/2

where D·· is defined as the symmetric deformation ratetensor arid nij is defined as the anti-symmetric spin orrotation tensor. (Notation is not consistent in thesedefinitions and the factor 1/2 mayor may not beincluded. Sij is frequently used to represent the rate ofstrain (deformation rate) tensor rather than Dij• Thisproblem is beginning to be addressed by the Symbolsand Terminology Group of the IITC. [56]) Notethat the deformation rate tensor, D, is invariant to allobservers (for Galilean transformations), while the spintensor, n, is not invariant.

For a perfect (ideal) fluid, i.e, the viscosity equals 0, theshear stresses vanish in equation 1, and

divT = -grad p

The resulting Euler Equations of Motion weredeveloped by L. Euler in 1755-56, over 65 years beforethe stress principle of Cauchy (Serrin [54]) and are givenin vector form

pDV /Dt = pf - grad p (3)

The notion of viscous fluidity was developed 90 yearslater by G.G. Stokes in 1845. Stokes stated,

"...that the difference between the pressure on aplane in a given direction passing through anypoint P of a fluid in motion and the pressurewhich would exist in all directions about P if thefluid in its neighborhood were in a state ofrelative equilibrium depends only on the relativemotion of the fluid immediately about P and thatthe relative motion due to any motion ofrotationmay be eliminated without affecting thedifference of the pressures above mentioned."(bold mine) [57]

The statement in bold font indicates that Stokespostulated that "material invariance" was an essential

concept when defining material properties of viscousfluids long before Noll [58] formalized the "axiom ofobjectivity" for constitutive equations used in continuummechanics. Serrin [54] defines a large class of real"Stokesian Fluids" as those which have the followingcharacteristics,

1. The stress tensor Tij is a continuous function ofthe deformation tensor Dij and the localthermodynamic state, but independent of otherkinematical quantities.2. The fluid is homogeneous.3. The fluid is isotropic.4. When there is no deformation, the stress ishydrostatic, i.e. Tij = -POij .

A Newtonian fluid is defined as a linear Stokesian fluid,which for incompressible flow has the followingconstitutive equation (compressible flows are onlyslightly more complex).

Tij= -POij + /.LDij (4)

where j), is the dynamic viscosity.

Combining equations (1) and (4) one obtains theNavier-Stokes Equations for incompressible flow,

pDV IDt = pC - grad p + div(/.LD) (5)

or8v,18t + vv, ' = f - grad pip + (L1DiJ) i (6)) 1 ),1 ) ,

since the kinematic viscosity LI = /.LIp

For laminar flow of an incompressible fluid withconstant kinematic viscosity, 1/ can be factored outsidethe derivative with respect to Xj

pDV IDt '" pf - grad p + /.L divD

A limited set of laminar flows in simple geometries haveexact solutions of Eqn (7). For inviscid flow, Eqn (7)reduces to the Euler Equation (3).

4.2.2 Turbulence Modeling

To discuss the "Reynolds-Averaged Navier-Stokesequations" (RANS) using ordinary text mode on a wordprocessor, the overscore on time-averaged quantities{available only as a macro or in graphics mode forequations) will be replaced by the superscript A (arecent proposal of M. Schmiechen [56]) added to thenotation for Reynolds stress, Rit., used by Luml~y [~9]in his 1970 paper, "Toward a Turbulent ConstitutiveRelation". Thus we define the time varying velocityvector components as

11

(8)

whereUA = time averaged local velocity,

)

Uj = fluctuating velocity, and

Rt = -(pujUjt = -pu i Uj (9)

so that

RA = time averaged Reynolds stress tensor whosecomponents are given by Eqn (9)

(The above definition may not outlast this symposiumbecause of possible confusion with the correlationcoefficient RA,B defined by Hinze [60] and others, butsomething needs to be done in an age when authors usetheir own personal computers to write papers.)

The derivation of the RANS equations may be found inHinze [60], Tennekes and Lumley [61] and othertexts and will not be repeated here. Both Hinze andTennekes and Lumley begin their derivations withCauchy's equation of motion (Eqn (1), above) and usethe rules for evaluating the effects of time averaging onthe substitution of Eqn (8) in Eqn (6). The resultingReynolds stress term, R/, which, comes from theacceleration terms in Eqn (6), is combined with Eqn (4)to obtain the following constitutive equation

T = _pAoh + IID,A + RA1) 1)'" I) I)

(10)

or

(1)

This equation in combination with Cauchy's equation ofmotion for the mean flow yields the Reynolds-AveragedNavier-Stokes (RANS) equation without any assumptionsabout eddy viscosity.

Many papers concerning CFD for viscous flows use theso called "Boussinesq assumption" that the Reynoldsstress can be linearly related to the average deformationrate by using an "apparent" or turbulent eddy viscosity,LIt, i.e.

(12)

so that the implied constitutive equation for turbulentflow having a linear relationship between mean stressand mean deformation rate 'becomes

(13)

12

which gives the following form for the RANS equationbased on a linear eddy viscosity model

Boussinesq assumed that the eddy viscosity was a scalarquantity, but turbulence measurements have justified thisassumption for very few flows of interest to the designcommunity. Although the normal kinematic viscosity isa material property of the fluid, the eddy viscosity isclearly a function of the nature and local kinematics ofthe turbulent flow.

For complex flows associated with actual vehicles (seefor example, Odabasi [62]), the eddy viscosity variesin space and time so various levels of turbulencemodeling have been established. Standard references onturbulence modeling include Bradshaw [63], Lumley [64]and the many review papers in the three volumeProceedings of the 1980-81 AFOSR-HITM-StanfordConference on Complex Turbulent Flows (Kline[65]). More recent detailed review papers are byLakshminarayana [66], Wilcox [67], and Patel[68). All these reviews support the conclusions ofsection 3 of this paper concerning the lack of adequateturbulence modeling for all but the most simple flows.

In his 1986 review paper, Lakshminarayana discusses thelimitations of seven categories of turbulence modelsfrom the simple but widely used algebraic eddy viscositymodel of Baldwin and Lomax to the various higher ordermodels. The various kinetic energy-dissipation equation(k - E) models assume that the eddy viscosity, "l' isrepresented by

(15)

wherek = (u iU i )/2 = turbulence kinetic energy which

can be measured as well as modeledby transport equations

E turbulent dissipation which cannot bemeasured directly, but is modeled bytransport equations and evaluatedindirectly

C = assumed to be a "scalar constant" for isotropicturbulence, but probably a vector (as aminimum) for non-isotropic turbulence

Lakshrninarayana [66] states,"...the empirical constants used in the turbulencetransport equations, which are based on two-dimensional simple flows, are invalid orinadequate for complex flows... The concept of

'eddy viscosity' is phenomenological and has nomathematical basis."

Lakshminarayana discusses the so called algebraicReynolds stress model (ARSM) and the Reynolds stressmodel (RSM) which are used to solve Eqn (11) directlyat no small cost in complexity. He recommendsinvestigating the use of combination (k- f)/ASRMmodels since rotation and curvature effects are captureddirectly instead of through modeling.

Wilcox [67] advocates replacing the turbulent dissipationrate, E, with what he calls the rate of dissipation ofturbulence per unit energy, w, as a way to incorporatethe effects of surface roughness in the calculations. Healso recommends at least 5 mesh points between y+ =o and 2.5 in the viscous sublayer. In a second paper inthe same AIAA Journal, Wilcox [69] proposes a"Multiscale Model for Turbulent Flows" on the premisethat,

"...turbulence can be described by representingthe turbulence energy spectrum in terms of anupper and a lower partition with the upperpartition corresponding to lowest frequency,energy-bearing eddies. This notion is used inlarge-eddy simulation work where small eddies(corresponding to the upper portion of thespectrum) are numerically simulated. Bycontrast, we model both the small and the largeeddies. We use the general observations thateddies in the lower partition are expected tocontain most of the vorticity, to be isotropic, andto dissipate rapidly into heat. A key feature ofeddies in the upper partition is that they aremore or less inviscid." [69]

Even more sophistication has been proposed by Orzag [70]using Renormalization Group Theory (RNG) which isclaimed to avoid the difficulties of algebraic models thatcontain singularities which diverge near wall regions andseparation zones.

The more complex formulations may model theturbulent flow field more accurately, but thecombination of a complex turbulent model applied tocomplex geometries is cost prohibitive and probablycannot be validated because of the non-algorithmicconsiderations discussed in section 4.1.

There have been several attempts at matching carefullycontrolled experiments with precise CFD analyis inorder to "validate" various turbulence models. Asignificant problem in building an experimental databasefor CFD validations is the difficulty in makingturbulence measurements at sufficiently high Reynoldsnumbers. This is because it is difficult to obtain reliable

readings in thin boundary layers and hot wire/film andlaser doppler velocimeter (LDV) measurements aresubject to considerable uncertainty at y+ < 10 in theimportant viscous sub-layer. To overcome theseproblems, measurements have been made in oil-filledchannels by Ecklemann [71] and his colleagues atGottingen,

Perhaps the best known attempts at validation are thestudies at NASA Ames Research Center, such as thoseby Kim, Moin and Moser [72], in which the unsteadyNavier-Stokes equations were solved numerically at aReynolds number of 3300. The computation was carriedout with 192 x 129 x 160 grid points. 250 hours ofCRAY-XMP CPU time were used for the timedependent calculations. Although the calculated energyspectra and two point correlations indicated adequategrid resolution,

"..the drop-off of the computed spectra of highwave numbers is not sufficient evidence that thecomputed results are unaffected by the smallscale motions neglected in the computations. Itis not clear what significant dynamical roles, ifany, these small scales would play if included inthe computations. Numerical experiments withmuch finer resolutions than those used herewould presumably clarify this issue. Suchcomputations are very difficult and timeconsuming to carry out with the presentcomputers ....detailed comparison in the wallregion reveals consistent discrepancies. Inparticular, the computed Reynolds stresses areconsistently lower than the measured values,while the computed vorticity fluctuations at thewall are higher than the experimentalvalues ....One source of the discrepancy might berelated to the measurement of the wall-shearvelocity ur. When the mean-velocity profiles arerenormalized with the corrected (experimental)wall shear results, excellent agreement among theexperimental results and the computed results isobtained ...(however) the computed turbulenceintensities, except the streamwise fluctuations,remain lower than measured values ....it isimportant to resolve the differences if the use ofthe computer-generated databases orexperimental data in studying turbulencestructures and in developing improved turbulencemodels is to be continued." [72]

The NASA group [73] then used the database of [72]to compute the distribution of the turbulence kineticenergy k and the dissipation rate f. The steps involvedin validating the k - e model are well illustrated in thispaper.

13

Precise validation experiments using simultaneous two-component laser velocimeter measurements were madein the incompressible turbulent flowfield following anaxisymmetric sudden expansion, which involves separatedflows. (Gould [74]) For this moderately complexflow, the measured turbulent normal stresses weresignificantly different from the calculated ones. Gouldconcluded,

"A major shortcoming of the k - f model and, inparticular, the modified Boussinesqapproximation is that it does not correctly modelthe anisotropy of the flow...One must rememberthat the k - E model does not directly predictturbulent stresses; it solves two additionalconservation equations, one for turbulent kineticenergy and one for turbulent dissipation rate, inorder to define an eddy viscosity. Knowing theeddy viscosity allows one to estimate theunknown turbulent stresses in each of theconservation of momentum equations andtherefore close the problem." [74]

4.2.2.1 The Question of "Material Indifference"

Although Lumley [59] acknowledged the importance ofthe concept of material indifference for constitutiveequations involving material properties, he continues todispute the contention of Speziale [75], [76] thatone should seek invariant forms of the constitutiveequations for turbulence because,

"The Reynolds stress transport equations are notform invariant under time-dependent rotations ofthe spacial frame of reference while the Reynoldsstresses have been proven to be invariant underthis group of transformations." [75]

Lumley responds,"I believe it is clear [59] that requmng aconstitutive relation to satisfy the principle ofmaterial indifference is equivalent to ignoring theinertia of any motions responsible for thedevelopment of the stress, since the relation isrequired to be the same in inertial andnoninertial frames. Since all matter has inertiathis means that the principle must be anapproximation in any situation in which.unresolved motions contribute to the stress ....Themotions responsible for the development of theReynolds stress are not microscopic (as in thecase of ordinary viscosity), but of quite finiteamplitude, and definitely subject to inertia." [64]

In his 1970 paper, Lumley [59] discussed the possiblesignificance of the ratio of the spin tensor to the strainrate tensor n/S (n/D) in establishing the form of theconstitutive equations for turbulence. One possible

14

Reynolds stress formulation was a function of a whichwas used as an argument to reject material indifferencerequirements for turbulent constitutive equations.

