19 views

Uploaded by Rogerio Guahy

Engenharia aeroespacial

save

- Wind Tunnel Testing Aerospace
- 7.AIAA.Goldschmied.78
- Huang Paper
- bannwart2013
- Sr Cosmos
- Expt1a
- Aerospace
- CAtxtChap1.pdf
- Lee
- AGARD-R-783
- ARL-TR-2850
- Advanced Marine Hydrodynamics 1
- introcfd.pdf
- EAEC 2011 Presentation - Ruben Jacob D
- CFX13_03_Physics1
- Physics 1.2 kinematics or mechanics
- MCQ fluid mechanics
- CFD Modelling of Entropy Generation in Turbulent Pipe Flow
- sun-gang
- Mechanics of Fluids - Lecture Notes, Study Materials and Important questions answers
- Physics for You - December 2017
- Lab10 Complete
- A Practical Guide to Wig Models
- Chapter 3
- PTA_Fekete_PTA_Course_(Mattar_2004)well test interpretation.pdf
- Applied Cfd Hyun
- IRJIE::Three Dimensional Non-Hydrostatic Free-Surface Flow Simulation Over Curved and Sharp Crested Weirs
- Revised B.tech Civil Engg. 2013 dfg fg dfg dfgdfgd f dfgdfg dg
- Rr210801 Fluid Mechanics
- Aerodynamic Analysis and Drag Coefficient Evaluation of Time-Tria
- Ada 002918
- Analise tecnica e fundamentalista.pdf
- A teoria da perspectiva e as mudanças de preferência .pdf
- 19980228064
- 19700033422
- Ada 294980
- 20030065862
- 19930081443
- (Aerospace Series) Denis Howe-Aircraft conceptual design synthesis-Professional Engineering Pub (2005) (1).pdf
- ADA294980.pdf
- - Marine CFD 2005 - 4th International Conference on Marine Hydrodynamics (2005, The Royal Institution of Naval Architects)
- Ada 294979
- Design of a Wing Section in Ground Effect_ Application to High Sp
- 19930091333
- Como Acumular Riqueza Nas Fases Da Vida GuiaInvest
- teoria dos jogos 2.pdf
- 222399-Analise-Fundamentalismo-Basico.doc
- Embarcações de alta veloc.pdf
- SALVAMENTO E COMBATE A INCÊNDIO EM AERONAVES .pdf
- [Ebook]_Guia_de_Sucesso_na_Bolsa_de_Valores.pdf
- Theory and Design of Air Cushion Craft[Yun].pdf
- nomenclaturanaval.pdf
- Wave resistance for high-speed catamarans.pdf
- Groun Effect on a lift generating body[KOHFH].pdf
- 2014 SSCI Goldminer
- Design of a Hoverwing Aircraft .pdf
- Distribuiçao de sustentaçao_arrasto_efeito_solo.pdf
- Ground Effect Aerodynamics[Cui_Zhang].pdf
- WIG Guidance
- 5th Sem Syllabus & Scheme 2014
- Chapter 3, CFD Theory
- GM5EI-2011-CHARLE-Mémoire
- 1172
- Apsley-Turbulence modelling in CFD
- vol1no3-7_RanjithG
- 5-Mylonas Keel Paper Revised
- Terziev Etal JFS 2018 Behaviour and Performance of Ships Advancing Through Restricted Shallow Waters
- Numerical Model Fsw
- Solution Algorithm
- Drag Coefficients
- 155992.pdf
- on the Influence of Viscosity on Esp Performance - Valdir Estevam
- 2_08_osullivan
- HyperMesh and BatchMesher
- 83_10_SinghSN
- Irrigation in tics Haapasalo Et Al 2010-1
- wu2014
- CFD ANALYSIS.pdf
- 455984.fucak_fp
- cr020718s
- openfoamrotatinggeometryenglish20140518-140517222323-phpapp01.pdf
- Sathesh Prof. Thesis
- Robert C. Ripley- Cylindrical Explosive Dispersal of Metal Particles: Predictive Calculations in Support of Experimental Trials
- The Design Process of an Aerodynamic Package for an FSae Car
- Contam User Guide and User Documentation Version 3.2
- COSMOSFloWorks Introduction
- Flow-Science-Brochure-Aerospace-WEB.pdf
- Natural Ventilation in Buildings
- hsl0679.pdf

You are on page 1of 366

AC/323(AVT)TP/9

NORTH ATLANTIC TREATY ORGANIZATION

10

Q.

6i-

**RESEARCH AND TECHNOLOGY ORGANIZATION
**

BP 25, 7 RUE ANCELLE, F-92201 NEUILLY-SUR-SEINE CEDEX, FRANCE

RTO MEETING PROCEEDINGS 15

**Fluid Dynamics Problems of Vehicles
**

Operating Near or in the Air-Sea Interface

(Problemes de dynamique des fluides des vehicules evoluant

dans ou pres de 1'interface air-mer)

**Papers presented and discussions recorded at the RTO Applied Vehicle Technology Panel (AVT)
**

Symposium (organised by the former AGARD Fluid Dynamics Panel), held in Amsterdam, The

Netherlands, 5-8 October 1998.

m

DISTRIBUTION STATEMENT A:

Approved for Public Release -

19990315 026

Distribution Unlimited

Published February 1999

Distribution and Availability on Back Cover

**The Research and Technology
**

Organization (RTO) of NATO

RTO is the single focus in NATO for Defence Research and Technology activities. Its mission is to conduct and promote

cooperative research and information exchange. The objective is to support the development and effective use of national

defence research and technology and to meet the military needs of the Alliance, to maintain a technological lead, and to

provide advice to NATO and national decision makers. The RTO performs its mission with the support of an extensive

network of national experts. It also ensures effective coordination with other NATO bodies involved in R&T activities.

RTO reports both to the Military Committee of NATO and to the Conference of National Armament Directors. It comprises a

Research and Technology Board (RTB) as the highest level of national representation and the Research and Technology

Agency (RTA), a dedicated staff with its headquarters in Neuilly, near Paris, France. In order to facilitate contacts with the

military users and other NATO activities, a small part of the RTA staff is located in NATO Headquarters in Brussels. The

Brussels staff also coordinates RTO's cooperation with nations in Middle and Eastern Europe, to which RTO attaches

particular importance especially as working together in the field of research is one of the more promising areas of initial

cooperation.

The total spectrum of R&T activities is covered by 6 Panels, dealing with:

• SAS Studies, Analysis and Simulation

• SCI Systems Concepts and Integration

• SET Sensors and Electronics Technology

• 1ST Information Systems Technology

• AVT Applied Vehicle Technology

• HFM Human Factors and Medicine

These Panels are made up of national representatives as well as generally recognised 'world class' scientists. The Panels also

provide a communication link to military users and other NATO bodies. RTO's scientific and technological work is carried

out by Technical Teams, created for specific activities and with a specific duration. Such Technical Teams can organise

workshops, symposia, field trials, lecture series and training courses. An important function of these Technical Teams is to

ensure the continuity of the expert networks.

RTO builds upon earlier cooperation in defence research and technology as set-up under the Advisory Group for Aerospace

Research and Development (AGARD) and the Defence Research Group (DRG). AGARD and the DRG share common roots

in that they were both established at the initiative of Dr Theodore von Kärmän, a leading aerospace scientist, who early on

recognised the importance of scientific support for the Allied Armed Forces. RTO is capitalising on these common roots in

order to provide the Alliance and the NATO nations with a strong scientific and technological basis that will guarantee a solid

base for the future.

**The content of this publication has been reproduced
**

directly from material supplied by RTO or the authors.

®

Printed on recycled paper

Published February 1999

**Copyright © RTO/NAT01999
**

All Rights Reserved

ISBN 92-837-0004-X

**Printed by Canada Communication Group Inc.
**

(A St. Joseph Corporation Company)

45 Sacre-Cceur Blvd., Hull (Quebec), Canada K1A 0S7

$Q f99-06-/PS-

**Fluid Dynamics Problems of Vehicles Operating
**

Near or in the Air-Sea Interface

(RTO MP-15)

Executive Summary

This Symposium concentrated on issues associated with vehicles operating near the sea surface, issues

common to aeronautics and hydrodynamics, issues involving air-sea interactions; and on providing the

opportunity of bringing together scientists and engineers from Western Europe, North America,

Australia, Russia, and Ukraine.

One-third of the papers were authored by scientists from Russia and Ukraine. The two major topics

covered during the Symposium included: Aerodynamics around Ships, and Non-Classical Aircraft

flying near the air-sea interface i.e. Ekranoplanes. In addition, there were papers on Surface Effect

Ships, Ship Stabilization, Hydrofoil Boats, Underwater Missile Launch, and Ship Bow Waves and

Water Entry.

Aerodynamics around ships is militarily important because ship superstructures create highly unsteady

3-D flows with massive regions of flow separation behind them, and this situation can create severe

problems for the landing of helicopters on the deck. It was very apparent from the presented papers that

current means of calculating these flows are developing, but they are currently very limited. Many

research opportunities exist involving unstable and separated flows and their simulation.

Several papers covered ground effect aerodynamics, specifically Russian technology in this area, as

evidenced by the development of Ekranoplanes. By utilizing ground effect, the gap between slow and

inexpensive ships and fast but expensive aircraft can possibly be filled. However, economic studies,

which include sea state operability, need to be performed. An innovative means of reducing drag during

take-off and thus enabling a reduction in engine power requirements was presented. There were

proposals to use Ekranoplanes in a grand air/sea rescue system, and also a proposal for a future

Symposium on "Marine Vehicles for Rapid Disaster Response".

For vehicles operating in the air-sea interface, load control and vertical control are very difficult and

extraordinarily important. It is necessary to keep vertical accelerations very small, under 1/1 Oth of a G

and, therefore, expectations of 1/4 of a G for some Surface Effect Ships and Hydrofoils present a very

serious problem. A new implementation of an automated moving weight scheme together with

sophisticated fins, which reduced ship motions on an Air Craft Carrier, while ensuring navigational

capability, was presented.

The material assembled in this report was prepared under the combined sponsorship of the NATO

Partnership for Peace Program, the RTO Applied Vehicles and Technology Panel, the United States

Office of Naval Research International Field Office - Europe, and the United States Air Force European

Office of Aerospace Research and Development.