During the 1981 conference on Complex TurbulentFlows, Peter Bradshaw made the following observation

"However, Launder has successfully added non- 'invariant terms to his dissipation transportequation in order to improve predictions for thinshear layer .....Therefore, non-invariant allowancesfor the mean extra strain rates of interest shouldnot be regarded as illegal, immoral or eveninconvenient." (Kline [65], Vol. 2, p 707)

In his more recent papers on nonlinear k - e models forturbulent closure, Speziale [77] continues to insist onthe use of invariant quantities for constitutive equations,but finds that the a/S ratio may be significant inpredicting the equilibrium states of the turbulent kineticenergy, k, and the dissipation rate, €.

I believe it is time to revisit Lumley's 1970 paper [59]and recast the discussion of constitutive equations fornu~erical ~al~sis at the macro scales of typicalattamable gnd sizes and turbulent vorticity scales ratherthan at the fluid particle level as is the case for thee~s~i~g math~~atical models. It is. very likely thatSIgnificantvorttcity (a > 0) is present within many of thesmallest practical grid sizes in the high shear regionsnear the wall for both thin and thick boundary layers.This could partially explain why computed Reynoldsstresses are consistently lower than measured values.

A frequently overlooked requirement when setting meshsizes in numerical hydrodynamics is that the fluid isconsidered to be non-polar, i.e. the fluid is such that thetorques within a "fluid particle" arise only as themoments of direct forces. The assumption is then madethat angular momentum is conserved at the fluid particlelevel thus making the stress tensor symmetric. ([53] p102) ,

Since extremely fine mesh sizes are required in highshear layers to prevent "numerical vorticity", thenecessary grid sizes for complex flows become too largeto be "algorithmic in a useful way". A possibleresolution of this dilemma would be to modify the~urbulence. constitutive equations (10) and (13) byincorporating some function of the local spin tensor todeformation rate tensor (a/D) ratio determined at thelocal mesh size. For closure this would require addinga transport model for vorticity distribution as a functionof the shear gradients present in the flow. Althoughsuch a scheme may only substitute another apriori

problem for the existing ones, hopefully one would beable to use an optimization scheme to trade off grid sizeagainst vorticity distribution complexity. This suggestionmay be related to some of the methods involved in theviscous-inviscid interaction method described by Haaseand Seibert [78], but time did not permit me toinvestigate the latter method.

5. BOTI'LENECKS FOR INTEGRATING CFD ANDCAD

Recent interviews with CFD oriented vehicle designersreveal that there are still many "bottlenecks" in thedesign process. On the CFD side these includesituations involving

5.1 Interfaces between CFD and CAD SurfaceGeometry Codes

Common to nearly all CFD/CAD systems is the timeconsuming and frequently labor intensive delayassociated with repanelization and grid generation. Thisis required to accommodate changes in surface geometryand for situations where dynamic flow conditions requiretime domain solutions. It is an area where artificialintelligence/expert system developments can providemore cost effective design solutions. It is also an areawhere industry-wide standards on formats for thedescription of surface geometry would make innovativecode developments transportable from one CFD/CADsystem to another.

Difficult geometries, such as intersections with complexvehicle appendages and the underside of an automobile(see for example, Haase [78] and Han [79]), mayrequire nearly as much time for grid generation as wouldbe needed for a wind tunnel or ship model test. I wasinvolved with a flow code validation study (Hoyle [46])in which, for a short period, the model test program hadto wait on the CFD code grid generation process.

5.2 Code Validation Bottlenecks

As discussed in section 2.2, nearly all exchanges ofhydrodynamic data are accomplished with "hardcopy"data exchanges. To insure consistency in comparativevalidation studies, hullform geometry and hydrodynamicperformance data should be exchanged in standardizedformats (see section 6.1). Standard format dataexchanges could significantly reduce the time and cost ofCFD code validations.

I would like to suggest that the use of "RANS" as amarketing "buzzword" for unvalidated viscous flow codesbe discouraged unless sufficient qualifying assumptionsare stated in the same paragraph. A step in the right

direction can be found in the Summary of the GroupDiscussion on Navier-Stokes Solvers on pages 721 to 724of the Proceedings of the 5th International Conferenceon Numerical Ship Hydrodynamics in 1989 [47-50].

5.3 Improved Hull Optimization Algorithms

Combined Wave Resistance, Viscous Resistance andSeakeeping Code Systems are being developed ([25],[26]) but are not yet readily available to the ship designcommunity outside the group which wrote the codes.Cost models for the impacts of seakeeping on shipoperating economics (Sellars [80]) have been J?utforward, but not integrated into many hull form des~gnsystems. Simple fuel cost models based o~ operatingprofiles should be avail~ble from a menu ~nven ~ystemwhen making comparauve hullform analysis studies.

Finally, algorithmic loads, slamming and deck wetnesscodes plus seaworthiness and operability crit~ria are stillthe subject of considerable de~ate. R~so~utlOn.o~.the~eproblems will require more rational cntena definition III

order to be incorporated into integrated CFD/CADsystems.

6. FUTURE DIRECTIONS

6.1 ITIC Standard Formats For Neutral DataExchange

As discussed in Johnson [2], the amount of informationto be stored in hydrodynamic hullform design databasetables will be -very large, especially if CFD validationdata is to be shared between various hullform designsupport organizations. The number of possible attributes(columns) and individual tests (rows) in these tablesnumbers in the thousands. This is a situation where thedevelopment of neutral standard formats could makedata exchanges between multiple organizations moreeconomically feasible. Responding to the need for dataexchange standards expressed by various organizationsassociated with the International Towing TankConference, the ITTC Symbols and Terminology Grouphas proposed to undertake the following tasks during the1990 to 1993 period:

1. To develop a standard neutral format for theefficient exchange of hydrodynamic performance datadefined in the ITTC Standard Symbols andTerminology List. [56]

2. To utilize the format specifications being developedby ISO/STEP (STandard for the Exchange ofProduct model data) as the basis for an interimstandard neutral format for the efficient exchange of

15

data concerned with the definition of hullform,propeller and appendage geometry.

The development of mc standard formats for theexchange of hull form design and attribute data ~lrequire interactions with organizations normally outsidethe ship design community. This is especially true forthe development of standard formats for the efficientexchange of data concerned with the CFD codevalidation and the definition of hullform, propeller andappendage geometries.

In this respect, interaction with the parallel developmentofIGES/PDES/ISO STEP standards including those forthe Computer Graphics Metafile (CGM) concept(Mumford [81], Owen [82], Brandli [83]) andnew database technologies (Yamamoto [84], Ohsuga[85]) may prove fruitful to all parties concerned. Obj~ct-oriented DBMS which would appear to be a possiblesolution for the exchange of surface definitioninformation, as yet have no equivalent to the relationaldatabase oriented ANSI standard SQL (StructuredQuery Language) which is quite adequate for storing andretrieving single valued quantities. There is also thechallenge of developing standards for the exchange ofrecords and analysis procedures for time series datawhich includes seakeeping, maneuvering and turbulencemodeling data.

The 1987 Draft Standard Symbols and Terminology List(including computer compatible symbols) has beenupdated for the 1990 conference. Although the list isnot yet complete, the computer compatible sy~bolscontained in the 1990 draft can be used as the basis fordefining standard attribute names for a neutral controlfile format.

6.2 Intelligent Ship Design Systems

A fascinating view of future CAD systems is containedin the paper "Towards Intelligent CAD Systems" by Dr.Setsuo Ohsuga of the Research Center for AdvancedScience and Technology at the University of Tokyo(Ohsuga [85]). Dr. Ohsuga points out that .,.

"...current applications of AI (artificialintelligence) are limited to rather small-scaleproblems, and are not proving a powerful meansuseful for innovative design."

What is needed is a new generation of informationprocessing systems for "knowledge-proc7ssing"technology. An intelligent CAD system, he continues,

"...must be able to represent explicitly an objectand be able to manipulate it. The representationof an object is called an 'object model', and thesystem, a 'model based problem-solving system"

16

In addition to representing the 'object model', theintelligent CAD system must also contain a model of theDesign Process (Figure 2, adapted from [85]) which cananalyse and evaluate the object model from multipleviewpoints in relation to the given requirements.Although numerical codes for many aspects of the shipdesign process are well developed, the design processitself is not yet algorithmic, and may never be if Penrose[52] is correct. (In this sense, I prefer Figure 2 to theclassic "design spiral" which could be interpreted asimplying that the design process is algorithmic.) It isstill up to the individual designer to come up with thebest design to satisfy the goals and constraints of theproblem.

Thus Ohsuga states that the goals of the intelligent CADsystem are:

"To computerize as many individual operations(shown in Figure 2) involved in a design processas possible andTo integrate these different operations into aprocess. As a matter of course, a human designermust be able to intervene in the process at anytime." (85)

Engineering design requires not only quantitativecomputations but also qualitative ones; innovation,creativity, exploration of new non-standard designs, andreasoning about goals, requirements and constraints.Penrose [52) and others outside the "strong AI"community regard many of these qualitive activities asnon-deterministic and certainly not algorithmic in a usefulway. Artificial intelligence, although it can not offer acomplete solution to the problem, can provide valuableassistance and methods for representation of designs,domain knowledge acquisition, and reasoning aboutgoals and constraints. Expert systems have successfullycaptured the knowledge and problem solving techniquesused for CFD (Andrews [86]) and during the designprocess (Tong [87]). The ship design process hasbeen outlined by naval architects (see for example,Schaffer (88)), and expert systems have been used tosupport portions of the ship design process (MacCallum(89), Duffy [90), van Oortmerssen (18), Koops[20], and several papers at this symposium). ArtificialIntelligence methods will eventually be used to provideassistance to the designer in the generation of possibledesign solutions and to control the algorithmic aspects ofthe design process. Based on the opinions of Penrose[52) and Ohsuga [85), AI is unlikely to replace thecreative (conscious) input of human designers.

7. CONCLUSIONS

The foregoing paper has touched upon many aspects ofCFD and CAD codes used for hull form design and the

developments needed to integrate these tools into a costeffective ship design system. The conclusions can besummarized as follows:

1. Previous attempts, such as the IPAD effort, todevelop fully integrated vehicle design systems builtaround the efficient use of a central databaseexceeded the capabilities of the hardware andsoftware commercially available at that time. TheWorks tation/Mainframe /RD BMS /VNIXenvironment is now sufficiently developed toaccomplish this task at a reasonable cost.

2. The work on international CAD/CAM/CIM dataexchange standards by projects such as theISO/STEP, NIDDESC, and CALS should beextended to include other CAE/CFD and model testdata frequently used during the concept designphase. The IITC community should join this effortto encourage international cooperation in ship designand to provide even more cost savings during shipand marine vehicle design.

3. A ship design equivalent to AGARD, possibly aninternational consortium, is needed to bridge the gapbetween basic CFD research efforts and the presentdesign process. There is a great need to make thosewho are in a position to fund integration work awareof the long term cost benefits which could result byreducing the bottlenecks in the concept designprocess discussed in section 5. Cost sharing betweenthe parties involved in "open system" code and codeinterface development is a viable alternative to thedevelopment of proprietary CAE/CAD systemswhich mayor may not find a niche in the ship designmarket.

8. ACKNOWLEDGEMENTS

Much of the work described in this paper was carriedout under the sponsorship of the Defense AdvancedResearch Projects Agency, Machine Intelligence Branch.

9. REFERENCES

1. Johnson, B., and Saylor, R., "An Attempt toCharacterize the 'Turbulence Burst Phenomena'using Digital Time Series Analysis", Proceedings ofthe Symposium on Turbulence in Liquids, Univ, ofMissouri-Rolla, Oct. 1971.

2. Johnson, B., Beaver, W., Fitzsimmons, P., McCallum,D., Glinos, N., and Anderson, N., "Database Systemsfor Hull Form Design", Transactions, SNAME, 1990.

3. Boylston, J. W., deKoff, D. J., and Muntjwerf, J. J.,"SL-7 Containerships: Design, Construction, andOperational Experience", Transactions, SNAME,1974.

4. Kloetzli, J., and Billingsley, D., "NIDDESC: Meetingthe Data Exchange Challenge Through aCooperative Effort", SNAME NSRP 1989 ShipProduction Symposium, Arlington, VA, Sept. 1989.

5. Streiff, M. A, and Cada, D. G., "Applications ofDigital Transfer of Computer-Aided Design Data forProduction Usage", Jour. of Ship Production, Vol. 5,No.1, Feb. 1989, pp 16-21.

6. Port, 0., Schiller, Z., and King, R W., "A SmarterWay to Manufacture",Business Week, No. 3157, April30, 1990, pp 110-117.

7. Fulton, R E., and Salley, G. C., "IPAD: A UniqueApproach to Government/Industry Cooperation forTechnology Development and Transfer", NASA-TM-86422, June 1985.

8. Miller, R E., "IPAD-Integrated Programs forAerospace Vehicle Design", NASA ContractorReport 3890, May 1985.

9. Blackburn, C. L., Storaasli, O. 0.,and Fulton, R E.,"The Role and Application of Data BaseManagement in Integrated Computer-Aided Design",Jour. Aircraft, Vol. 20, No.8, August 1983, pp717-72?

10. Wilson, P. R, "A Short History of CAD DataTransfer Standards", IEEE Computer Graphicsand Applications, Vol. 7, June 1987, pp 64-67.

11. Dube, R P. and Smith, M. R, "ManagingGeometric Information with a DatabaseManagement System", Special Edition of IEEEComputer Graphics and Applications, Oct. 1983, pp57-62.

12. Taggert, R, ed. Ship Design and Construction,SNAME, New York, 1980.

13. Johnson, B., Glinos, N., and Anderson, N., "AProgress Report on a Proposed IITC 'OpenArchitecture' Standard Format for HullformDatabases", Proceedings of the 22nd AmericanTowing Tank Conference, St. Johns, Newfound-land, Aug. 1989.

14. Koops, A, "Hull Form Definition and Computer-Aided Design", Proceedings, 5th Int. Conf. onComputer Applications in the Automation of

17

Shipyard Operation and Ship Design (ICCAS),Trieste, Sept. 1985.

15. Oomen, AC.W.J., and van Oossanen, P.,"Development of a New Computer-Aided Systemfor the Conceptual Design of Ships", 6th WEMTSymposium, Lubeek-Travemunde, June 1987.

16. van Oossanen, P., "Preliminary Ship Design Usingthe HOSDES CAD System", 7th WEMTSymposium, July 1988.

17. Koops, A, et aI, "HOSDES: A New Computer-Aided System for the Conceptual Design of Ships",Bicentennial Symposium, Sydney, Australia, Jan.1988.

18. van Oortmerssen, G., and van Oossanen, P., "ANew CAD System for the Design of Ships",Proceedings of CADMO 88, the 2nd Int. Conf. onComputer-Aided Design, Manufacture andOperations in the Marine and Offshore Industries,Southhampton, Sept. 1988.

19. Glijnis, B., "HOSDES/MARDES: A New Conceptfor CAD Systems", Schip en Werf, Vol. 56, June1989, pp 196-202.

20. Koops, A, Oomen, AC.W.J., and Bras, B.A,"Computer Based Design Assistance and DecisionSupport", Proceedings, 6th Int. Conf. on ComputerApplications in the Automation of ShipyardOperation and Ship Design (ICCAS), Shanghai,Sept. 1988.

21. Parker, M.N., Odabasi, A Y., Fitzsimmons, P. A,and Goggin, C.J., "Advanced Technology in ShipDesign Analysis and Production", Naval EngineersJournal, May 1984.

22. Odabasi, A Y., Patterson. D. R., Williams, E. A,and Davison, G. H., "Britships-Application ofAdvanced Technology to Shipbuilding CADCAM",6th WEMT Symposium, Lubeck-Travemunde, June1987.

23. Archer, D. J., and Marshall, G., "PreliminaryShip Design: A Rational Approach UsingMicrocomputers", Proceedings of CADMO 88, the2nd Int. Conf. on Computer-Aided Design,Manufacture and Operations in the Marine andOffshore Industries, Southhampton, Sept. 1988.

24. Ames, R M., and Lynaugh, K. M., "A Review ofHullform Design Systems for the Marine Industry",Proceedings of CADMO 88, the 2nd Int. Conf. onComputer-Aided Design, Manufacture and

18

Operations in the Marine and Offshore Industries,Southhampton, Sept. 1988.

25. Fritts, M., Comstock, E., Lin, W.c., and Salvesen,N., "Hydro-Numeric Design: PerformancePrediction and Impact on Hull Design",Transactions, SNAME, 1990.

26. Larsson, L., Broberg, L., Kim, K, and Zhang, D.,"A Method For Resistance and Flow Prediction inShip Design", Transactions, SNAME, 1990.

27. Tinoco, E. N., "CFD Application to ComplexConfigurations", Chapter 15 of AppliedComputational Aerodynamics, AlAA Progress inAeronautics and Astronautics, Vol. 125, 1990.

28. Goldhammer, M. I., and Rubbert, P. E., "CFD inDesign: An Airframe Perspective", AlAA paper89-0092, 1989.

29. Jameson, A, "Successes and Challenges inComputational Aerodynamics", Papers of the AlAA8th Computational Fluid Dynamics Conference,Hawaii, June 9-11, 1987, pp 1-35.

30. Sacher, P.W., "Technical Evaluation Report on theFluid Dynamics Panel Symposium on Validation ofComputational Fluid Dynamics", AGARD AdvisoryReport No. 257, 1989.

31. Slooff, J. W., "Technical Status Review on DragPrediction and Analysis from Computational FluidDynamics: State of the Art", AGARD AdvisoryReport No. 256, 1988.

32. Boppe, C. W., "CFD Drag Prediction forAerodynamic Design", AGARD Advisory ReportNo. 256, 1988.

33. Boppe, C. W., "Computational AerodynamicDesign: X-29, the Gulfstream Series and a TacticalFighter", SAE Technical Paper Series 851789,1985.

34. van Oossanen, P., "The Development of the 12Meter Class Yacht Australia /I", Proceedings, 7thChesapeake Sailing Yacht Symposium, Annapolis,MD,1985.

35. Slooff, J. W., "On Wings and Keels", AlAA 12thAnnual Symposium on Sailing (The AncientInterface XlI), Seattle, WA, Sept. 1985.

36. Scragg, C. A, et aI, "The Analysis of WaveResistance in the Design of the Twelve MeterYacht Stars & Stripes", 8th Chesapeake SailingYacht Symposium, SNAME, March 1987. AlsoMARINE TECHNOLOGY, SNAME, Oct. 1987.

37. Boppe, C. W. et aI, "Stars & Stripes '87;Computational Flow Simulations for HydrodynamicDesign", 8th Chesapeake Sailing Yacht Symposium,Annapolis, MD, March 1987.

38. Oliver, J. C., Letcher, 1. S., and Salvesen, N.,"Performance Predictions for Stars & Stripes",Transactions, SNAME, 1987.

39. Letcher, 1.S. et al, "Hydro-Numeric Design ofWinglet Keels for Stars & Stripes", MARINETECHNOLOGY, Vol. 24, No.4, Oct. 1987.

40. Greeley, D. S., and Cross-Whiter, J. H., "Designand Hydrodynamic Performance of Sailboat Keels",MARINE TECHNOLOGY, SNAME, October,1989.

41. Morgan, W. B., and Lin, W.-c., "ComputationalFluid Dynamics, Ship Design and ModelEvaluation", Proceedings of the 18th IITC, Vol. 2,1987.

42. Rickets, M. V., and Gale, P. A, "On Motions,Wetness, and Such: The U. S. S. Midway BlisterStory", Transactions, SNAME, 1989.

43. Lin, W.-M., and Yue, D.KP., "Numerical Solutionsfor Large-Amplitude Ship Motions in the TimeDomain", Proceedings, 18th Symposium on NavalHydrodynamics, Ann Arbor, Aug. 1990.

44. Dommermuth, D. G., et aI, "Deep-water PlungingBreakers: a Comparison between Potential Theoryand Experiments", Jour. Fluid Mech., Vol. 189,1988, pp 423-442.

45. Kjeldsen, S. P., "The Practical Value of DirectionalOcean Wave Spectra", Technical Digest, Vol. 11,No.2, Johns Hopkins Univ. Applied Pbysics Lab.,1990.

46. Hoyle, J. W. et al, "A Bow Bulb DesignMethodology for High Speed Ships", Transactions,SNAME, Vol. 94, 1986.

47. Cheng, B. H., "Computations of #D Transom SternFlows", Proceedings, 5th Int. Conf. on NumericalShip Hydrodynamics, Hiroshima, Sept. 1989.

48. Noblesse, F., Lin, W.-M., and Mellish, R.,"Numerical Evaluation of a Ship's Steady WaveSpectrum", Proceedings, 5th Int. Conf. on NumericalShip Hydrodynamics, Hiroshima, Sept. 1989.

49. Larsson, L. et aI, "New Viscous and Inviscid CFDTechniques for Ship Flows", Proceedings of the 5th

Int. Conf. on Numerical Ship Hydrodynamics,Hiroshima, Sept. 1989.

50. Hoekstra, M., "Recent Developments in a ShipStern Flow Prediction Code", Proceedings of the5th Int. Conf. on Numerical Ship Hydrodynamics,Hiroshima, Sept. 1989.

51. Larsson, L. et al, "SHIPFLOW - A CFD System forShip Design", 4th International Symposium onPractical Design of Ships and Mobile Units, Varna,1989.

52. Penrose, R, The Emperor's New Mind, OxfordUniversity Press, New York, 1989.

53. Aris, R, Vectors, Tensors, and the Basic Equationsof Fluid Mechanics, Prentice-Hall, EnglewoodCliffs, NJ, 1962.

54. Serrin, J., "Mathematical Principles of ClassicalFluid Mechanics", in Handbuch der Physik, Vol.VIII/I, Springer-Verlag, Berlin, 1959, pp 125-263.

55. Shapiro, A. H., The Dynamics and Thermodynamicsof Compressible Fluid Flow, Vol. 1, Ronald Press,New York, 1953.

56. International Towing Tank Conference, StandardSymbols and Terminology List, Draft, 1990.

57. Stokes, G. G., "On the Theories of the InternalFriction of Fluids in Motion and the Equilibriumand Motion of Elastic Solids", Trans. CambridgePhil. Society, Vol. 8, 1844-1849, pp 287-319.

58. Noll, W., "A Mathematical Theory of theMechanical Behaviour of Continuous Media",Arch. Rational Mech. and Analysis, Vol. 2, 1958,pp 197-226.

59. Lumley, J. L., "Toward a Turbulent ConstitutiveRelation", Jour. Fluid Mechanics, Vol. 41, 1970,pp 413-434.

60. Hinze, J. 0., Turbulence, Second Edition,McGraw-Hill, New York, 1975.

61. Tennekes, H., and Lumley, J. L., A First Course inTurbulence, MIT Press, Cambridge, MA 1972.

62. Odabasi, A. Y., and Davies, M. E., "Structure ofthe Turbulent Shear Flow in Ship BoundaryLayers", 2nd Symposium on Numerical and PhysicalAspects of Aerodynamic Flows", Calif. State Univ.,Jan. 1983.