**Problemes de dynamique des fluides
**

des vehicules evoluant dans ou pres

de Pinterface air-mer

(RTO MP-15)

Synthese

Ce symposium a porte essentiellement sur l'exploitation de vehicules evoluant ä proximite de la surface

de la mer. II s'agit de questions relevant ä la fois de l'aeronautique et de l'hydrodynamique, avec des

interactions air-mer. Des scientifiques et des ingenieurs de l'Europe occidentale, de FAmerique du

nord, de l'Australie, de la Russie et de l'Ukraine ont pu etre reunis ä cette occasion.

Un tiers des communications a ete presente par des scientifiques russes et ukrainiens. Les deux

principaux sujets couverts par le symposium comprenaient: l'aerodynamique ä proximite des navires et

les aeronefs non-conventionnels evoluant ä proximite de la surface de la mer, c'est ä dire les

ekranoplanes. En outre, des communications ont ete presentees sur les navires ä effet de surface, sur la

stabilisation des navires, sur les navires hydropteres, le lancement de missiles ä partir de sous-marins,

les vagues d'etrave et la penetration dans l'eau.

Les conditions aerodynamiques ä proximite des navires sont d'un grand interet militaire, car les

superstructures creent, en aval, des ecoulements tridimensionnels fortement instationnaires avec des

zones de decollement tres importantes, ce qui pose parfois de serieux problemes pour l'appontage des

helicopteres. Des communications presentees, il est apparu tres clairement que des methodes de calcul

de ces ecoulements sont en cours de developpement mais restent tres limitees. De nombreuses

possibilites de recherche existent, centrees autour des ecoulements instationnaires et decolies et autour

de leur simulation.

L'aerodynamique des effets de sol a fait l'objet de nombreuses communications. La technologie russe a

bien ete mise en evidence avec le developpement des ekranoplanes. Par le biais des effets de sol, il se

pourrait que l'ecart qui existe entre les navires lents mais peu coüteux, et les avions rapides mais chers,

puisse etre elimine. Cependant, cela impliquerait la realisation d'etudes economiques, entre autres sur la

tenue ä la mer. Un moyen novateur de reduire la trainee au decollage, permettant ainsi de reduire les

besoins en propulsion, a ete presente. Des projets ont egalement ete proposes visant l'utilisation

d'ekranoplanes dans un vaste Systeme de sauvetage en mer, ainsi qu'une proposition de symposium sur

«les vehicules maritimes pour une reaction rapide en situation de catastrophe ».

Pour les vehicules ä proximite de la surface de la mer, le contröle de la charge et le contröle vertical

sont extremement difficiles et importants. II faut reduire les accelerations verticales au minimum

(moins d'un dixieme de g) et, par consequent, les previsions d'un quart de g, pour certains navires et

hydropteres ä effets de surface, posent un probleme serieux. Une nouvelle configuration a ete presentee,

comprenant un Systeme automatise de poids amovible, comportant des derives sophistiquees, qui

permet de reduire les mouvements d'un porte-avions, tout en conservant 1'aptitude ä la navigation.

Ce rapport a ete elabore sous l'egide conjointe du Programme OTAN de Partenariat pour la paix, de la

Commission RTO de technologies appliquees aux vehicules, du bureau exterieur international - Europe,

du Directorat des Etats-Unis de la recherche navale, et du Bureau europeen de recherche et

developpement aerospatial de l'armee de Fair americaine.

Contents

Page

Executive Summary iii

Synthese iv

Recent Publications of the Former Fluid Dynamics Panel vii

Applied Vehicle Technology Panel Programme Committee ix

Reference

**Technical Evaluation Report T
**

by M.P. Tulin

**Recent Developments in the Analytical Investigation of Shipboard Rotorcraft Engage and 1
**

Disengage Operations

by E.C. Smith, J.A. Keller and H. Kang

**CFD Predictions of the Influence of External Airflow on Helicopter Operations when Operating 2
**

from Ship Flight Decks

by N.H. Wakefield, S.J. Newman and P.A. Wilson

**Higher Order Accurate Solutions of Ship Airwake Flow Fields Using Parallel Computers 3
**

by L.N. Long, J. Liu and A.V. Modi

**Simulation and Analysis of LHD Ship Airwake by Navier-Stokes Method 4
**

by T.C. Tai

**Prediction of Ship Air Wakes Over Flight Decks using CFD 5
**

by P. Tattersall, CM. Albone, M.M. Soliman and C.B. Allen

**Aerodynamic Perturbations Encountered by a Helicopter Landing on a Ship - Effects on the 6
**

Helicopter Flight Dynamics

by A. Taghizad, Ch. Verbeke and A. Desopper

**Analysis of Patrol Frigate Air Wakes 7
**

by S.J. Zan, G.F. Syms and B.T. Cheney

**Modelling and Simulation of Ship Air Wakes for Helicopter Operations - A Collaborative 8
**

Venture

by CH. Wilkinson, S.J. Zan, N.E. Gilbert and J.D. Funk

**The Experience of Aerodynamic Disturbances Research behind an Aircraft-Carrier Ship with 9
**

Elements for Safe Operation of Ship-Based Aircraft

by L.A. Maslov, N.O. Valuev and A.V. Zharinov

**Simulation Tools in the Calculation of Aircraft-Ship Interface Operational Limits 10
**

(Application des outils de simulation pour le calcul des limites operationnelles de Pinterface

dynamique aeronef-navire)

by B. de Ferrier and B. Langlois

**Piloting of a VTOL-UAV to Shipboard Recovery 11
**

by C. Reboulet, P. Mouyon, B. de Ferrier and B. Langlois

**Motions and Added Resistance Due to Waves of Surface Effect Ships 12
**

by J.C Moulijn

**Cobblestone Effect on SES 13
**

by T. Ulstein and O.M. Faltinsen

**Heel Compensation for the Charles de Gaulle Aircraft Carrier: Principles and Control 14
**

Structure

by S. Kummer, G. Hardier and C. Lambert

**New Advances in Sailing Hydrofoils 15
**

by F. Lefaudeux

**Unsteady Flow Around a Hydrofoil Beneath Waves 16
**

by U.P. Bulgarelli, M. Greco, M. Landrini and C. Lugni

**Hydrodynamic Characteristics of Rudders Operating in Air-Sea Interface 17
**

by V.T. Savchenko

**Hydrodynamical Characteristics of an Ekranoplane Wing Flying Near the Wavy Sea Surface 18
**

by V.G. Byelinskyy and P.I. Zinchuk

**Theoretical Analysis of Dynamics of a Wig Vehicle and Extreme Ground Effect 19
**

by K.V. Rozhdestvensky

**Longitudinal Stability of Ekranoplans and Hydrofoils Ships 20
**

by V.l. Korolyov

Aerodynamic Scheme of Ekranoplane Optimization with Reference to New Areas of Application 21

by A.V. Nebylov and E.Th. Zhigalko

**Conceptual Bases of WIG Craft Building: Ideas, Reality and Outlooks 22
**

by E.A. Aframeev

Progress Report on Aerodynamic Analysis of a Surface Piercing Hydrofoil-Controlled Wing-In- 23

Ground Effect SEABUS Configuration

by CM. van Beek, B. Oskam and G. Fantacci

**The First Commercial Ekranoplan "Amphistar" and Prospects for the Development of 24
**

Passenger Ekranoplans

by D.N. Sinitsyn and A.I. Maskalik

**A View of the Present State of Research in Aero- and Hydrodynamics of Ekranoplans 25
**

by A.I. Maskalik, K.V. Rozhdestvensky and D.N. Sinitsyn

**On the Prediction of Nonlinear Free-Surface Flows Past Slender Hulls Using 2D+t Theory: 26
**

The Evolution of an Idea

by E. Fontaine and M.P. Tulin

**The Complex Boundary Integral Equation Method for a Problem of Entry of a 2D Solid Body 27
**

in an Incompressible Liquid

by D.I. Cherniy

**Evolution de la bulle de culot et jets rentrants derriere un projectile lance en immersion 28
**

by J.B. Paquet, J.P. Flodrops and A. Dyment

**Undisturbed Motion of Vehicles in the Fluid 29
**

by Yu.N. Savchenko

**The Hoverwing Technology-Bridge between WIG and ACV 30
**

by H. Fischer and K. Matjasic

General Discussion GD

vi

**Recent Publications of the Former
**

AGARD Fluid Dynamics Panel

AGARDOGRAPHS (AG)

Turbulent Boundary Layers in Subsonic and Supersonic Flow

AGARD AG-335, July 1996

Computational Aerodynamics Based on the Euler Equations

AGARD AG-325, September 1994

Scale Effects on Aircraft and Weapon Aerodynamics

AGARD AG-323 (E), July 1994

Design and Testing of High-Performance Parachutes

AGARD AG-319, November 1991

Experimental Techniques in the Field of Low Density Aerodynamics

AGARD AG-318 (E), April 1991

Techniques Experimentales Liees ä l'Aerodynamique ä Basse Densite

AGARD AG-318 (FR), April 1990

A Survey of Measurements and Measuring Techniques in Rapidly Distorted Compressible Turbulent Boundary Layers

AGARD AG-315, May 1989

REPORTS (R)

Fluid Dynamic Research on Supersonic Aircraft

RTO Report EN-4, November 1998

Hugh Speed Body Motion in Water

AGARD R-827, February 1998

Turbulence in Compressible Flows

AGARD R-819, Special Course Notes, June 1997

Advances in Cryogenic Wind Tunnel Technology

AGARD R-812, Special Course Notes, January 1997

Aerothermodynamics and Propulsion Integration for Hypersonic Vehicles

AGARD R-813, Special Course Notes, October 1996

Parallel Computing in CFD

AGARD R-807, Special Course Notes, October 1995

Optimum Design Methods for Aerodynamics

AGARD R-803, Special Course Notes, November 1994

Missile Aerodynamics

AGARD R-804, Special Course Notes, May 1994

Progress in Transition Modelling

AGARD R-793, Special Course Notes, April 1994

Shock-Wave/Boundary-Layer Interactions in Supersonic and Hypersonic Flows

AGARD R-792, Special Course Notes, August 1993

Unstructured Grid Methods for Advection Dominated Flows

AGARD R-787, Special Course Notes, May 1992

Skin Friction Drag Reduction

AGARD R-786, Special Course Notes, March 1992

Engineering Methods in Aerodynamic Analysis and Design of Aircraft

AGARD R-783, Special Course Notes, January 1992

ADVISORY REPORTS (AR)

A Selection of Test Cases for the Validation of Large-Eddy Simulations of Turbulent Flows