19

63. Bradshaw, P., Cebeci, T., and Whitelaw, J. H.,Engineering Calculation Methods for Turbulent Flow,Academic Press, London, 1981.

64. Lumley, J. L., "Turbulence Modeling", JournalApplied Mechanics, Vol. 50, Dec. 1983, pp 1097-1103.

65. Kline, S. J., Cantwell, B. J., and Lilley, G. M.,Proceedings, The 1980-81 AFOSR-HTTM-StanfordConference on Complex Turbulent Flows, 3Volumes, Stanford University, 1982.

66. Lakshminarayana, B., ''Turbulence Modeling forComplex Shear Flows", AlAA Journal; Vol. 24,No. 12, Dec. 1986, pp 1900-1917.

67. Wilcox, D. C., "Reassessment of the Scale-Determining Equation for Advanced TurbulenceModels", AlAA Journal, Vol. 26, No. 11, Nov.1988.

68. Patel, V. C., Rodi, W., and Scheuerer, G.,"Turbulence Models for Near-Wall and LowReynolds Number Flows: A Review", AlAAJournal, Vol. 23, No.9, Sept. 1985.

69. Wilcox, D. c., "Multiscale Model for TurbulentFlows", AlAA Journal, Vol. 26, No. 11, Nov. 1988.

70. Orzag, S. A. et al, "RNG Modeling Techniques forComplex Turbulent Flows", Proceedings of the 5thInt. Conf. on Numerical Ship Hydrodynamics,Hiroshima, Sept. 1989.

71. Eckelmann, H., "The Structure of the ViscousSublayer and the Adjacent Wall Region in aTurbulent Channel Flow", Jour. Fluid Mech., Vol.65, 1974.

72. Kim, J., Moin, P., and Moser, R, "TurbulentStatistics in Fully Developed Channel Flow at LowReynolds Number", Jour. Fluid Mech., Vol. 177,1987, pp 133-166.

73. Mansour, N. N., Kim, J., and Moin, P., "Near-Wallk - f Turbulence Modeling", AlAA Journal, Vol.27, No.8, Aug. 1989.

74. Gould, R D., Stevenson, W. H., and Thompson, H.D., "Investigation of Turbulent Transport in anAxisymmetric Sudden Expansion", AlAA Journal,Vol. 28, No.2, Feb. 1990.

75. Speziale, C. G., "Invariance of Turbulent ClosureModels", Physics of Fluids, Vol. 22, 1979, pp 1033.1037.

20

76. Speziale, C. G., "Closure Relations for thePressure-Strain Correlation of Turbulence", Physicsof Fluids, Vol. 23, March 1980, pp 459-463.

77. Speziale, C. G., and Mhuiris, N. M. G., "On thePrediction of Equilibrjum States in HomogeneousTurbulence", Jour. Fluid Mech., Vol. 209, 1989, pp591-615.

78. Haase, W., and Seibert, W., "ComputationalAerodynamics for Passenger Cars", FluidsEngineering Division paper, ASME AnnualMeeting, Boston, Nov. 1983.

79. Han, T., "Computational Analysis of Three-Dimensional Turbulent Flow Around a Bluff Bodyin Ground Proximity", AIAA Journal, Vol. 27, No.9, Sept. 1989, pp 1213-1219.

80. Sellars, F., and Setterstrom, C., "Impacts ofSeakeeping on Ship Operating Economics", paperNo. 22, Int. Symp. on Ship Operations,Management and Economics, Kings Point, Sept.1987.

81. Mumford, A. M., "Why care about the ComputerGraphics Metafile?", Computer-Aided Design, Vol.19, No.8 Oct. 1987.

82. Owen, J., and Bloor, M. S., "Neutral Formats forProduct Data Exchange: the Current Situation",Computer-Aided Design, Vol. 19, No.8. Oct. 1987.

83. Brandli, N., and Mittelstaedt, M., "Exchange ofSolid Models: Current State and Future Trends",Computer-Aided Design, Vol. 21, No.2, March1989.

84. Yamamoto, Y., and Murahashi, Y., "CommonLanguage for Multilateral Communication BetweenDifferent CAD CAM Drawing Databases",Computer-Aided Design, Vol. 21, No. 10, Dec. 1989.

85. Ohsuga, S., ''Towards Intelligent CAD Systems",Computer-Aided Design, Vol. 21, No.5, June 1989.

86. Andrews, A. E., "Progress and Challenges in theApplication of Artificial Intelligence toComputational Fluid Dynamics", AIAA Journal,Vol. 26, No.1, Jan. 1988.

87. Tong, S. S., "Coupling Symbolic Manipulation andNumerical Simulation for Complex EngineeringDesigns", Proceedings, Conference on ExpertSystems for Numerical Computing, Int. Assoc. ofMathematics and Computers in Simulation, PurdueUniversity, Dec. 1988.

88. Schaffer, R, Byers, D., and Slager, J., "Towards anImproved Hull Form Design Methodology", NavalEngineers Journal, May, 1983.

89. MacCallum, K 1., "Creative Ship Design byComputer", 4th International Conference onComputer Applications in the Automation ofShipyard Operation and Ship Design (ICCAS),Annapolis, MD 1982.

90. Duffy, A. H. B., and MacCallum, K. J., "ComputerRepresentation of Numerical Expertise forPreliminary Ship Design", Marine Technology, Vol.26, No.4, Oct. 1989.

General Postulate for a Material Volume

Differential Equation [Flow - Field) AnalysisI - ... ------,- I

Only normal surface Surface stressesstresses (pressure) include viscous

are considered (shear) stresses

Non - ViS~OUS Flow ViSCOUf Flow

Apply Continuity Equation andDivergence (Green -Gauss) Theorem

Integral Analysis

Apply Reynold'sTransport Theorem

Integral MomentumEquation for aControl Volume

Euler'sEquation of Motion

Cauchy'sEquation of Motion

Consider flow alonga given streamline,

Postulate a linearrelationship between

stress and deformationrate (constitutive eqn.

for Newtonian fluid)

[- Navier-Stokes

Equation

Bernoulli Equation I

FIGURE 1. LINEAR MOMENTUM EQUATIONS

21

22

Requirements

refine

refine

Model n

Analyze andevaluate

(conceptual)

desrgn

Analyze andeva luat.e

Cpreliminary)desrg n

Generateinformation

for PlanningMan ufacturing

Tesling

Modifyand<refine

FIGURE 2.

(Detail\

\yesigV

PlanManufacture

Test

INTERACTIVE DESIGN PROCESS

SESSION I

CAD DEVELOPMENTS

106

4. CONCLUSION

It has to be seen that Chantiers del'Atlantique have been involved in the prac-tical use of that three dimensional modellingsoftware for a short period of time.

This experience confirmed that it is quitesuitable for the hull surface description. Itis a first step towards the complete threedimensional modelling of the ship geometry.

It can also be seen as an opened door to thefuture, both for surface modelling of ship'sstructure than for heavy CFD computations.

ACKNOWLEDGEMENTS

The author would like to thank his colleaguesfor their contribution to this article, aswell as Societe SISTRE who designed theprogram.

CFD and CAD in Ship DesignG. van Oortmerssen (editor)© Elsevier Science Publishers B.V., 1990

107

IMPACT OF CFD ON AERODYNAMIC DESIGN

N. VoogtManager Computational Aerodynamics

Fokker Aircraft B.V.P.O. Box 7600, 1117 ZJ Schiphol-Oost

The Netherlands

The increasing importance of CFD in aerodynamic design and analysis isdiscussed and important elements in CFD are highlighted. Several relevant CFDapplications in the Fokker 50 and Fokker 100 development phases are describedand their relationship with windtunnel and/or flight testing is emphasized.Finally a recent application of a newly developed code is shown and effortsare discussed which are needed to transform this code into an effective futureaerodynamic tool.

1. Introduction

During the last decades Computational FluidDynamics has been emerging as an important toolin aerodynamic design of aircraft configurations.As such it has obtained an essential role in thedesign process complementary to windtunnel- andflight testing. In the past the first phase ofthe design process relied heavily on windtunneltesting and aerodynamic shapes were optimized ina time-consuming trial-and-error cycle.Present-day CFD techniques can assist the designprocess by producing and analysing candidateshapes. On the basis of CFD analyses the numberof windtunnel models can be reduced signifi-cantly. At the same time detailed analysis offlow fields encountered in windtunnel- and freeflight environments can lead to a better under-standing of critical flow phenomena and evenreveal details which can not be obtained other-wise.In an industry environment the combined use ofCFD, windtunnel and flight testing is essentialto reduce design cycle times as well as potentialdevelopment risks.

The benefits of CFD can only be exploited ifthere is a suff icient conf idence 1eve 1 in thecodes and if a hardware/software infrastructureis available in which the codes are embedded.This infrastructure allows routine applicationswith short turnaround times and requ ires adequatepre- and postprocessing facilities.

At Fokker experience in CFD applications hasaccumulated over the years and especially duringthe development phases of the Fokker 50 andFokker 100 projects all the available codes wereapp 1ied. F1ight measurements performed onprototypes of both aircraft provided uniqueopportunities to compare computed and measureddata.

The present paper describes some essentialelements of CFD, high 1ights a number ofapplications and presents some aspects of futureapplications.

2. Description of the design process

The aerodynamic design process is aimed at full-filling a number of aerodynamic requirementsunder certain geometric constraints. It involvesa computational cycle and experimental investiga-tions and has essentially an iterative character.In the computational cycle a configuration isoptimized for a set of selected parameters at thedesign condition. The windtunnel is then used toassess the aerodynamic characteristics for a widerange of operational conditions. Results from thewindtunnel investigation usually lead to a newdesign cycle with adapted design parameters.

As an example the present-day wing design prac-tice is illustrated in figure 1.

EXPERIMENTS COMPUTATIONS

~ ~ E!f{,~ ~~ M Q

~ ~~.~_j

Figure 1

108

The process leading to the optimum wing geometryis essentially different from the traditional oneshown on the left. Starting point here was asuitably selected wing geometry for whichwindtunnel experiments determined the aerodynamiccharacteristics. By making successive modifica-tions to the windtunnel model and evaluatingthese in the windtunne 1 the best design wasestablished.In the present computer aided design process thestarting point is a selected flow field aroundthe wing at the design condition and the geometrywhich has to produce this type of flow iscomputed by means of inverse methods. The un~er-lying philosophy is that the selected flow fleldfor the design condition determines the aero-dynamic characteristics for the whole operationalrange of the wing. After the computational loopis finished the aerodynamic characteristics aredetermined experimentally and if necessary a newcomputational cycle starts with adapted design(pressure distribution) parameters.

In the computational cycle the design iterationsfollow three main steps:• choice of aerodynamic parameters for the

design conditionwing geometry design using inverse methodsflow analysis at off-design conditions usingdirect methods

In the first step the required flow field at thedesign condition is translated into a pressuredistribution on upper and lower surfaces of thewing.

The differences between this target pressuredistribution and the pressure distributioncomputed for an estimated initial wing geometryare minimized in the iterative design processthrough a sequence of inverse and directcomputations.

In the inverse procedure the geometry is theresult of the computations and in order toprevent unrealistic shapes constraints can be puton geometric characteristics such as localincidence, local thickness, etc. [1].For the new wing geometry flow field computationsare then made to assess the aerodynamicperformance in off-design situations ranging:fromlow-speed, high-lift cases to high-speedtransonic flow cases with strong shock waves. Thecomputational cycle is continued until bothgeometric and aerodynamic characteri stics aresatisfactory.

The windtunnel is still an essential element inthe design cycle because not all flow conditionscan be simulated by CFD. This is specifically thecase for low-speed, high-lift and high-speedcases with local areas of separated flow on thewings. On the other hand CFD can assist inunderstanding flow phenomena which occur inexperiments and it can also show details whichcan not be found directly from windtunnelexperiments.

3. Elements of CFD

A prerequisite for CFD to be an effective toolfor aerodynamic design in an industry environmentis the availability of a well-established hard-ware/software infrastructure. Key elements insuch an infra-structure are different types ofcomputers, graphic terminals and workstations andflow codes in combination with pre- and post-processing software. Key player is the CFDspecialist using his graphic terminal or work-station.Computational models stored in the database formthe basis for input to the flow codes.Preprocessing software is used to arrange geome-tric information as required by the differentflow codes and to add information determining theflow condition and numerical parameters.

Flow codes are run on different computer systemsranging from general purpose VAX computers forinteractive processing and small batch jobs tothe IBM 3090 mainframe computer for compute-intensive number crunching and the NEC SX-2supercomputer at NLR for specia 1ised app 1icationsof advanced flow codes.

Resu lts of the computat ions are stored in thedatabase and can be visualized on graphicterminals or workstations. Such facilities areessent ia1 .for processing the ever increas ingamount of information which is generated by someof the codes.Examples of visual postprocessing are illustratedin figure 2a showing a changing isobar pattern ona wing when the slipstream effect from aprope 11er is inc 1uded and in figure 2b wherelocal flow directions are shown on the Fokker 100front fuselage. The latter information isimportant when external devices have to bemounted outs ide the fuse 1age without interferencewith the flow field.

Figure 2aPropeller on

Figure 2b

3.1 Geometry modelling

Close ly connected wi th CFD is geometry mode 11ing.CFD codes require geometric input in the form ofspecially arranged sets of points. For mostapplications these points represent the surfaceof an aerodynamic shape. At Fokker and many otheraeronautical industries the three-dimensionalsurface modelling and definition system CATIA isused for CAD purposes. Because CATIA has ananalytical surface definition it can not be useddirectly to provide input for CFD codes.

Therefore for the specific purpose of modellingfor aerodynamic computations an in-house devel-oped system - GAMMA - is being used. It hasinterfaces with CATIA such that grids of surfacepoints extracted from CATIA can be used in GAMMAand conversely a surface can be fitted in CATIAthrough points defined in GAMMA. Figure 3 illu-strates the different surface representations inCATIA (left) and GAMMA of the complex Fokker 50engine nacelle.

,-- ..~

-. w"/(_ ->" ~,.' .~--, '

..~;_..-;(// -, .".,

Figure 3

Through the GAMMA-CATIA interface the linkbetween CFD and windtunnel is also establ ishedbecause CATIA is also used by Research Institutessuch as NLR in the process of producing wind-tunnel models.GAMMA uses an underlying data structure whichenables efficient access through preprocessing toa range of aerodynamic codes. The geometricmodelling can easily be adapted for a specificflow problem as shown in figure 4 with differentmodelling schemes for the Fokker 100 as appliedfor different flow codes based on panel methods[2]. Both are wire frame models which are createdby connecting all defining points by straightlines. In this way the surface is co~ered by flatpanels.

Figure 4

109

For the different models the computations servedifferent purposes. In one case the purpose is todetermine overall aerodynamic characteristicssuch as spanwise load- and moment distributionsby means of vortex lattice methods and a coarsepanelling is sufficient. In the other case a moredetailed investigation is required into aerody-namic interference between wing, fuselage andengine nacelles. To obtain accurate pressuredistributions a finer panelling is needed andalso the actual wing surface has to be modelled.The GAMMA system was originally set up forpreparing geometric input and many functions wereimplemented for all sorts of geometric mani-pulations. It was later extended for visualizingcomputed results in combination with thegeometry. Beside visualizations such as thoseshown in figure 2 other powerful appl icationsinclude visualizing pressure fields by differentco lours each represent ing a range of pressurevalues.

4. CFD applications for Fokker 50 and Fokker 100projects

During the development phases of the Fokker 50and Fokker 100 projects CFD has played animportant and sometimes indispensable role. Thesevere time constraints dictated by the rapidsuccession of both projects required shortsdesign cycles. Under such circumstances theavailable infrastructure as described above is anessential element. Some of the many applications[3,4,5] will be discussed.

4.1 Fokker 100 wing design

CFD has been a vital element in the design of theFokker 100 wing. In fact aerodynamic computationshad first pointed out the potential benefits ofmodifications to the original F28 wing. Theaircraft was designed in the early sixties witha design condition for the wing at a modest Machnumber M=.73 and a low liftcoefficient CL=.2. TheFokker 100 wing design required a designcondition at higher Machnumber and much higherliftcoefficient. An additional design requirementwas that the original F28 torsion box should beretained, so that wing modifications had to belimited to front and rear sections of the wingcontours.Computations indicated that the drag developmentwhich occurs on a representative F28 section canbe reduced significantly by a leading edgeextension. Figure 5 shows that there is a largedrag increment at a 1iftcoefficient of C1=.5between M=.5 and M=.7 for the F28 section.

This is due to the development of a strong shockwave at the front of the airfoil (see figure 6 onthe left). By means of the leading edgemodification shown in figure 5 local curvaturesare reduced and the shock disappears at that flow

112

The computations show that the displacementthickness of the wake which pushes the outer flowaside is suddenly increasing when the downstreampressure gradients caused by the deflected flapbecome too strong. Due to this thickening wakethe circulation around the airfoil can no longerbe increased and maximum lift is reached.

4.4 Ice accretion on Fokker 50 inlets

In the design of modern aircraft much attentionis being given to problems of ice accretion onaircraft components in flight. When ice accretionoccurs on wings and tailplanes these surfaces canloose much of their effectiveness which - ifuncorrected 1eads to dangerous f 1ightsituations. Similarly ice accretion can not beallowed to block vital air intakes.During the aerodynamic development of the Fokker50 main emphasis was put on designing new enginenacelles and much attention was paid to flowinvestigations in those areas such as inlet ductswhich are sensitive to ice accretion.

Figure 12 shows the very complex shape of theinlet duct system and several cross sections. Theshape of the main duct shows a gradual transitionfrom the cross-section at the throat A to theengine inlet flange B. Before the flow enters theengine inlet a part is branched off into a bypassduct C which discharges at the bottom of thenacelle and which protects the engine fromingesting foreign objects.

c

Figure 12

To investigate ice accretion in the main inletduct a model was put in the icing tunnel of NAEin Canada. In those tests the tunnel walls werevery close to the nacelle inlet and there weredoubts whether in that set-up the free-flightsituation was sufficiently simulated. To analysethe prob lem simp le computat iona1 methods wereused which predict ice accretion on the basis ofcomputed droplet impingement. Information fromsubsonic flow field analyses is used in methodsto solve streamline and droplet trajectoryequations.In such equations empiricism plays an importantrole: it is applied in the determination of thedroplet drag force and in modelling dropletbouncing against a wall.

For the analysis three-dimensional computationalmodels were used with cross-sections shown infigure 13. The configuration on the leftrepresents the inlet in the icing tunnel and theother one the nacelle as installed on the wing.To simulate flow through the duct uniform inletflow velocity distributions are specified at thecontrol surfaces as indicated.Three-dimensional flow field computations weremade to simulate the flow condition in the icingtunnel. Angle of attach was selected such thatlocal flow directions upstream of the nacelle areidentical in both cases.

F---~, .. ,

~ ". ." -"".L . - •

1'--Icing tunnel Free flight

Figure 13Figure 13 shows the computed droplet trajectoriesand clearly indicates the need for anti-icingprovisions in the air intake. The results alsoshow that when the intake iss ituated in thevicinity of the tunnel walls the dropletimpingement changes significantly. In that casethe airflow follows the curvature of the tunnelwall and drives the droplets in a downwarddirection.These computations showed that results from suchtests have to be interpreted carefully and thecomputed trajectories were used together with the'results from the icing tunnel for a number ofcritical flight- and atmospheric conditions todetermine the best locations for the de-icingblankets.

The problem of ice accretion also came up duringa redesign of the inlet to the oilcooler locatedon the lowers ide of the nacelle. Experimentsshowed that the original NACA type submergedinlet could not maintain sufficient air flowthrough the oilcooler unit and because at thesame time increased engine power requirementsdemanded more coo 1ing air the intake was re-designed.As a result of the modification shown schema-tically in figure 14 the inlet is no longersubmerged so that the possibility of ice

-la--original modified

Figure 14

accretion had to be investigated. This was firstdone computationally and later checked in flighttesting. Two-dimensional computations were madefor a cross-section which also included thepropeller spinner and the main engine inlet.Droplet trajectories were computed for variousdroplet diameters in a flow situation at 15000 ftwith an aircraft speed of 100m/sec.Figure 15 shows the result of a computation fordroplets of 30 micrometer diameter. The resultindicates that ice accretion will occur on theleading edge and inside the main engine inlet andon the leading edge of the inlet to theot lcoo ler .

Figure 15

On the basis of the computations it was decidedto leave the .oilcooler inlet lip unprotectedbecause the droplets do not enter the intake. Thecomputational result was confirmed by fl ighttests which showed that although spiky accretionoccurred around the inlet leading edge asufficient airflow could be maintained throughthe oilcooler unit.

5~ Preparing for future CFD applications

In the applications described above use was madeof CFD codes many of which were developed in theseventies and early eighties and for which a wideuser experience has been accumulated. Tofacilitate application of these codes pre-processing software has been developed whichautomates most of the sometimes laborious inputpreparation and enables the user to optimizegeometric input graphically and interactively.Similarly postprocessing software is availablewhich enables the user to prepare and presentresu lts graph ically without hav ing to extract andinterpolate data manually from the vast amountwhich is created by the computational process.For future applications this aspect of the userbeing interactively involved in processingincreasing amounts of data is an essentialelement for reducing turnaround times.

Especially the advanced CFD codes which are beingdeveloped and evaluated at present [6] requireextensive geometry processing for generatinggrids around aerodynamic configurations. For acomplex shape consisting of fuselage, wings andengine nacelles grid generation can take as muchas several weeks whereas the actual computationis done overnight and postprocessing takes a fewdays.

113

Before these new codes become cost effective andhave short turnaround times much effort will haveto be invested in speeding up these preparatoryprocesses. The justification for all theseefforts is that the new codes provi de betterphysical modelling of flow problems and thereforebetter prediction of critical flow situations.

An example of such a new CFD tool which is beingdeveloped and evaluated in close cooperationbetween NLR and Fokker is a code based on theEuler equations.This method describes rotational flow and isespecially suited for the prediction of propellerslipstream interference with the aircraft. Fokkeris preparing for future applications of this codein two major areas: in geometric preprocessingand in validating the code.

The importance attached to the new code can beillustrated by the fact that specific flighttesting was done to measure slipstream effects ona Fokker 50 prototype. The purpose was toinvestigate the effect of the propellerslipstream on wings, nacelles and tailplanes andto compare these measurements with Euler flowcomputations [7].

The computational cycle contains the followingsteps:- geometry (re)definition- block decomposition- grid generation- flow computations- result visualization

The computations require a grid of pointsconstructed in the flow field around aconfiguration. For reasons of flexibility thecode is set-up such that the grid generationprocess is subdivided and executed in a number ofseparate blocks. To be able to define the blockboundaries a redefinition of the surface grid isoften required. Figure 16 shows that thecomputational model of the Fokker 50 as definedby GAMMA and used for input to other CFD codeshas been modified substantially. The surface wasredefined in CATIA and the new surface grid hasbeen transferred to GAMMA.

Figure 16

114

The topological and geometrical subdivision of aflow domain is a highly iterative procedure usinginteractive graphical methods. The eventual gridgeneration process where grids are defined ineach block is controlled by many parameters andis also highly iterative. A number of partialcross-sections of the grid is shown in figure 17.For the computation the finite flow domain aroundthe Fokker 50 was divided into 556 blocks and forsufficient resolution a maximum number of 2.2million grid points was used. For computationswithout slipstream effects usually only a halfconfiguration is needed for reasons of symmetry.In this case however the complete configurationhad to be modelled because the propellers areco-rotating.

Figure 17

The effect of this asymmetry is illustrated infigure 18 which shows the spanwise distributionof lift computed for a typical cruise conditionwith slipstream effects as compared to thecomputed distribution for the wing alone.Although the total amount of wing lift is notmuch increased as a result of the slipstreameffect the resu 1t shows that very 1argedifferences in local lift occur.