AGARD AR-345, April 1998

Ice Accretion Simulation

AGARD AR-344, Report of WG-20, December 1997

Sonic Nozzles for Mass Flow Measurement and Reference Nozzles for Thrust Verification

AGARD AR-321, Report of WG-19, June 1997

Cooperative Programme on Dynamic Wind Tunnel Experiments for Manoeuvring Aircraft

AGARD AR-305, Report of WG-16, October 1996

Hypersonic Experimental and Computational Capability, Improvement and Validation

AGARD AR-319, Vol. I, Report of WG-18, May 1996

Aerodynamics of 3-D Aircraft Afterbodies

AGARD AR-318, Report of WG-17, September 1995

April 1990 Computational Methods for Aerodynamic Design (Inverse) and Optimization AGARD CP-463. December 1991 Manoeuvring Aerodynamics AGARD CP-497. July 1991 Rotary-Balance Testing for Aircraft Dynamics AGARD AR-265. Report of WG10. April 1993 Aerodynamic Engine/Airframe Integration for High Performance Aircraft and Missiles AGARD CP-498. September 1993 Theoretical and Experimental Methods in Hypersonic Flows AGARD CP-514. February 1996 Aerodynamics and Aeroacoustics of Rotorcraft AGARD CP-552. Vols. March 1990 Applications of Mesh Generation to Complex 3-D Configurations AGARD CP-464. Support Interference. Report of WG-15. May 1990 CONFERENCE PROCEEDINGS (CP) Missile Aerodynamics RTO Report MP-5. October 1990 Aerodynamics of Combat Aircraft Controls and of Ground Effects AGARD CP-465. July 1994 Computational and Experimental Assessment of Jets in Cross Flow AGARD CP-534. December 1994 Wall Interference. Report of WG-13. May 1998 Aerodynamics of Wind Tunnel Circuits and Their Components AGARD CP-585. November 1991 Vortex Flow Aerodynamics AGARD CP-494.A Selection of Experimental Test Cases for the Validation of CFD Codes AGARD AR-303. April 1996 Aerodynamics of Store Integration and Separation AGARD CP-570. November 1996 Progress and Challenges in CFD Methods and Algorithms AGARD CP-578. November 1993 High-Lift System Aerodynamics AGARD CP-515. March 1990 Fluid Dynamics of Three-Dimensional Turbulent Shear Flows and Transition AGARD CP-438. September 1992 Effects of Adverse Weather on Aerodynamics AGARD CP-496. and Flow Field Measurements AGARD CP-535. August 1995 Application of Direct and Large Eddy Simulation to Transition and Turbulence AGARD CP-551. June 1997 The Characterization & Modification of Wakes from Lifting Vehicles in Fluids AGARD CP-584. April 1989 . I and II. November 1998 Advanced Aerodynamic Measurement Technology AGARD CP-601. July 1994 Air Intakes of High Speed Vehicles AGARD AR-270. Technical Status Review. Report of WG-14. September 1991 Appraisal of the Suitability of Turbulence Models in Flow Calculations AGARD AR-291. July 1991 Missile Aerodynamics AGARD CP-493. Report of WG11. August 1994 Quality Assessment for Wind Tunnel Testing AGARD AR-304. December 1990 Calculation of 3D Separated Turbulent Flows in Boundary Layer Limit AGARD AR-255.

PURTELL (Deputy Chairman) Program Officer .T. Jan L. Javier JIMENEZ Universite d'Orleans Escuela Technica Superior de 6. Dr. Dr.Germany Middle East Technical University Inono Bulvari PK: 06531 Assoc. Prof. Ir.O.O. Dr. Department D-38108 Braunschweig . Box 3267 D-38022 BRAUNSCHWEIG . of Mechanical Engineering National Technical University of Athens Mr. Patrick. Richard J. Dr. R. Cahit CIRAY Lilienthalplatz 7 Aeronautical Eng. LOVELL P.Office of Naval Research Delft University of Technology 800 North Quincy Street Kluyverweg 1 Arlington.France Departamento de Mecanica de Fluidos Plaza del Cardenal Cisneros 3 Dr. Ing. Engenharia Mecanica Carleton University Instituto Superior Tecnico Ottawa./NATO BP 25.Canada 1096 Lisboa Codex . DECUYPERE Prof. Horst KÖRNER Direktor Institut für Entwurfsaerodynamik des DLR e. Box 64070 Research Manager Applied Aerodynam. C. GOLIA Defence Research Agency CIRA Farnborough Hants GUI4 6TD .GERMANY PROGRAMME COMMITTEE Prof.Turkey Dept.Norway Prof. Rue Ancelle UnitPSC 116 F-92201 Neuilly-sur-Seine Cedex APO AE 09777 France Tel: 33 (1) 55 61 22 75 Fax: 33 (1) 55 61 22 98 .NTNU . rue Eudoxe Marcille Ingenieros Aeronauticos 45000 Orleans . Dr. A. van INGEN Energy Conversion Division Dept. FALCAO Aerospace Engineering Depart.T.United States 2629 HS Delft . MASURE (Chairman) Prof.Greece Aero/Structures Dept X80 Building Prof. Prof.Netherlands PANEL EXECUTIVE Mr. R.Spain Direktor Institut fur Entwurfsaerodynamik der DLR e. VA 22217-5660 . S. P. MOLLOY Mail from Europe: Mail from USA and Canada: R.K. Avenue de la Renaissance The Norwegian University of B-1000 Brussels-Belgium Science and Technology N-7034 Trondheim . Ontario K1S 5B6 .Italy Dr. 15710 Zografu-Athens . de O.Mechanics and Prof. of Aerospace Engineering Code 333 . KIND Department of Mechanical and Prof. V. V. Tor YTREHUS Ecole Royale Militaire Division of Applied Mechanics 30. Applied Vehicle Technology Panel CHAIRMAN: Prof. L. 7. TSANGARIS Ankara .F.A.Portugal IGA B. David A. J.United Kingdom Via Maiorise 81043 CAPUA (CE) . Horst KÖRNER 28040 Madrid .A.

" meeting was most unusual. Tulin Director. In this respect the with vehicles operating near the sea surface. were: Moscow (TsAGI). papers on a variety of important and even remarkable development of other technologies lightly covered (C): Ekranoplane (ground effect) vehicles. and his other organizers and the spanned the interface itself. "con.edu 1 Introduction The meeting brought together about 125 engi- neers and scientists from 16 countries. and about this difficult fusion as they have in such a which are therefore. and the Chairman. were chosen: (A) . in addition. T-l Technical Evaluation Report Applied Vehicle Technology Panel Symposium on Fluid Dynamics Problems of Vehicles Operating Near or in the Air-Sea Interface Professor Marshall P. permitting a good eval- tion of Russian engineers and scientists from uation. organized by an international rope plus the US and Canada. by conferences on either aeronautics or naval hy- An important and significant feature of this drodynamics. Aerodynam- • Hydrofoil boats (2 papers) ics meets Hydrodynamics at the Sea Interface.e. • Ship bow waves. as from aeronautical research almost equally di- centrated on fluid dynamic issues associated vided between the two fields. can be summed up as. and St. the Russian air-sea interface. Nizhny-Novgorod (Technolo- (A) Aerodynamics around ships (10 papers) gies Transport JSC). and (B) Non-classical aircraft flying near the Ukranians from Kiev. In particular. Ocean Engineering Laboratory University of California Santa Barbara." Specific well-covered examples of Symposium was the very active participa- the latter technologies. "not traditionally covered pleasant and effective venue. the organiz- ers are to be thanked for addressing this theme. and water entry (2) It is to be understood that the two major • Miscellaneous (2) topics covered A & B above. Petersburg. i. There was an emphasis on technologies which B. • Surface effect ships (2 papers) The general theme of this symposium. Dutch hosts are to be congratulated for bringing dynamic and hydrodynamic phenomena. Masure. 14 in Eu- This Symposium. in ad- dition to the statements made in the first para- • Ship stabilization (2 papers) graph above. Ekranoplanes(ll papers) participation allowed a very good view of their There were. There was a very program committee under the Chairmanship of good representation from ship research as well IG A Bernard Masure (Orleans. In view of the military and commercial impor- • Underwater missile launch (1) tance of operation at the interface. combining aero. CA93106-1080 USA email: mpt@vortex.ucsb. Prance).

a section on hydrody. This paper is extremely im. including a Summary of have been conducted in the US. stated in Paper # 9. "An air wake model of ap- tailed aerodynamic problems. Russian (TsAGI) work on the earlier. The shape of the bluff body features above the deck. It is clear that like frigates. mensional character. and because of the complexity of these flows create problems for helicopter oper- the aerodynamics behind the ships structure ations. as in the paper on underwater missile noteworthy.. and present status of the sub.. especially from small ships like frigates. Australia and the USA. and particularly its superstruc. which is usually located behind the superstruc- cially in Russia. it is remarkable how often for helicopter operations." From a longer range point It seemed entirely appropriate to include to. launch or on hydrofoil boats. it is clear that effective simulation of gether with the mainly aerodynamic consider. a good view of the evolution of ture.an essential element of a to stability and control. As thinking and techniques. So a wind tun- the history. be derived in pricipal from either full scale or der (C) do not permit a reasonable evaluation wind tunnel observations and measurements. Canada. ment in the wake characteristics through careful namic aspects with a close connection to aero. the forward end of aircraft carrier decks is very niques. Full scale measurements are difficult and costly to obtain and the conditions 2 Aerodynamics Around cannot be repeated or called up at will. role. a particularly noisy place to operate. tanking area a large separated flow region forms . The Scope of technical coverage structures like drilling platforms. espe. from computer simulations (CFD). limited model-scale tests " are also conducted on A very good reportage from participants in the each new ship entering service. Problems must also arise in tial of wind in ground effect vehicles. behind the last large deck structure before the ture leads to highly unsteady flows character. training simulation. broad and the presenters sufficiently expert to The training of military helicopter pilots permit in the case of (A) an assessment of the through the use of flight simulators requires an overall problem and an understanding of the accurate simulation of the atmospheric environ- complex fluid dynamics involved and the state ment and this requires effective modeling of the of present research. and in the case of (B) an ship's wake and especially near the landing deck. of view. These originate from all of and by winds near the sea surface. In nautics. Research in this field this kind of difficulty arises during field work at is largely carried out as part of an international sea. especially related propriate fidelity is. Each poses its own difficulties. these complex flows could also lead to improve- ations mentioned above. it was re- Ships counted in Paper # 9 that Australian attempts at full scale measurements in 1989 on a frigate This main sector of the meeting has focussed on failed because of "benign" wind conditions dur- the modelling and simulation of ship air wakes ing much of the test. alterations in the design of superstructures. ized by massive regions of flow separation and ing of helicopters and small naval combatants large unsteady vortical patterns. These were included in (C) and listed this connection. volve unstable shear regions leading to large un- The flow around ships is induced both by their steady vortical structures of distinctly three di- forward motion. Many of these have a clear and imme. It is apparent that An air wake model for simulator use could the lightness of the coverage of the items un. nel data base is growing. and a fair view of de.T-2 because of the military importance of the mat. it is discussed later here. and (B) because of the and they do in fact limit helicopter operations military and commercial interest in the poten. overview of actual vehicle development. approaching the deck. their seaway induced motions. the ship itself. Extensive wind tunnel experiments have cooperative program (TTCP) involving four na. so-called heavy wind operations from commercial ocean Ekranoplanes. or of those subjects. full scale installation of vortex control devices at diate connection with aeronautics and its tech. A variety of serious efforts have been made at portant for anyone interested in this subject at numerical simulations of these flows." Further tests TTCP was presented. which in- this time. in stormy weather. Finally. been conducted in the UK and in Canada and tions: the UK. ject (Paper #9). although these of these two subjects was certainly sufficiently were not discussed at this Symposium.

T-3

**of finite length, no doubt partially stabilized by which must be imposed. Calculated veloc-
**

the presence of the horizontal landing deck sur- ity perturbations around a deck structure were

face. The simulation is complicated by uncer- shown, but no comparisons with experiments

tainties in the wind field over the ocean, which were made; in these calculations; the viscous

is unsteady and effected by atmospheric strati- driving terms were neglected. This approach

fication. A further complication arises from un- to unsteady simulation seems worthwhile, al-

certainties due to the motion of the ship in a though the influence of the initial disturbances

seaway. is very worrisome. In these calculations, wind

Finally, the action of the helicopter rotor it- turbulence was used to drive the unsteady deck

self can have a large influence on the deck flow flow. However, it seems highly unlikely that the

and vice versa, as the rotor induces substantial deck flow unsteadiness is, in essence, controlled

velocities directed downwards over an area often by wind turbulence. As in the Karman vortex

not very much larger than the helicopter. Most strut, it seems likely that an inherent instabil-

of the simulations were made in the absence of ity of the shear flows produced by separation

the rotor flow and these are first discussed be- from the deck structure reaches some equilib-

low. rium quasi-oscillatory state through the balance

How effective is the state of the art in CFD of non-linear processes, with little effect from

simulation for these turbulent flows? Five pa- small ambient disturbances.

pers reflect on this question (#2, 3, 5, 6, 8 One of the most interesting approaches to un-

and 9). A major limitation at this time is that steady separated flow calculation, paper #6, in-

current RANS codes (Reynolds averaged Navier volved the use of the incompressible Euler equa-

Stokes) are constituted to handle steady, i.e. tions, allowing for the distortion and propaga-

mean flows only. How well do these RANS code tion of vorticity fields, once created, in the ab-

simulate mean flows? How well do they compare sence of viscosity. In this connection we can

with each other? The latter question cannot think of the simulation of vortex sheet roll-up

be answered at all since comparsions were not by inviscid theory, or of von Karman's predic-

made, although similar separated region topol- tion of vortex street geometry by an inviscid

ogy can be seen among several papers. The for- analysis of stability. In the present case, the

mer question can be answered only roughly. In question arises as to the origin of the vorticity

#3, oil flow streaklines on deck are compared which appears behind the deck structures. In

with flow vector simulations in a qualitatively fa- the absence of viscous boundary layers here, the

vorable way. In #8 favorable comparisons of the vorticity which presumably appears here in the

topology of the aft deck separation region are form of shear layers at separation points must

made. However, it was noted that the measured be an artifice of the numerical calculation, and

shear regions were much heavier than in the sim- must presumably depend upon the roundness of

ulations. It is recommended that priority the corners where separation occurs, the grid

be given to careful comparisons between size, and the numerical scheme. It is never-

the numerical results of different research theless a most interesting phenomena and one

groups and with the results of a bench- to be taken advantage of, as was done here.

mark experiment. This has proved effective Once created, the distortion and propagation

in other applications, as for instance in the cal- of the inviscid vorticity field will occur accord-

culation of ship stern flows. Finally, it is possi- ing to the Euler equations. It is not surprising

ble, as mentioned by one of the authors, that the that highly unsteady flows appear once vorticity

RANS codes are insufficient for these separated is introduced at separation points, since vortex

flows and that large eddy simulation or other sheets tend to be inherently unstable. The in-

more sophisticated approaches are necessary. teresting question occurs as to the connection

The possibility of extending the RANS mean between these unstable, inviscid, vortical fields

flow calculation to the simulation of unsteady and the real flow fields with viscosity. In sim-

components was discussed in #3. This involved ple separated wakes at high Reynolds number it

the solution of the so-called non linear distur- is known that the length of the separated flow

bance equations, (NLDE), for the unsteady flow behind the body depends on Reynolds stresses

as driven by a known mean flow (RANS). The acting on the boundary of the separation cavity,

unsteady flow depends upon initial disturbances and the success of an inviscid simulation would

T-4

**therefore depend on the inviscid simulation of 3 What Price Speed? Ekra-
**

these stresses. These questions concerning in- noplanes

viscid simulations deserve careful study in view

of the potential value which lies in this approach. The real transport efficiency of ships decreases

with increasing speed beyond some optimum

Two papers (#2, 6) showed techniques for in-

speed which is determined by the balance be-

cluding the effect of rotor action on the flow

tween fuel and fixed cost including amortization

fields and vice versa. One of these (#6) showed

of both ship and cargo. At this time, the op-

the strong effect of the rotor and helicopter fuse-

timum speed is in the range of 20-35 kts, and

lage in modifying the separation regions behind

the transport efficiencies are much higher than

the deckhouse. These studies both create the

those of modern aircraft. Of course, there is

incentive for including the helicopter action in

a premium to be paid for speed in the case of

future simulations and show the way to do this.

very valuable cargos or those whose value de-

The bottom line regarding computer creases with time. Therefore people and mail

simulations is that qualitative pictures are carried long distances by aircraft, while oil

and effects are reproduced, but the de- and automobiles, etc. are carried by large ships

gree of accuracy has not yet been estab- at less than 1/10 the speed, but at considerably

lished. It seems very likely, considering reduced cost.

the fluid dynamic complexities here, that For speed in water over about 50 kts severe

highly accurate numerical simulations lie new problems arise. These have led to alter-

some distance in the future. This stresses native vehicles beyond the simple monohull dis-

the importance of detailed wind tunnel placement ships: planing hulls, catamarans, hy-

measurements, which certainly seem fea- drofoil boats, and surface effect ships (SES). The

sible and capable of providing the infor- boldest projections of feasible speeds that could

mation needed by the training simulators. be achieved in regular practice by any of these

marine vehicles does not exceed 100 kts., and

It is to be noted that none of the helicopter- the feasible hydrodynamic efficiences do not ex-

related papers on ship wake flows were con- ceed those of a large modern aircraft; structural

cerned with modification of the deck structures weights in the water also tend to be high.

to control the flow and thereby improve the It has for at least 70 years been known, that

engage-disengage environment. Here one can the induced resistance of a wing decreases con-

think in terms of the modification of sharp cor- tinuously as it approaches the ground; the so

ners on the aft of the deck structure or of re- called ground effect. As a result, when a wing

shaping it at the rear; here automotive know- is optimized in ground effect, its aspect ratio

how would be very instructive. (span/chord) decreases, and its left coefficient

In the above connection, a very interesting increase relative to flight at altitude. This held

Russian paper from TsAGI is to be noted (#11); forth the theoretical possibility of ground effect

it concerned "provision of aerodynamic capa- wings of smaller area and span, and higher effi-

bility of ship-based aircraft and carrier," whose ciencies than conventional optimized aircraft at

"aim was design and realization of measures of the same speed.

reduction of (wind flow deck) disturbances and The gifted Russian hydrofoil boat designer,

...gradients up to level required for take off land- Rostislav Alexeev from Gorki, began in the

ing operation safety." In the first step, the ori- 1960's the development of large vehicles of 120-

gin and trajectory of large rolled up vortices 550t based in the ground effect, which he called

were determined; these largely originated from Ekranoplanes. These were, however, not de-

the lateral edge of the main deck, the sponson, veloped with commerical use foremost in mind,

and the island structure. Recommendations re- but at the order of the Russian Navy for pre-

sulted: for decrease of the island aspect ratio sumed amphibious use. The earliest open tech-

and its relative width, and to move it forward; nical publication on the subject may be a pa-

for the inclusion of flat and curved path deflec- per by P.E.Kumar on "Some Stability Problems

tors and suitable fairings on the deck and spon- of Ground Effect Vehicles in Forward Motion,"

son edges. These innovative approaches deserve Aeronautical Quarterly, vol. XVII, Feb. 1972.

detailed attention. At the present time, given the reality of a va-

T-5

**riety of Russian vehicles as well as some others of the best aircraft figure, 5, shown by Aframeev,
**

abroad, coupled with a high current interest in Fig. 8 of #29. Of course, actual values must

increasing the speed of marine vehicles, there take into account the necessity of operation in

seems a high interest in Ekranoplanes, and their reasonable seaways.

future. For example, the list of References in Do special missions exist fitting the technical

paper #30 refers to a variety of relevant confer- strength of the Ekranoplane, which is to oper-

ences in the last ten years. ate near the sea at speeds up to 400 km/h and

The papers presented here, notably but not in weights up to, perhaps, 1000 t? The use of

only the Russian contributions offer a good op- Ekranoplanes to provide a launching and land-

portunity for an evaluation, and particularly ing base for reuseable aerospace planes or Space

#32 (a review of the present state of research), Shuttles was discussed in paper #27. The re-

and #29 (an objective and critical look at quirement of high launch payload (600 t) and

technical performance and possibilities by E.A. high speed (220 m/sec) could, they suggested

Aframeev from St. Petersburg). be carried out by a large Ekranoplane.