Computations such- as these require extensivecomputer resources. For sufficient accuracy onecomputation on the fine grid level consumes 10CPU hours on the NEC SX-2 supercomputer of NLRwhereas some 120 CPU hours would be required onthe IBM 3090 computer at Fokker. Because accessto this mainframe computer for large-scalescientific computing is limited it will be clearthat such computations can not yet be done on aroutine basis.For the measurement of slipstream effects a totalof 10 pressure belts were mounted at variousspanwise stations on the wing of a Fokker 50prototype. Each pressure belt included 30chordwise distributed static pressure orificesconnected to scan iva 1ves. The fl ight test programcovered a matrix of conditions with variations inspeed, angle of attack, flapsetting and propellerthrust.

IMach = 0.36 • Tc = 0.101

Figure 182~_'.~~_~.~~_.~~_~'~~_2--~--~~~-+~~.~~'.Xle-I ET....

Several computations were made corresponding witha number of in-flight measurements. Figure 19gives a comparison of computed and measuredpressures on the wing at several locations behindthe propeller for a typical cruise condition. Incorrespondence with figure 18 it can be seen thatthe upwards rotating propeller causes an increaseof lift and of local velocities on the wing.

For this case the influence of the propellerslipstream is only moderate and generally a verygood agreement is shown. For other casesespecially those encountered during climb wherethe slipstream influence is much stronge~comparison between experiments and computationsindicated that some modifications are required inthe modelling of the slipstream.

srMaOLS: FL!GHT TESTLlNE : EULER RESUL,S

F[NE GRID

A./ 1 \.,1V '" '....._

I =,~!

I~

C -,/

_".r-~ -.r <,~

Figure 19

In the near future much attention will be givento such deta i1s as well as to improving theefficiency in the computational cycle. Mucheffort will be spent on improving preprocessingand on speed ing up the actua 1 computat ions bydeveloping more efficient algorithms so thatfuture applications can be performed more cost-effectively.

6. Conclusions

The vital role which CFD is playing in modernaerodynamic design and analysis is clearlydemonstrated by the applications made during therecent development phases of the Fokker 50 andFokker 100 projects.Some conclusions are:- CFD methods have to be used in close

combination with windtunnel- and flighttesting.

- combined application of various geometrymodelling techniques is essential for anefficient flow of information betweencomputational- and windtunnel models as wellas for grid generation.

much effort has to be spent on the infra-structure in which CFD codes are embedded toensure fast and efficient applications.

115

7. References

[1] J.M.J. Fray and J.W. Slooff, A constrainedinverse method for the aerodynamic design ofthick wings with given pressure distributionin subsonic flow. AGARD CP285 paper 16, 1980.

[2] H.W.M. Hoeijmakers, Panel methods in aerody-namics; some highlights. NLR MP87028 U, 1987.

[3] E. Obert, The aerodynamic development of theFokker 100. ICAS-88-6.1.2, 1988.

[4] N. Voogt, W.J.A. Mol, J. Stout, D.F. Volkers,CFD applications in design and analysis ofthe Fokker 50 and Fokker 100. AGARD CP437paper 20, 1988.

[5] J. van Hengst, N. Voogt and G.J. Schipholt,Aerodynamic design and testing of Fokker 50nacelle and intake ducts. AIAA-89-2483, 1989.

[6] J.W. Boerstoel, A.E.P. Veldman, J. v/d Voorenand A.J. van der Wees, Trends in CFD foraeronautical 3-D steady applications: TheDutch Situation. NLR MP86074 U, 1986.

[7] J.L. Kuijvenhoven, Validation of propellerslipstream calculations using a multi-blockEuler code. AIAA-90-3035, 1990.

CFD and CAD in Ship DesignG. van Oorttnerssen (editor)@ Elsevier Science Publishers B. V., 1990

PRACTICAL APPLICATIONS OF AN INTEGRATED DESIGN SYSTEMIN SHIP HYDRODYNAMICS

Ilmo KUUTTI

Masa-Yards Inc., Turku New Shipyard, P.O. Box 666SF-20101 Turku, Finland

A short description of the development of a computer aided design system for navalarchitects is given. The NAPA integrated approach is described and the operational anddevelopment experiences are discussed. The basic architecture of the system - consistingof user interface, data base system and application modules - is presented, and the mainfeatures of the subsystems are mentioned. The hydrodynamic part of the NAPA system isdescribed and the capabilities of the integrated ship design system are presented in formof solved hydrodynamic problems.

1. BACKGROUND 1.2. Requirements for the new system

1.1. A short long history The experience from the users and the knowledgefrom the maintenance and development of the oldST -system helped very much the system analystto set the t'equirements for the new one. Someof the main ideas for the new system were:

The development of the NAPA system is a goodexample of how large software systems are beingdeveloped for engineering purposes. Shipyardsin Finland have been using computers since theadvent of first models. Hydrostatic calcula-tions were suitable applications to thecomputers of that time, as a lot of calcula-tions were required for a limited amount ofinput data, and the results could be presentedby a few numerical tables.

It must be easily adapted to the needs in thefuture:

basic solutions must allow versatiledesign and development of the systemhighly modular architectut'e must befollowed, without exceptions, to renderpossible modifications and extensions

Since the good-old-days, the navalarchitectural calculations have used allavailable capacity of the hardware, and thesystem developers have tried to keep thesystems up to date.

Easy to maintain and to port to severalcomputers and operating systems:

the software should last much longet'than the hardware on which it will bedeveloped and usedFortran -77 language was selected, as itwas the most standardized language forrnathematical programmingdetailed internal programming standardswere developed and adopted to allsoftware workthe system structure was carefullydesigned to avoid commitment to aspecific hardwarethe hardware or operating systemdependent operations were separatedcat'efully from the main parts of thesystemthe software must use all avallablepower in the hardware system

The first Finnish integrated software packagefor naval architectural calculations wasdeveloped in the late sixties and earlyseventies at the Wartsila shipyards. Thedevelopment of the system is described bySaarilahti in [1]. The so called ST -packagewas used by most of the Finnish shipyards andego by the Kockums shipyards and the Royal Navyin Sweden.

In the late seventies the limitations of theold ST system were so hard to overcome and somuch could be done with the new era of mini-computers, that it was decided to replace thesystem by a totally new one starting fromscratch. The new system got the name NAPA,which stands for the Naval ArchitecturalPackage.

117

118

some limitations of the Fortran -77standards were avoided by programmingpractical tools for dynamic memorymanagement, character handling and formaintenance of the software code anddocuments

The system must be developed for a long lifespan:

further extensions, modifications w1dimprovements must be easy to makeit must be possible to utilize newpossibilities of the modern hardwareit must be possible to replace part ofthe system without disturbing otherparts, which makes it possible to updatethe system step by stepall the main parts of the system must bedeveloped as well as possible

The program must fulfill all the needs of allusers:

the vast development costs can be sharedamong several users if the system servestheir needswhen there are more users to give theirfeedback, the system will become capableto handle more versatile problems andtherefore serve all users betterit is impossible to limit the futureneeds, so it is better to try to developthe system to handle all possible tasksrequired by any user

1.3. Developing organization

Traditionally large systems are developed in anindependent department or a separate companythat is not using the system in their everydaywork. Paper documents and specifications areused to describe what the system should looklike and what it should do. This kind ofdevelopment is not efficient in large, com-plicated engineering systems were the users donot really know what they want until they havesomething to start working with.

The development of the NAPA system is mainlybased on using prototypes, were a problem firstis solved by a relatively simple method thatcan be used for testing the idea, developing itf'urt.her and under-standang what is reallyneeded, after which a final solution can bedescribed. This leads to a short time from thestart to the first production use of a newapplication, but it can also easily lead to anendless loop of testing, improving andmodifications.

The prototyping also requiI:'esthat the end-Llsers are running the prototype in their normal

work and the communication between the devel-opers and the users is easy and lively. In somecases the optimum is when the useI:'is also thedeveloper of the system, as the communicationis not needed at all!

In the NAPA development work some six to tensystem analysts have been working together withabout the same number of users testing theprototype in real environment with real data.The communication problem has vanished in thecourse of time because the staff has beenalmost the same during these ten years ofdevelopment and they now know each otheI:'well.The development and the testing was done bothin Helsinki and Turku, which required someadditional work in communication in the begin-ning. Later a network could be arranged toconnect the two sites separated by a distanceof 165 km, and the communication problem wassolved completely.

2. NAPA INTEGRATED APPROACH

2.1. Basic structure

The layout of the NAPA system is presented inthe figure 1.

The system is controlled by the monitor, whichtakes care of all administration tasks withinthe system.

The data management is handled by two systems:1) the Data Base system stores and retrievesthe data from external data bases located onthe disks and 2) the Data Management takes careof the dynamiC memory management during therun, allowing programmers to use powerful toolsto generate and manage complicated datastructures in the applications.

The user interface is handled by the followingsystems:

AI Alphanumeric and InputAP Alphanumeric PrintGR Graphical Input and Output

Auxiliary tools for development are:

MN MonitorAD AdministrationER Error handlingCH Character handlingGB Basic geometryGM Geometric modellingIN Integrals and basic mathematics

•

APPLICATIONSGeometr~Stab iLiE~CapacitiesSpeed/PowerSeakeeping

USER INTERFACECommandsListsGraphicsNAPA basic

DATA BASE AUXLIARYSYSTEMS

AdministrationError hand LingMathematicsTooLsDATA MANAGEMENT

Figure 1

2.2. Application systems of NAPA

GM - Geometry

The geometry subsystem handles definitions ofall geometric objects like surfaces, rooms andarrangements. Any hull form can be describedwith the system including catamarans, twinskegs and even off-shore structures.

The geometric objects are built mainly byreferring to other objects, which guaranteesthat the definitions are consistent aftermodifications to some parts.

The geometry system allows also transforma-tions, translations and rotations of allobjects defined in the data base.

SM - Ship Mode!

With the ship model the user can create aproduct model of the project. The generalarrangements, compartmentation, parameters forall compartments, tank lists and arrangementdrawings are handled with the ship model.

The user can select an interesting subset ofthe objects and ask enquiries about thevolumes, areas, centers of gravities or

---------,.,

119

capacities concerning of the particular subsetof interest.

HD - Hydrostatics and stability

All hydrostatic and stability information ofthe ship is presented either by tables orplots.

CP - Compartments

Contains definition of sounding devices,calculation and presentations of sounding andullage tables for a heeled and/or trimmed ship.Produces also tonnage calculations and compart-ment hydrostatics, including the real and IMOfree surfaces.

LD - Loading conditions

Contains definition and handling of loadingconditions, plotting of 'loaded' arrangementand presentation of stability curves.

DA - Damage stability

Damage stability calculations for severalalternative loading conditions presentedagainst all necessary stability criteriaincluding those defined by the user.

CR - Stability criteria

Both the static and dynamic stability criteriaincluding the user defined criteria. Plots ofrequired GZ-curves and required GM includingwind-, passenger-, and turning moments ifnecessary.

MisceLLaneous

In addition to the main tasks, the NAPA systemcontains several auxiliary tools that can beused for almost any kind of work. The two mostoften used tools are described in the nextpar'agr'aphs.

The table calculation (spr'ead sheet) tool canbe used together' with all subsystems in NAPA.All quantities defined into the system aredirectly available in the table calculation,and the user' can define easily new quantitiesand formulas. The user selects the data itemsfrom the data base which he is interested in,loads the data from the data base into thetable and processes the data further. Theregr'ession analysis tools are directly avail-able within the table calculations, so the user'can create new formulas and calculate correla-tions and other statistical values. The resultsin the table can be presented either with thetable output facilities or in the graphicalform with diagrams and bar charts.

120

The text editor is used normally for preparingthe calculation input data, but can be used forother purposes as well. The text processorincluded in the editor is an effective tool forhandling large texts mixed with drawings anddiagrams.