Did the Ekranoplanes in actual practice re- A clearcut vision for the future development

alize an improvement in the aerodynamic ef- of Ekranoplanes was provided by E.A. Eframeev

ficiency (L/D) of modern aircraft? The an- from the Kylov Shipbuilding Research Institute,

swer is no improvement, only similar per- paper #29. He discounted competition with air-

formance, depending on the degrading effect craft which seem, if anything, technically supe-

of increasing sea states, Fig. 1 of #29. In rior to the Ekranoplane. He stressed the ability

some other performance aspects however the of the latter to operate in two-modes both as a

Ekranoplane is quite inferior: payload effi- ship and as an airplane. He therefore proposed

cience, thrust/weight requirements, and struc- a new direction of development in which large

tural weight of both hull and wing, see figures Ekranoplanes would be designed for full and sus-

4, 5, 8 of #29. The main reason for this infe- tained marine operation at cruising speeds up

riority lies in requirements arising from marine to, say, 15 kts, and flight speeds of, say 250 kts in

operation, for instance the necessity to augment seas up to 3.5 m significant wave height. Loaded

take-off and to provide for water loads. weight of 750 t and payload of 250-300 t with

Can present performance shortcomings be a range of 3-4000 km were projected. For this

overcome? The Russians had augmented lift for two-mode vehicle he imagines use for commer-

take-off by diverting the engine jet to stagnate cial passenger and cargo transportation, includ-

underneath to wind. In paper #30, progress on ing car-passenger ferries, rescue, and tourism

a hybrid hydrofoil-Ekranoplane was described. (perhaps, cruise ships). He stressed a special

The prognosis does not sound good as it was application to create a global international sea

stated that "the resulting operational efficien- rescue system. This idea and this paper de-

cies puts this vehicle in the helicopter class." serve careful study. At minimum, Aframeev

This is not surprising as structurally and sea- has provided what seems to be a critical evalu-

way effective hydrofoils at speeds over 100 kts ation of Russian Ekranoplane technical perfor-

will have efficiencies for below those of wings mance, as already touched upon in preceeding

in air. A more promising approach to improve paragraphs.

take-off is being taken by Fisher-Flugmechanic, The systematic technical development of

paper #37, who propose to use the propellers Ekranoplane aerodynamics was carried out fol-

to pump air underneath the wind and seal lowing their introduction in the 70's. Consid-

it partially underneath with a hovercraft type ering the high degree of stability required, it is

bag-seal which can be retracted after take-off. not surprising that stability in ground effect has

They call this vehicle a Hoverwing. A hybrid been intensively studied. This is briefly reviewed

hovercraft-Ekranoplane results. In terms of per- in #32, and a more detailed asymptotic analysis

formance, paper #37 shows higher thrust weight is presented by Kiril Rozhdestvesky of St. Pe-

ratios for WIGS (Ekranoplanes), at between 4 tersburg in paper #22. Fortunately, a high de-

and 5, then Aframeev in #29, between 3 and 4, gree of longitudinal and laterial stability seems

and Fisher projects a value of about 7 for his obtainable through the application of known re-

Hoverwing-80, of which a model is already fly- sults; indeed, this makes the Ekranoplane fea-

ing. This value would be substantially in excess sible. It is particularly fascinating that a par-

T-6

**ticular airfoil shape (S shaped camber) allows senger and car transport in ground effect will be
**

stability without a tail surface; this was men- very much brightened.

tioned by both Rozhdestvensky and Fisher but

is not documented in the written papers. Fi- 4 Other Interface Vehicles and

nally, the relative role of the upper and lower

surfaces in providing lift and stability are most Problems

important and interesting. In paper #30, the

A variety of isolated papers on SES vehi-

wing investigated at the NLR in the Nether-

cles, hydrofoils, ship control, missile water exit,

lands shows a thick foil (thickess/chord about

and special problems in interface hydrodynam-

20%) with maximum thickness far forward and

ics were presented. These contributed to the

with a flat bottom and trailing edge flap, Fig. 7;

breadth, balance, and interest of the Sympo-

this foil develops substantial negative pressures

sium, without allowing for evaluations of the

on its upper surface over the "bump" there, and

subjects involved. A paper, #15, on SES accel-

it seems remarkable that the upper surface pres-

erations (the cobblestone effect) due to dynamic

sure distribution is almost unchanged in and out

compression of the air cushion by waves illus-

of ground effect. On the other hand, "when a

trated again the problems of comfort associated

wing flies very close to an underlying surface,

with operation in a seaway. A number of the

the prevailing contribution to its aerodynamics

papers presented interesting hydrodynamic phe-

comes from the channel flow underneath." This,

nomena and ideas, #35 and #36, and #33 pre-

of course, simplifies design and analysis.

sented a review of an important emerging tech-

Waves inevitably play a large role in determin- nique for the simulation of certain high speed

ing the design and inevitably degrading the per- non-linear features at the interface.

formance of marine interface vehicles; for exam-

ple, main shortcomings of the Ekranoplane (pa-

per #29, section 3.2) include "the complexity of 5 Concluding Remarks

take off, particularly at rough sea, the necessity

The Symposium was highly successful in that

for qualified pilot team and their constant train-

it gathered together experts from both aero-

ing." Other authors also mentioned the high

dynamics and hydrodynamics, in almost equal

mean vertical accelerations during flight over

number, to discuss interface operations.

waves. It was therefore welcome to see data

Helicopter-ship mating poses severe problems

on wing performance as effected by flight over

requiring accurate simulation. Further research

waves, paper #21 from Kiev, including extensive

on the aerodynamics around and behind ship su-

data on smooth sea performance, too. The most

perstructures is necessary, both theoretical, nu-

important aspect of these data are the fluctua-

merical, and experimental, as are validation of

tions in verticial force due to flight over waves,

numerical results. Systematic wind tunnel data

from which the seaway induced vertical accel-

bases are desirable. Further fundamental re-

erations may be deduced. Since human toler-

search on unsteady three dimensional separated

ance to sustained mean accerlations is limited

flows and its application would be helpful.

to about 0.10 g, limitations on performance in a

The participation of a substantial Russian del-

seaway are implied. This has always been a se-

egation made possible a good review of the sta-

rious problem for hydrofoil craft and led to the

tus of the Ekranoplane. Technical performance

adoption in the US of fully submersible control-

exceeding that of aircraft has not so far been

lable foils. A monograph (in Russian) exists on

the dynamic control of Ekranoplanes, including achieved. Emphasis was put on equal devel-

a brief history which was published in St. Pe- opment of both hull borne and wing borne op-

eration (joint mode) for special applications as

tersburg in 1994; the author is A.V. Nebylov.

global marine disaster response.

Finally, a very upbeat presentation was made Other subjects such as SES, hydrofoils, missile

by H. Fisher, paper #37, describing the devel- water exit were briefly treated; further attention

opment of Ekranoplane vehicles, utilizing hov- to interface vehicles and their technical problems

ercraft ideas for take-off, as mentioned earlier. by this Panel would certainly seem warrented.

High performance was promised for a 100 kt ve-

hicle for the transport of 80 passengers. Should

this development succeed, the prospect for pas-

1-1

**RECENT DEVELOPMENTS IN THE ANALYTICAL INVESTIGATION OF
**

SHIPBOARD ROTORCRAFT ENGAGE AND DISENGAGE OPERATIONS

Edward C. Smith*, Jonathan A. Keller1, and Hao Kangt

**Department of Aerospace Engineering
**

The Pennsylvania State University

University Park, PA 16802-1401

**This paper presents an overview of recent developments in an effort to predict transient aeroelastic
**

rotor response during shipboard engage and disengage sequences. The blade is modeled as an elastic

beam undergoing deflections in flap, lag, extension and torsion. The blade equations of motion are

formulated using Hamilton's principle and they are spatially discretized using the finite element

method. The discretized blade equations of motion are integrated for a specified rotor speed run-up

or run-down profile. Blade element theory is used to calculate quasi-steady or unsteady aerodynamic

loads in linear and nonlinear regimes. The analysis is capable of simulating both articulated,

hingeless, and gimballed rotor systems. Validation of the rotor code is discussed, including

correlation with droop stop impact tests and wind tunnel experiments. Predictions of safe

engagement and disengagement envelopes, limited by excessive blade tip deflections or hub moments,

are presented. Future directions of study are also discussed.

**Introduction them. Most tunnel strikes cause minor damage
**

THIS paper presents an overview of the such as denting or tearing the synchronization

research that has been conducted at Penn shaft cover or cracking the synchronization shaft

State University since 1995 to predict the mounts. Major damage involves severing the

transient aeroelastic response of rotors during synchronization shaft, blade failure, or droop

shipboard engage and disengage operations. stop failure.

Unique and often hazardous conditions are For gimballed rotor systems, such as the V-

encountered when rotorcraft are operated from 22 Osprey tiltrotor, the entire hub can tilt and can

ship based platforms. One troublesome contact the gimbal restraint, resulting in high

rotorcraft/ship interface problem can occur loads on the gimbal spring and hub component.

during the engagement and disengagement of the Potentially, these loads could be much larger

rotor system while the aircraft is on the flight than the peak design loads.

deck. In this low rotor speed region, the Shipboard engage and disengage behavior

centrifugal stiffening is low but aerodynamic has been the subject of much research at the

forces may be large, particularly in high wind University of Southampton1"6. The transient

and sea states. Excessive aeroelastic flapping blade response was investigated using a flapwise

elastic rotor code for a hingeless rotor system

due can be the result. This problem often limits

using analytic run-up and run-down rotor speed

the safe operating envelope for the rotorcraft and

time histories and quasi-steady aerodynamics,

can severely restrict flight operations at sea.

including a trailing edge separation model. The

For articulated rotor systems, the rotor blade

ship airwake environment included ship roll

can deflect several feet and contact the fuselage

motion effects on wind and simple deterministic

of the helicopter, resulting in a "tunnel strike" in

gusts which were developed from model scale

tandem rotor configurations; or the tailboom,

wind tunnel surveys correlated with full scale

resulting in a "tailboom strike" in single rotor

ship airwake data. The aerodynamic model was

configurations. The H-46 Sea Knight, a tandem

later improved by including flap-torsion coupling

rotor helicopter used by the U.S. Navy and

and the deterministic gust model was correlated

Marines, has encountered over 100 tunnel strikes

with additional wind tunnel experiments. The

since 1964 and still occasionally experiences

*Associate Professor, Member AIAA, AHS.