2.3. The user interface of the NAPA system

NAPA has an interactive command-driven userinterface rich with powerful commands availableto the user at all times. Part of the commandsal~ for general use and therefore availableeveri~ere in the program. These transparentcommands are used to control the total environ-ment (eg. graphics) and general purpose toolslike the function calculator and the macrolanguage. The total number of commands exceeds2000 and each command has several options, butthe user has to know only a fraction ofpossibilities in his normal work.

A menu of available commands, help- texts,examples, list of command parameters and aninteractive tutoring module is available on-line in case the user does not know how tocontinue. Also a special tourist mode isavailable in the general tasks.

The main features of the user interface are:

powerful for professional userscommand menu, explanations and messagesto the user are presented to the userwhen neededinteractive help feature for tutoringpurpose and for quick learning of newtasksdetailed instructions and examplesdirectly available when neededuser can define new permanent tailoredcommands of his owntemporary user-defined commands can bedefined at any time during the runpossibility to create graphical tabletmenusinteractive macro programs can be writtenby the user with the NAPA BASIC languagea short error message at an errorcondition, longer error explanation whenaskedtakes care of all engineering unitconversions both at the input and outputand allows dynamic formatting of numbersand character strings

Output facilities comprises of:

versatile colour graphicspowerful report generators for definingand modifying purpose-adapted lists andgraphics

all data is stored into the database, soit is possible always to return to thecalculations and prepare newpresentations of the resultscombining results from severalcalculations into one presentation caneasily be donesupport of several hardware devices foralphanumeric print and graphics

The whole command set of the user interface,the data base contents and the main calculationfUnctions are directly available in the macrolanguage.

Graphical input can be used for showingcoordinates in the geometry, pointing objectson the screen and for giving commands from thegraphical menus.

2.4. Experience on an AI based user interface

The NAPA system is mainly designed for theprofessional user who knows the thing he isworking with and knows also how to use NAPA.

A large Finnish project in the artificialintelligence (AI) was started about four yearsago to study intelligent user interfaces forlarge software systems. The hydrodynamic sub-system of NAPA was used as one test case of theintelligent user interface. The test case isdecribed by Linnainmaa and Priha in [2).

The AI user interface was developed by theTechnical Research Center of Finland with theKnowledge Engineering Environment (KEE) runningon the Symbolics computer. The NAPA system wasrunning on a VAX computer connected via networkto a Symbolics Lisp computer.

The user interface was to contain knowledgeboth about the hydrodynamics, and also how theNAPA system should be used to make the studies.The system was working like an expert user de-fining input data for calculations, selectingwhat methods to use and to judge if the resultswere good or bad.

The prototype was developed within about oneyear, and was tested by both experts and noviceusers, but it has never been utilized for realshipyard use.

Some conclusions drawn from the experiment are:

the development of such a user interfacetakes an enormous amount of workit is difficult to combine two differentarchitectures, one written in Fortranlanguage algorithms and the other basedon rules and statements written in LISPlanguage

rthe hardware requirements in -87 were toomuch for a realistic investment due tohigh prices of lisp computers, but nowthe situation is much better in thatrespectmajor difficulty is in keeping theknowledge up to date when the applicationprograms develop further and theexperience in the hydrodynamics isincreasing.

After this experiment, the development in theuser interface has focused on helping the userto find the right documents and the help textsin the system. Also much can be improved bywriting more detailed error and warning mess-ages, testing of more possible misuse of thesystem and by increasing the knowledge of thesystem with the traditional programming tools.

2.5. Data base solution

In an integrated software system the data basesolutions are of vital importance, as the shar-ing of data among the applications is takencare of by the data base system. Modern hard-ware allows users to create and maintain verylarge data bases, which are efficient for thetotal performance if used properly.

The data organization in NAPA follows the reallife in ship design. All data is stored intodata bases according to projects. The database for one project is divided further intoversions similarly to the design alternativesof a ship project. Typically each version hasdifferent main dimensions, hull form or achanged general arrangement.

In one version each task can further split thedata base into independent packages, which arecalled descriptions in the NAPA terminology.Descriptions can be separated by their name,type or according to their contents, and withinone description a user or the system can storewhatever data of any type and as much as need-ed. A description is thus only a package thatcontains data belonging logically into one set,and which as a whole belongs to one version ofa design project.

Most of the description names are handled bythe system, but the user can also use his ownnaming convention to organize his data forlater use.

With the user interface and the data basesystem it is possible for example to:

select an interesting subset of designcasesspecify what studies should be made foreach case

121

perform all operations for all selectedcasespresent the interesting quantities inone presentation.

Even though the NAPA data base can be usedsimilarly to a modern relational data base, itis not a relational data base in its internalsolutions.

As it is shown for example by Stanley andAnderson [3], the relational data bases aresuitable for business and administration dataprocessing where there are many instances for arather limited number of different recordtypes. In the knowledge engineering and in theartificial intelligence the situation is theopposite: there is a large number of differentdata types with only a few instances each. Atypical situation in the large engineeringapplications contains the worst combination ofthe two: there are many instances of manydifferent data types.

The solution developed for NAPA has proved tobe efficient and practical for typical condi-tions allowing flexible operations with lessoverhead than some other engineering systemsusing commercially available relational databases.

3. HYDRODYNAMICS SUBSYSTEMS

3.1. General features rThe number of different quantities handled inthe hydrodynamic calculations is much largerthan in other naval architectural calculations.Hydrodynamics are full of nondimensional ordimensional parameters required in the input,and in the output the results can be presentedin many alternative ways requiring own quan-tities for each item. Also, the use of nondi-mensional merit coefficients helps the user tosee that the results seem to be correct. Thenumber of quantities in the hydrostatics andstability is 277 and in the hydrodynamics atotal of 2548 different quantities are defined!

The interesting quantities differ from oneproject to an other, and therefore the hydro-dynamic programs in NAPA produces all possibleresults into the data base, from where the usercan select the results he is most interestedin.

New calculation procedures are developed allthe time and the hydrodynamic programs shouldchange respectively. Therefore a lot ofattention was paind to designing the run timeenvironment for the hydrodynamic subsystems.

Adding of new tasks or modifying of an existingone is made so easy, that the all new modifi-

122

cations will be produced directly into NAPAproduction version. Separate main programs withindipendent input or output fuctions are notused, which leads to minimal programmingeffort.

The hydrodynamic calculations in NAPA aregrouped into subsystems according to what kindof studies they are intended for. In the nextchapters the main characteristics of the appli-cations of each subsystem are described.

3.2. Resistance and propulsion

The basic system SH contains tasks for geo-metric data handling, more than ten differentresistance prediction methods, propelleroptimizer, a propulsion analysis task andseveral methods for prediction propellerinduced vibration excitation.

When the hull form is defined into the database, necessary input data for resistancecalculations is directly available based on thesame data that is used in the hydrostatics etc.Therefore the misinterpretation and inaccur-acies in the input are avoided and the calcu-lations can start immediately. In case only themain dimensions are known, the statisticalvalues are calculated, or the user can definehis own values.

All calculation tasks use the same definitionfor the input data, so all methods can becalculated directly. Wind resistance, append-ices, form factor l+k and extrapolation allow-ances Ca can be defined by the user or thedefault values are calculated by the ,system.

The calculated values can be corrected accord-ing to the correlation factors derived from themodel test results and calculations for earlierdeSigns.

The propulsion coefficients can either be cal-culated within the propulsion calculations, aconstant value defined by the user is used orthe coefficients measured in the model testscan be entered into the system. Open watercharacteristics can either be from Troost-Bopen propeller series, several due ted pro-pellers or the model tests results. All import-ant corrections due to hub ratio, surfaceroughness, blade thickness etc. can be includedif necessary.

In the propeller excitation predictions, theactual blade geometry and the hull form aretaken into account automatically, so theresults are accurate even if the pitch setting,propeller diameter or other main parameters arevaried in the calculations.

3.3. Seakeeping

The geometry routines of the seakeeping systemSHS contain one task to extract 2D strip pro-perties from the hull surfaces to be used inthe strip theory calculations for slendervessels, and another routine to generate planefacets on the wetted surface to be used in the3D sink-source method.

The seakeeping properties of the vessel arethen calculated either by the strip theory orby a three dimensional sink-source method, orthe transfer functions measured in the modeltests can be used for further studies.

The post-processor of the seakeeping systemallows versatile statistical and mathematicalstudies of the performance of the vessel inarbitrary sea conditions. Studies like:

Significant valuesdown-time analysislimiting wave heightnumber of occurrencesscatter diagrams

can be made for:

motions, velocities andaccelerationsrelative motionsbottom slammingflare impac tswave added resistancedrift forcesetc.

3.4. Manoeuvring

Manoeuvring subsystem SHM can be used for alltime domain studies of the ship motions. Itcontains easy to use tasks for normalmanoeuvring simulations:

turning circlespull-out testszig-zag testsspiral and reverse spiraltest.directional stabilitysteering quality indices

Several thrusters, rudders, rotatable thrust-ers. open propellers and ducted propellers canbe described into the vessel, so all slow speedmaneuvers like harbour manoeuvring can bestudied.

Environmental conditions including the effectsof wind, waves, current and shallow water canbe taken into account, if necessary.

The manoeuvring characteristics of the hull canbe calculated by statistical formulas or thecoefficients measured in the model tests areused in the calculations.

The results can be shown either in the form oftables, diagrams or in trace plots.

3.5. Information system

Information system IS contains tools fordefining "user company" - based data bases forstoring all kinds of information on the builtand tested ships. At Masa-Yards the model testresults are stored into the NAPA data base withthe information system to help comparing thecalculated and new model test results to thestatistics of the older designs.

The Fairplay International Research Services(FIRS) delivers large ship files of all vesselsdelivered, under construction or on order. TheFIRS files are converted into the NAPA database form by the routines included in theinformation system to allow further studieswith regression analysis, scatter plots etc.

3.6 Auxiliary tools

An essential tool in the normal use of thehydrodynamic programs is the function calcu-lator of the NAPA user interface. The userneeds only to write the formulas to make theneeded calculations, as the values in the database are directly available in calculations,and the results will also be stored into thedata base. For example to calculate the Froudenumber values for all speed values, the userneeds only to give the following transparentcommand:

!CALC FN=VS/SQRT(LWL*9.81)

The length of waterline (LWL) and all speedvalues VS are fetched from the data base andthe result goes directly into the data base. Asall values are stored into the data base in SIunits, no unit conversions are needed. Forexample, the unit for LWL can be feet at theinput and output and the speed values in knots,but the Fn values are calculated correctly!

The other most used tool within the hydro-dynamic calculations is the diagram plottingenvironment. It also gets the values from thedata base and the default formats and header

123

texts from the system data base, so the userinput can be minimized.

For example, to plot the resistance curve onlythe following commands are needed:

DRAWINGDIAGRAMARGUMENT VSFUNCTION PEPLOT

The result could look like follows:

4. EXAMPLES

:3I:

LQJ 1830a.QJ:>...,oQJ......

w

5

I'I

·1

II:

12 14 16 1S 28 22Ship speed KNOTS

Figure 2

4.1. Main dimension variation

This example describes shortly a main dimensionvariation carried out in a cruise liner projectto see, for example, the effect of maindimensions on:

power performance, lightweight. buildingcost and on the operation costsoptimum combination of main dimensionsas function of design speedsensitivity of the main parameters tothe main dimensions.

124

CALCULATION MODEL

The study is based on the regression formulasfor lightweight, material costs and man hoursetc. derived from the statistics of the cruiseliners built earlier by Turku and Helsinkishipyards.

A NAPA BASIC macro program was written togenerate a design variation and to run theanalysis calculations for the alternative.

The macro was then repeated by four DO loopsinside each other to vary all the requireddesign alternatives.

Totally more than 200 alternatives weregenerated to the data base, each with its ownhull form transformed from the same existingcruise liner. The displacement and, thepropulsion power was iterated into balance,which means that all variants are possible fromdisplacement and powering point of view.

Calculation of the lightweight, material costand manhours is based on exact volumes, areas,powers, main dimensions etc. calculated by NAPAgeometry and hydrostatics subsystems.