^Rotorcraft Fellow, Student Member AIAA, AHS.

tVisiting Scholar, Nanjing University of Aeronautics and Astronautics

**Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the
**

Air-Sea Interface", held in Amsterdam, The Netherlands, 5-8 October 1998, and published in RTO MP-15.

1-2

**elastic flap code was modified to model The assumptions in the original analysis in
**

articulated rotor systems in which blade motion Refs. 7-10 included that: the total rotor thrust is

is constrained from excessive upward and close to zero which precludes the need for inflow

downward flapping motion by mechanical flap and ground effect models; gravity forces are

and droop stops. Recent studies were conducted necessary since the blade operates at low rotor

to validate the rotor analysis in which a model speeds; ship motion effects on blade inertial

rotor system with rigid, teetering blades was loads are negligible due to low ship motion

placed aboard a scaled ship deck and tested in a frequencies relative to the rotor frequency; and

wind tunnel. The influence of the ship's airwake lead-lag degrees of freedom are neglected

was proven to be very important to the blade's because the portion of the engage/disengage

behavior. sequence in which the majority of tunnel strikes

occur is where Q.(t) is less than 20% of £20 . In

Overview of Previous Research this region Coriolis forces are small. However,

RESEARCHERS at Penn State University, in axial and lag degrees of freedom have since been

conjunction with the US Navy, have also added to the analysis and are detailed in Ref. 17.

been developing simulation tools to predict Quasi-steady airloads, including

transient aeroelastic rotor response during noncirculatory effects and either linear or

shipboard engage/disengage operations7' . This nonlinear separation using a Kirchoff model can

research differs from the efforts in Refs. 1-6 in be predicted. In addition, unsteady aerodynamic

several ways. The rotor model included both loads, including attached flow, unsteady trailing

flap and torsion elastic blade motions with edge separation, and dynamic stall effects can be

Bernoulli-Euler beam bending theory. Using the included.

finite element method, radial variations in elastic, The following sections will give a brief

inertial, and aerodynamic properties are easily introduction to the analytical development of the

accommodated; and different hub types are helicopter structural and aerodynamic model,

modeled with greater ease. Either a quasi-steady examples of validation of the model with

or a time domain unsteady aerodynamic model experimental data, and examples of the

based on indicial response is used to calculate parametric studies conducted to date. For

airloads in the attached flow, nonlinear example, the effects of collective and cyclic

separation and dynamic stall regimes. control settings, droop stop angle, and ship

Experimentally measured rotor run-up and run- motion on blade response were investigated " .

down rotor speed time histories were employed A preliminary study of the effectiveness of a flap

in the prediction of the transient blade response. damper in reducing tip deflections was also

As an example, the measured rotor speed run-up conducted9'10. The importance of torsion and

profile for the H-46 Sea Knight can be seen in unsteady aerodynamics to the simulation of the

Figure 1. blade response was also determined 9,10

**Engagement and Disengagement Analysis
**

I. Structural Model Formulation

The partial differential equations governing

blade motion are derived using the generalized

Hamilton's Principle

\ 1 1 1 1 h m=l''(sr-su+sw)dt=o (i)

0 2 4 6 8 10 12 14 16 18

Time (seconds)

where ST is the variation of kinetic energy, ÖU is

Figure 1: Measured H-46 Rotor Speed Run- the variation of strain energy, and SW is the

up Profile

virtual work due to external forces. Detailed

Given a ship airwake environment, the blade formulations of all terms are presented in Ref. 7.

equations of motion were integrated in either The variation of strain energy and virtual work

physical or modal space using a fourth order are composed of aerodynamic forces, gravity

Runge-Kutta integrator to solve for the transient forces, pitch link stiffness, and droop stop

aeroelastic response. stiffness and are expressed below

1-3

**6U = 6UB+6UDS+SUPL+6UFS (2) shipborne helicopters such as the HUP-1 through
**

6W = SWG+SWAF+SWFn (3) HUP-4 series to reduce excessive blade flapping

during low rotor speed operation, but are not in

where the subscript B is for the blade, DS is for use on the H-46. The flap damper is assumed to

the droop stop, PL is for the pitch link, FS is for act through the entire angular range between the

the flap stop, G is for gravity, AF is for the droop and flap stops. The virtual work

aerodynamic forces, and FD is for the flap performed on the blade due to the flap damper is

damper. Brief formulations of the variation of

droop and flap stop strain energy and flap <5Wrn = -C„Aw 6Aw' (5)

damping are presented below.

The rotor system features most important to where Cß is the flap damper strength, and Aw is

the engage/disengage process for an articulated the flap hinge velocity

rotor system are the droop and flap stops. The II. Aerodynamic Model Formulation

droop stop, shown in Figure 2, restrains the The shipboard aerodynamic environment

downward rotation of the flap hinge during low consists of ship roll motion effects on ambient

rotor speeds and at rest. The flap stop restrains winds and deterministic gust distributions over

upward flap hinge motion using the same type of the rotor disk. Except for roll motion, the ship is

mechanism.

assumed to be stationary with respect to the

/ Counterweight

inertial coordinate system,. The atmospheric

Rotor Shaft winds relative to the ship are uniform and are

termed the relative Wind-Over-Deck (WOD)

conditions defined by VWMl the wind velocity,

\p wod the wind direction, and Vven, the vertical

wind component as seen in Figures 3a-b.

Droop Stop ™Y/ad

Figure 2: Droop Stop Schematic

**The droop and flap stop interactions with the
**

blade are modeled using a conditional rotational

spring located at the flap hinge. While the droop

stop is extended, the rotational spring stiffness is

zero for hinge angles greater than the droop stop

angle and large enough to restrict the flap hinge

rotation to less than 0.1° for hinge angles equal to

or below the droop stop angle. The variation in

the strain energy for the droop stop is expressed

as

tUDS = Kß (Aw' - w'DS )5Aw' (4)

**where Kß is the droop stop spring stiffness, Aw' is
**

the flap hinge angle, and Aw'DS is the droop stop

angle.

Results of Refs. 7-8 indicate that under

certain wind conditions the blade violently strikes

the droop stop, causing a transfer of system

energy from kinetic to strain. Preliminary rotor

engage simulations are performed investigating Ship

the possibility of reducing the blade deflection by Figures 3a-b: Wind Over Deck Convention

adding a rotational damper acting at the flap

hinge. Flap dampers have been used on early

1-4 Recall that ship motion effects on blade where KlM is the vertical wind component and x is inertia forces were assumed negligible. is used to discretize them. The ship roll motion aerodynamic individual components are defined as model consists of the transformed relative WOD Vr =V„dcos(aw)cos(¥wJ (10a) conditions due to the ship roll angle and a wind component at the hub height due to the angular Vr' = -V^ cos(or„. the stall model calculates the additional lift and large linear gust developed in Ref.swla (10c) \ w /) sinusoidal variation of $s. damping.J (9a) IV. linearly to a downward wind component on the 1. In the linear gust of freedom are q.=0«xSin (6) P follows an unsteady thin airfoil theory development by Johnson". is used to formulate the blade equations of leeward half of the disk. and can be expressed as where aw is the angle of the wind relative to the r horizontal. The attached flow. the ship roll angle.) (10b) roll velocity. Ship roll motion is governed by Vzg""=-Vwoa. Discretized Blade Equations of Motion windward half of the rotor disk transitioning The generalized Hamilton's principle. The unsteady height given by aerodynamic was developed by Leishman and Beddoes12'14. This time domain approach includes In Ixf attached flow. the rotor disk on the blade response. ship motion effects on blade airloads are In the uniform gust distribution the included. the motion and the finite element method is used to uniform gust detailed in Ref. Eqn. however. unsteady trailing edge separation. ■-</> max eg cos (7) and dynamic stall effects. The global discretized distribution the individual components are equation of motion becomes defined as Mq+Cq+Kq=F (ID Vr'=K^„^<¥. and over a ship's flight deck subjected to a crosswind unsteady boundary layer effects. Analysis vr'=-V^„„M¥ml) (9b) The blade transient response is calculated by Vzsu* = V-*«-Kcosfor -ye* yr^ ) (9c) integrating the discretized blade equations of . This gust consists of a and quadratic in torsion. y = £l0t includes the airloads due to virtual mass is the nondimensional time. 2. The unsteady trailing of the ship. The model can be expressed as elemental mass. The roll angular linear or nonlinear predictions using Kirchhoff's velocity causes a lateral wind velocity at the hub nonlinear separation model. 2ny^ The quasi-steady aerodynamics development 0. This gust consists of an upward wind component on the III. leading edge pressure lags. are spatially discretized by substituting shape In addition. Each flexible element has seven determine the effects of uniform vertical winds degrees of freedom and is cubic in flap bending on the blade response. the damping matrix C. The dynamic are explored in this analysis. unsteady model predicts both circulatory and where hcg is the height of the rotor above the eg noncirculatory airloads. the nondimensional distance along the blade. where the subscript H denotes the hub fixed and the stiffness matrix K7. The global degrees coordinates system. edge separation model accounts for trailing edge Two gust models that simulate the airflow separation. The first model. The wind velocity due to the either gust point Gaussian quadrature method. )sin(^ira. and P is the (noncirculatory) effects and allows for either nondimensional roll period. The second model. 7. The energy expressions uniform downward velocity across the rotor disk. and stiffness matrices Y«»« =ys""i +v* "j„+vrKH (8) and load vector are then assembled to form the global mass matrix M. is adopted to downward pitching moment created by the explore the effects of local wind variations over shedding vortex. each model features a lateral wind functions into the elemental virtual energy component in the direction of the relative WOD expressions and integrated in space using a six velocity. This development where <j>imx is the maximum roll angle.

The first and simplest technique is a direct solution of Eqn. Blade rotational spring is added to K to freeze the response can be computed in either the full finite hinge angle during blade contact with either the element space or modal space. The time contact with the droop stop. The second technique is a solution of Eqn. K. 1-5 motion. When the blade is not in recalculated at every time step. 12 into Eqn. two independent sets of natural illustrates the first four coupled flap-torsion frequencies and mode shapes are generated frequencies overlaid with test data for the second through the fourth flap modes from Ref. 15-16. An modal amplitudes investigation to the behavior of a model blade . Recall that a very stiff conditions for run-down solutions. In this method. Droop Stop Impacts amplitudes must be adjusted to maintain the An important feature of articulated rotor physical shape of the beam when the mode blades is the droop stop mechanism. Substitution of Eqn. C. 13 using only the hinged modes of the beam regardless of blade-droop stop contact. as the initial conditions for rotor run-up the flap hinge is restrained from further solutions. 7. The deflection of the beam can be system. and F which must be during such an event. The (0 = a>„ and if ß>ßns flap-torsion elastic analysis is in satisfactory (15) agreement with the blade data. 11 system. it is common to transform the global equations of motion into a Model Validation modal coordinate system I. The predicted Mä + Cä + Ka = F (13) static tip deflection of 15. 13 using either the hinged or Nondlmenslimal Rotational Speed cantilevered modes depending on droop stop Figure 4: H-46 Flap-Torsion Fan Diagram contact.one corresponding to the diagram is also developed to determine how well cantilevered condition of the beam and one the finite element approach models the H-46 flap corresponding to the hinged condition of the and torsion rotating natural frequencies. The boundary azimuth are functions of time and are used in the conditions of the blade change dramatically calculations of M.6 0. recommended for computational efficiency For computational efficiency. excessive flapping can cause the rotor approximated by either set of modal matrices and blade to repeatedly impact the droop stop. called a "Modal Swapping On" integration. called a "Physical Space" integration. called a "Modal Swapping Off" integration. the modal II. and ßDS is the droop stop angle. The third and most complicated technique is a 0.3 inches is within 2. where co is the natural frequency.S 1 solution of Eqn. Therefore. During the shapes are changed or discontinuities can arise in engagement and disengagement of the rotor the solution. but modal space is droop or flap stop. ß is the flap hinge angle. Blade Fan Diagram so(*M0 (12) The H-46 Sea Knight is a three-bladed tandem rotor helicopter.. to rotate and the blade acts as a hinged beam. Figure 4 beam.6 Two different stiffness matrices are used in percent of the Boeing test data average. The blade static deflection is used When the blade is in contact with the droop stop. The steady state response at the downward rotation and the blade acts as a operational rotor speed is used as initial cantilevered beam. but the present research where <D is the mode shape and a is the modal only models a single rotor blade of the aft rotor amplitude. the flap hinge is free integration is performed by a 4th order Runge. A fan this transient analysis . Recall that the rotor speed and blade was undertaken in Ref.4 0. 11 in full finite element space. Proper stiffness and mass distributions and multiplication by Or yields are verified by comparing the predicted static tip deflection to experimental data. The equations of motion can be solved using three different techniques. co = cor and if ß^ß. Kutta scheme.

B. Once the blade contacted the droop stop. Premultiplying Eqn.3 0. the blade was held in place by an response of the flap hinge angle is shown in electromagnet located just beyond the flap hinge. set at an angle 0. Once the blade tip reached During a transient analysis. This process was repeated 16 by ®^M yields four times for each initial flap hinge angle to determine the repeatability of the data. =<&TM<!>tah (17) At the beginning of each drop test.5 Time (s) of 0°. the mode shapes to cantilevered mode shapes is excess strain energy is then converted back into necessary when the beam contacts the droop kinetic energy in the upward rebounding of the stop. e is the hinge offset shapes is performed when the beam leaves and the subscript SD is the static deflection. Droop Stop Figure 5: Model Blade Schematic Each drop test commenced when the electromagnet was shut off. the measured tip accelerations can be focus on the impact between the rotor blade and used to calculate the measured tip displacement the droop stop. qlip (f = 0)= (L . $rMO. In each (20) drop test the root end of the blade was clamped between vice grips.a. initial velocity was set equal to zero since the A Froude scaled structural model of an H-46 blade was released from rest. Without anything to support it. Each drop test was performed at zero rotational since most H-46 tunnel strikes occur at ?(0=YM0+9*('-A0] less than 20% of the full rotor speed. the blade section outboard of the flap Droop Stop hinge rotated downward about the flap hinge until it contacted the droop stop.1 0. In order to specified.1-6 qs<M. and of about the flap hinge. Once the initial rotor blade was constructed from 6061-T6 position and velocity of the blade tip were aluminum and is shown in Figure 5.2 0. The contact with the droop stop.4 0. Figure 6a. The experimental data tracks well with the =K-'F analytical prediction before the impact and while WfpfO) in contact with the droop stop.sa>rar (16) the inertia of the blade caused it to continue bending elastically. the initial shape of the beam is a combination of the static If the mode shapes have previously been mass droop under its own weight and a known rotation normalized. a switch from hinged its point of maximum downward deflection. a series of drop tests were and tip velocity conducted. however. Then the section of the blade ?(0=-jfe*(f)+«*(f-A*)] outboard of the flap hinge was rotated upward and given an initial flap hinge angle ranging from The results for an example drop test from 2° to 10°.10 Potentiometer Accelerometer seconds is because the blade flexes while it falls. Figure 6a: Flap Hinge Angle Time History . (19) A similar procedure is used when a switch from the cantilevered mode shapes to the hinged mode where L is the blade length. Once the desired flap hinge angle was 9. upward rebounding. The transient reached. it could not rotate downward any further. The cause of the sudden upward Linear Motion rotation of the flap hinge angle at t = 0. the modal blade tip. When this situation occurs.e)sm(ßhingc (t = 0))+ q. ®jMOr = I. The initial condition for (18) the tip deflection is given by = <D^M®. The motion of the blade was measured amplitudes must also be adjusted to maintain the for one full cycle of downward bending and physical shape of the beam.7° are shown in Figures 6a-c.

Modal Swapping Off integration. i . 1-7 The transient response of the tip deflection is each drop test. Wind Tunnel Validation Studies Figure 6c: Strain Gage Time History Experimental data was gathered at the University of Southampton on the engagement As seen in the previous sections.2 0. . however. This ■—— Modal Swapping On g\ 4 1 E D Experiment reduces the length of the simulation to 10 hours.Static -| ßpf -| cd \ Droop 1 / p 1 simulation.1 0. .. Note that the CPU time can result using the three integration strain becomes negative at t = 0. the Modal Swapping Off measurements at five different positions were integration. .05 seconds. A dramatic difference in the measured and predicted strain. . .. and the Modal Swapping On taken without the helicopter model.. For each integration provide almost identical solutions for rotor hub location. i . i . of CPU time because only four normal modes "i? 0 \\ U Ss were required for the solution. the Modal Swapping On \ff\ U i i w i00 «V - 1 'pffi integration is the superior method for use in the .1 even the longest drop test solution are not very \T "M substantial.5 seconds of real time. winds were measured along . wüp/L. only 0.4 0. = 0. techniques in the aeroelastic analysis in Refs.2 0.15 p§3» stop impact event is a very small part of a larger ■ ..e. each of the and disengagement of a model helicopter on analytic techniques presented . Both the experimental data and time.2 picture . .5 Time (s) disengagement simulation. 6- indicating the beam is actually curved upwards.the complete rotor engagement and 0. -0. airwake Space Integration. the size of the 1500 —— Physical Space problem is reduced to 6 normal modes.. Much like the upward the computational efficiency of each solution movement of the flap hinge angle. The Modal ——Physical Space 0. i . because 41 degrees of analytical predictions capture this phenomenon. if the fan 1 f Modal Swapping On integration is used. i -0. -500 0 0. . Clearly. There is the aeroelastic rotor analysis make computational excellent agreement in the peaks and valleys of efficiency essential. The Modal Swapping Off The maximum tip deflection. i . 1000 which is still fairly substantial.5 Time (s) III. Finally. when the issue of 55 computational efficiency is considered along with accuracy. requiring 160 hours of CPU The maximum strain of 1400 |i£ is captured well time on an RS-6000 workstation. i . First. 30 L In: Wö minutes is required for the engagement — 500 .20 is shown in Figure 6c.3 0. i i i .1 . again the longest.3 0. If the captured by the experimental data and analytic engagement simulation is performed using the predictions. However. . freedom are used.0w*S Swapping On solution is the most efficient of the 5^ 0. of -0. .17 at t solution is the next most efficient model. . £ -0. 9.05 The differences in CPU time to generate CD : Static Droop a . the cause of method are of importance. Each drop test the upward motion of the blade tip at t = 0.4 0. .04 simulates 0. both the accuracy and shown in Figure 6b. . requiring only 13 seconds : ffn u . a total of 70 seconds. Many computationally intensive features such as blade Figure 6b: Tip Deflection Time History element unsteady aerodynamics which result in motion dependent terms and large numbers of The transient response of the strain gage at degrees of freedom including flap and torsion in x/L = 0. \ \ i J .15 .. . The Physical Space engagement solution is and is indicative of the blade flexing as it falls. The Physical seconds is due to the flexing of the blade while Space solution method requires the most CPU falling. . Instead of excellent agreement with the experimental data. rotor engagement/disengagement simulation.25 seconds for the analytic methods is in requiring 24 seconds of CPU time.the Physical board a scale ship deck5'6. the rotor blade and droop -0.05 \j ——~ Modal Swapping On a Experiment /jpf»* three analytic methods. i I i»*X ! 6 normal modes are required. integrating 41 coupled equations of motion. .1 0.

The variation over The experimental data gathered in Refs. 5-6 is shown on top and the Deck Position A theoretical predictions from the Penn State analysis are shown on bottom. For each figure.75 1 Radial Station (r/R) PSU/NAWCAD Analysis Figures 7a-b: Model Ship Airwake Measurements from Refs. 5-6 2000 Then. the experimental response measured in Refs. Deck Position E 3 90' Deck Position A 1 1 «o- 1 0 i -30 ^ \ Deck Position D| \ueck Position B 1 \ £ -60 | -90' J -120 ■ 10 20 30 40 50 60 -150 . . The results tend to response predicted by the Penn State aeroelastic indicate a very distinct division between laminar analysis. The rotor hub was tested in the 5 locations on the flight deck as shown in Figure 8. 0 10 20 30 40 50 60 Elapsed Time (sec) Figures 9a-b: Deck Position A Figures lOa-b shows the comparison for Windward Leeward deck position E.1X0 1 1 -1 1 V 1 1 Elapsed Time (sec) -1 -0. Correlation of the Penn State analysis and turbulent flow regions. 5-6 the rotor disk is shown for each of the five hub was then used to validate the transient dynamic locations (A-E) in Figure 8a. the theoretical predictions show good agreement with experimental results. —-""^^ o "3120. Tv Deck Position C K 1500 S 150. Figures 9a-b show the comparison for deck position A.75 -0.25 UK Wind Tunnel Data Radial Station (r/R) 30 i r 2000 3D 180. ^^. Note that the experimental data shows very little random behavior as might be expected in a laminar airwake. -0. Figure 8b illustrates results with the single-bladed scale model rotor the variation of flow inclination to the horizontal results was reported in Ref. This ship/helicopter combination was subjected to beam winds in the wind tunnel and a series of engage and disengage operations were performed.25 0 0. 9.1-8 the direction of wind flow. respectively. The comparisons between the Note the large change in flow inclination for the experimentally measured rotor response and the three leeward hub locations which is indicative of predictions are for deck positions A and E are reverse flow.5 -0.5 0. shown in Figures 9a-b and lOa-b. In this case. 5-6 airwake. as seen in the random blade response. Note the difference in behavior Figure 8: Model Ship and Rotor Wind Tunnel when the rotor is fully immersed in the turbulent Experiment Schematic from Refs.25 0 0. over the rotor disk for the five hub locations. a radio control helicopter model was mounted on the scale model flight deck.25 0.