The stability was not checked in this phase, asthe calculation of good estimation of thevertical center of gravity should have requireda detailed general arrangement drawing whichwas not possible to generate at this moment.Also the GM requirements are so time consumingto calculate that they are not feasible to berepeated for all design alternatives. A separ-ate study was set up to find the minimum sta-bility requirements. In this study only thetransverse metacentric height was calculatedbased on the hydrostatics of the generated hulllines.

Powering calculations are based on theHoltrop -84 resistance prediction, propulsioncoefficients calculated by Holtrop formulas andTroost B-series open water characteristics. Theresistance and propulsion figures were correct-ed according to the correlation factors derivedfrom the reference calculations and the modeltests. The fuel consumption calculationsincluded the service condition effects.

ASSUMED PARAMETERS

Based on the number of passengers, requiredcabin size and on the statistics of severalsimilar cruise liners, the following designparameters were fixed:

Deadweight at the design draughtInterior areaAuxiliary engine outputEstimated volume of deckhouseEstimated total volume.

Capital costs of investment and the propulsionfuel cost is calculated for:

Interest rateTime frame of calculationsSea marginSpecific fuel consumptionOperation timeFuel priceUSD/tonAccurate propulsion power

8 %10 years10 %

180 g/kWh365 d/a

93

The service profile is expected to depend onthe maximum design speed.

VARIATIONS

The following variations were calculated:

5 Lpp values 1655 Beams: 274 Design speeds:

175 190 20528 29 3021 22 23

22031.5

24

2 DeSign draughts: 6.8 7.3

Each variation was stored into its owndirectory, so it is possible to return to anydesign alternative afterwards. The hull form isnot stored, only the last one is available.Anyhow, it is always possible to reproduce thehull form if necessary.

RESULTS

For one deSign alternative about 100 quantitieswere produced and stored into the data base andas the number of alternatives was more than200, the total data base contents was verylarge. It took quite a long time to find a goodway to present the results, but in the end themost suitable way was to use diagrams, as inone page 100 data points could be presented andthe reader gets the idea about the trends ofvariables.

The figure 3 presents one example output, inwhich two diagrams are presented. The values ofdesign alternatives are connected by smoothcurves drawn through the calculated points.

One design parameter is fixed in each page. Theargument in the diagrams is either the lengthbetween perpendiculars, the beam or the designspeed. In the same diagram all curves for thetwo design draughts and for all parameter valesare shown. The following quantities arepresented in this example:

Lightweight: Estimated lightweight for thealternative.

FueZ+CapitaZ Cost in 10 Years: The calculatedcost of propulSion fuel and the capital cost ofthe investment over ten years including theinterest.

90-06-28IKu

Pr"oj

DoleMain DimensionsMasa-Vards

TurKu New Ship~ord

Sign

Weight and Casts

Beam27.028.029.030.031 . S

a 62~(j)::::lz;

61e

oilL0OJ see>-

IS)

c S90

0....,.- 580Q.0U+

OJ sve::Ju,

sse

SS0160

(j)z:ClI-

21000

....)

..c01

19000-

Design Draught6.8 7.3

Design Sp•• d2111!:1.0 knot.

18000-r----,-----~----r---_,----~----_r-----r----._----._----r_--~r_--_,----~

M

M

1?0 180 190 200 210 220

Length between perpend.

160 180 1901?0 200 210 220

Length between perpend_

Figure 3.

125

126

q.2 Downtime analysis

For passenger vessels the seakeepingcalculations are typically made for:

to compare different design alternativesaccording to how often certain passengercomfort criteria are exceededto judge the operability of the vessel inseveral sea areasto calculate the wave added resistance inservice conditionsto optimize the speed, GM and heading tofind the best possible operabilityto study if a project design is realisticfrom the seakeeping point of view.

For an offshore project the seakeeping calcu-lations are used for a wide variety of studiesdepending on the project. The studies made atthe Turku shipyard includes for examplehelicopter landing and take-off analysis,calculation of relative and absolute motionsand accelerations, calculation of the driftforces in an anchoring condition.

The attached drawing presents an example down-time analysis maqe for a small passengervessel, where the limiting factors are:

15 % of the passengers are seasick in theforward restaurantshipping of green water on deck 5 timesper houreight degrees significant roll anglefour bottom slams per hour.

The figure q contains wave scatter diagram andthe limiting wave height for each criteria andthe calculated probabilities to exceed eachcriteria.

q.3. Propulsion in service conditions

This example shows the final result-of a pro-pulsion study carried out for a passengervessel project.

At first the resistance was calculated byHoltrop -84 prediction method for the trialcondition and the optimum propeller was de-signed according to the owners' requirements.The resistance was corrected according to thecorrelation factors calculated for a similarvessel. The in service effects that were takeninto account are:

hull foulingincreased hull roughnesswind resis tance -wave added resistance.

The propulsion powers were then calculated forresulting effective power in service condition.After the studies of the Pd-N diagrams, thereduction gear ratio and the main engine wasselected and the fuel consumption could becalculated.

Figure 5 shows the final results, ie. the fuelconsumption per nautical mile for several speedvalues and for alternative pitch settings.

5. CONCLUSIONS

The development of a large software system is acomplicated process, which demands experiencedsystem analysts and enthusiastic and patientusers to test the system. It also requires manydemanding projects to test the system, a longtime period to complete and quite a big budget!

The experience of the NAPA system has proventhe following statements to be valid for anintegrated ship design system:

1) The user of a large integrated systemdoes not have to learn many userinterfaces of many smaller systems, orto jump from one system to another whileproceeding with the ship design.

2) A totally integrated data baseeliminates all kinds of data transfer,which leads to a dramatic reduction inthroughput time and also eliminateshuman errors in the data transfer.

3) An integrated system structure and theright use of data base systems helpskeeping the design data in the rightorder despite the lively and flexibledesign organization.

q) An integrated design system allows usersto do - efficiently and with minimumeffort - complete design studies formany alternatives thus helping to findbetter designs.

5) The user of an integrated design systemcan combine all important informationfrom several calculations from the database into one clear presentation, whichhelps to present the design informationto other p~rticipants.

6) Integration means less programming workin the end, as many of the modules andexisting'tools can be utilized at theprogramming of new subsystems into alarge software sytem.

Masa-Yards

Turku New ShJp~ord

Pr-c j NAPASHIP/G

90-06-27IL

DOWNTIME PER CENT Dole

Sign

NORTH SEA

HEADING 135.00 DEGREE LPP 100.00 MVS 10.0B KNOTS BWl 20.00 M

TA 5.500 MTF 5.500 M

TOTAL DOWNTIME 49.01 x

_..- ..- .._.._.._.._.- lSX OF PASSENGERS ARE SEA SICK AT PC8S 10 13)

-._.-.-.-._._.- 5 DECK WETNESSES PER HOUR

- - - - - - - - SIGN. ROLL ANGLE 8 DEG

4 SU'lMS PER HOUR

~ 11

10....JJ:zn(l) 9J:

(l)

::> 803

c 7r»II)

rn 6C

....J

:J 5II)(l)

~4

3

2

DOWNTIME 48.56 ~DOWNTIME 28. 34 ~

DOWNTIME 0.52 >:

DOWNTIME 3.18 ~

; , ,,;'

I it I I : \ I I- - - - - -, - - - - -11 - - - - -I - - - - -,- - - - - -, - - - -\- T - - - - - ,- - - - -1- - - --

I' I \ I I, \. , 1, 1 \ ,J I I I \ I

_____ 1 J\ J. , 1 \_I _ _ _ _ 1. : _

1 " I I I \1

: :, : : 1 1 1 \! I

1 I \ I I I "" I I_____ -t 1 • I- , -t 1\ ... , _

: \ :, 2 '\\1: 1 ': \ ,1 I ,2: \: :

~~"""_-t - t - - - T - - - - - .... - - - - -,- - - - -t - - -'- - T - - - - - ~ - - - --I' 1 I I I \ I I

i 2' 4 4 2\' 1 'I I \ I I \ I I1 J: I \ , ....- - - - - -, - - - - - ,. - - 'f - - - - - - -,- - - - - I - - - - - T - - - - - := = = ;: ;: =(- - - - ;:I I I I I '",-~--__ I 1 I I I 7 I 4 ' 1 '//I <, ' ': I I I I ~.

- - - - _1- .............__ T - J. - - _ - _ ~ - __ - _1 .J l. .,..,:: _

, ' " \ ' I , I I -: 1, c:,\,: 1 1 19 ' 14 6 ~./2, ' \1 I I 1 .........., 1

- - - - - ""1 - - - - - .J. - - - - '1'''' - - - - _t_ - - - - "1 - - - - - .1 - - .f - - t- - - - - _1- _I ' \'j\. I I 1 ·..., 1 , ...

I : 6 }'27 : 39 , 2&< 10: 3 ~~·/·1I ' '\ '" I.......... I , ..... ·· ... ··1

- - - - -,- - _ - _ ~ _ - - - - T -\~ - ~.,.~:.:._-: ...-: ...-~ - - - - ~ - - - _...: ....'&-.~.C - - - ~ - - - - -

: 17 : &~ 73: 4JL~~"~'r3:- - - - - -:- - - - - ~ - - - - - ~ - - - -~~----~~ ...:;- - - ~ - - - - - ~ - - - - -!- - - - -

, 3 ' 49 : 121: 99 40' 10 2 I• I I I I_____ , .J .1. , , J • , _,I I I I '

94 '41 10'

i r

3

19 86 2

2 5 6 11

SEC3 74 8 9 10

Zero crossing period

Figure 4.

127

'I

I

I

128

P2420/G

90-06-28

IKu

Pr-o ]

Dot.ePIC N plat afMasa-Vards

Turku New Shipyard

Sign

fuel cansumptian

Service effects Included

Lz:-,z:0t- 0.13

Q)

oCo

.._)0.12til

0

LQ)Q..

til 0.11C0U

Q)

::su,

0.1

r : : :[: : : 1: : : f : : : (: :r::!:::r : : ,: . .i:::r : : i:..1: : : [ :: :i:: :1I" - - -1- ~ - I - - - t- - - -1- - - "'1 - - - T - - -,- - - -I - - - T - - - ,- - - -1- - - 'I - - ~ r- - - -,- - -I I I I I I I I I I I I I I Ir I I I I I I I I I I I I I I II I I I I I r I I I I I I I I I~- - _:_- - ~- - - ~- - _:_- - ~- - - {-- - - ~ - - _:- - - ~- - - ~ - - _:- - - ~- - - ~_?Jj ,stfort:>I I I I I I r I I I I I I I

t I r I I I ~-""TI'r---"""'''--I I I I I I

~ - - -:- - - ~ - - - I" - - -:- - - ~ - - -

0.09

~ I ~__ -

, ' ,,, ' ,, I ,

r---:----1---I

___~--J--_! ~__I I I I,

," II I I I I I I

r- - - - 1- __ , - - _ L - - - 1- - - _I - - - T __ -

I J I I II

I I ,, I II - - -,- - - ., - - -

I

I I I I Ir---~--,---~---r--~---I I r I

II

, I- - -,- - - -, - - - T - - - ,- - -

I I I I

0.08

, I I

L : _J T __

,

0.07 - - -1- __

I

- - -1- - -

I

0.06I I I I

___ ~--J---~---L--~---I I

90 100 160

RPM110 120 130

Rate of rotation of140 1S0

prope l J er-

Figure 5.

-129

ACKNOWLEDGEMENTS

The author would like to thank all colleaguesat Turku shipyard and the staff of NAPA Oy fortheir diligent working for better softwaresystems, as well as in helping to prepare thispaper.

REFERENCES

[lJ Saarilahti, J., Compute~ P~ograms in ShipDesign, Shipping World and Shipbuilder(April 1973)

[2J Linnainmaa, S. and Priha,!., IntegratedKnowledge based Systems with an Ap-plication in Ship Design, Proc. STeP-88,Vol.1, Finnish AI Society, Helsinki,(1988)

[3J Stanley, S.M and Anderson, D.C.,FunctionaL specification for CADdatabases, Computer-Aided-Design, Vol 18number 3 (April 1986)