An H-46 baseline engagement envelope 46 operations. Representative blade to 75°. in Ref. was maximum downward tip deflection occurs at that different large scale airwake characteristics approximately 7 seconds into the engagement. in Figure 12. Typical Rotor Engagement and Disengagement II. the collective is 3°. First. (i. and the deflections. 10 20 30 40 50 60 Elapsed Time (sec) PSU/NAWC Analysis 0 10 20 30 40 50 60 Elapsed Time (sec) Figures lOa-b: Deck Position E Figures lla-b: Rotor Engagement and Disengagement Parametric Studies I. The standard control inputs are developed using standard control inputs and a used for all parametric studies unless otherwise uniformly distributed gust with a. Kvcn . envelope shows safe engagements for winds The unsteady aerodynamic model and dynamic speeds up to 40 knots for WOD directions of 45° stall effects were used. 7 showed that control inputs have a Note for rotor engagement. a series The engage envelope analysis determines the of parametric studies for the H-46 helicopter was blade-to-fuselage clearance for multiple wind- conducted. Engage Envelope Study After the validation was performed. the "auto" cyclic trim settings used in actual H. 5-6. and fKml = 90° was used. Results of sensitivity studies performed responses using are shown in Figures lla-b. The experimentally measured rotor run. Note that when the UK Wind Tunnel Data rotor brake is applied. blade transient response. with Vwnd = starboard WOD conditions. The predicted 40 knots.25%. and the lateral is composed of all wind-over-deck conditions cyclic is 0. the tip deflections moderate effect on maximum downward tip decrease as the rotor speed increases. different deck positions) are very important During rotor disengagement.5°. a typical rotor engagement and over-deck conditions using the maximum disengagement were examined. the blade to the blade sailing phenomenon deflections steadily grow. the tip deflections grow more rapidly and the maximum deflection occurs within the last two seconds. During all rotor downward tip deflection calculated from the engage/disengage sequences.=150 is shown noted. in agreement with Refs. These control inputs are termed with blade-to-fuselage clearances greater than 8 the standard control inputs and are derived from inches.07°.e. 1-9 The conclusion. The two following studies show the . A linear gust model. Note that the H-46 is limited to up profile in Figure 1 and the run-down profile wind speeds of 30 knots for port WOD with rotor brake application at 55% of full rotor conditions and is limited to 35 knots for most speed are used. The safe engage region the longitudinal cyclic is 2.

were investigated as means of The predicted H-46 engage envelope for a reducing the blade deflection. a preliminary investigation envelope shows a 5 knot reduction in the safe was conducted into the effectiveness of adding a engagement region for WOD directions of 45° to rotational damper at the flap hinge of the H-46 to 75° and 120° to 285°. lateral cyclic pitch increase of 2° from standard Tip deflection time histories for rotor control inputs is shown in Figure 14. or 17500 ft- lb/(rad/s). small changes in seconds after engagement in both configurations.1-10 effect of 2° variations in collective and cyclic collective and cyclic inputs can moderately effect pitch controls from the standard control inputs. A In an effort to reduce the frequency of tunnel comparison with the baseline engagement strike occurrences.5° to 10° and flap Figure 13: Safe Engage Envelope for 2° damper strengths ranging from zero to five times Collective Decrease the H-46's lead/lag damper strength. 330 ooo 030 330 000 030 300 300 060 270 270 4 090 240 240 120 "Safe Ehgagifl&sgiöh/y 210 180 150 210 180 150 Figure 12: Baseline Safe Engage Envelope Figure 14: Safe Engage Envelope for 2° Lateral Cyclic Increase The predicted H-46 engage envelope for a collective pitch decrease of 2° from standard III. of -. However. indicated that the flap damper. regardless of its strength. Unlike the engagement are presented in Figure 15. was not effective in reducing the blade deflection because the angular range of motion was too small. At very large damper strengths. raising the flap stop angle permits the damper to act through a larger angular range of motion and dissipate more energy. are examined. The results illustrate that during engagements in maximum negative tip deflection occurred 7 uniformly distributed winds. Flap Damping Study control inputs is shown in Figure 13. The flapping behavior of the rotor note the substantial reduction in engagement blade becomes excessive at both 2 and 6 seconds capability for port winds. WOD speeds are into the engagement because it is both rotating reduced by 10 to 15 knots for WOD directions into the WOD and experiencing a strong upflow from 120° to 150° and from 225° to 330°. These from the specified gust distribution.54° and 1. Preliminary results for the current H- 46 flap and droop stop settings.5° respectively. the safe engage region. the flap damper simply freezes the flap hinge angle because the energy necessary to rotate the highly damped hinge is much higher than the energy necessary to deform the blade. Combinations 210 180 150 of flap stop settings from 1. dissipate energy in the flapping degree of 330 freedom. Two small decrease in wind speed over large portions flap stop settings and flap damper combinations of the envelope for the collective pitch change. .

such as the V-22 Osprey. The motion of gimbal is expressed by using two A gimballed rotor has three or more blades degrees of freedom in the rotating system. is again used to The hub is attached to the rotor shaft by a develop the equations of motion. lag bending.e. on all blades contributes to gimbal M=l flapping moments so the motion of all blades is coupled together through the rotor hub. the H-46 Tunnel i 0 2 3 4 5 6 steady periodic response of the rotor can be 7 8 Time (s) obtained by considering the response of an C = OxLag Damper. The overall goal of this feature the use of a gimballed rotor system. Each beam element Figure 16: Gimballed tiltrotor consists of fifteen degrees of freedom18. The rotor i . the universal joint. attached to the hub without flap and lag hinges.) + SUCS +SUGR + «ft/. For this type of analysis. the motion of blades is non- Figure 15: Flap Damping Study periodic. or gimbal. 16. elastic twist. 1-11 For a flap damper four times the strength of the also result in high angles of attack at very low lead/lag damper and a flap stop setting of 10°. gimbal restraint impacts make the individual blades respond differently Current Research and the inflow distribution on all blades MUCH of the current research at Penn State contributes to gimbal motion. The blades are discretized into a number of beam elements using HUB UNDEFIECTED HUB DEFECTED the finite element method. fiGC and fias . the summed over each blade gimballed rotor has some unique structural and aerodynamic characteristics related to shipboard SU = (^SU. enough energy is dissipated to prevent a tunnel An analytical model for steady-state strike.5 deg equivalent single blade. They -40 Tip Deflection * are determined by the trim equations in steady- state analysis. Engagement and Disengagement Analysis PR0PR0T0R HUBv I. In SW = (£öW?) (21c) addition. disengage operations.-» : : : i 60 %% model included coupled flap-lag bending modes •' '. Compared to contributions to these energy expressions must be articulated or hingeless rotor systems. rotor speeds. associated with the ship 8T = {%8r:) (21b) airwake. Hamilton's Principle. The entire multi- has focused upon the engagement and bladed gimballed rotor has to be modeled to disengagement operations for tiltrotors17. However. The »1=1 inflow distribution.t (21a) engagement and disengagement operations. aeroelastic response and stability of gimballed tiltrotors was developed in Ref. The gimbal tilt /f\ \ I \ 20 was described by using the gimbal degrees of 0 /' ' \ ''' III freedom. Flap Stop = 1.. rigid wing/pylon). For transient engage and disengage analyses. Structural Model Formulation A fixed rotor support was considered in the present research (i. gimballed rotor blades are typically short and highly twisted. l\ 40 and torsion degrees of freedom.A 7 1 U roll of the rotor disk in nonrotating frame. research was to systematically investigate the Unique rotor characteristics include a gimbal transient response of gimballed tiltrotors during spring and a gimbal restraint as shown in Figure shipboard engage/disengage operations. the highly twisted blades introduce significant structural flap-lag coupling and can . In addition. Present analyze the transient response during engage and tiltrotor configurations. 18. The multi-bladed gimballed rotor is modeled as several slender elastic beams attached to a hub and undergoing flap bending. Eqn 1. and axial deflection. In addition. which are the pitch and -20 ■Z' \ V\ /\.

The rotational spring in nonrotating frame. Aerodynamic Model Formulation A quasi-steady aerodynamic model is used Figure 19: Global Degrees of Freedom in this analysis. For expressed as ten elements in each blade. The second element in each blade represents the blade flexure. there are 275 global ÖUa=K. The variation of gimbal restraint strain III. The first two interaction with the rotor using a conditional degrees of freedom describe the gimbal tilt. GR to the gimbal Ref7. pgmd <|> /? sin y + l> cosy ■Pitch Bearing. The gimbal degree of freedom in nonrotating and rotating coordinate systems are shown in Fig. fi ««. Discretized Blade Equations of Motion energy is given by An illustration of the blade finite element discretization is shown in Figure 18. 17.i "4 "2 li w'2 (4 Figurel8: Finite Element Model of Gimballed Rotor Figure 17: Gimbal Degrees of Freedom Figure 19 shows the global degrees of This analysis models the gimbal restraint freedom for the gimballed rotor.1-12 where A^ is the number of blades. Pilch Bearing ßas siny cosy Coning Flexure blade 2# blade 1# where y is the azimuth angle of rotating frame. Each of the 6UGS = Kßcc ßGCSßGC + Kfm ßGS SßGS (22) flexible elements consists of fifteen degrees of freedom. ßoc -siny COS\{f ß (23) twist {ft). and PL to the pitch link. ur T max = tan~ (26) / ß cos y/ . B refers to the shipboard aerodynamic environment are given in blade. restraint. The and three internal. and axial deflection (u). Detailed descriptions of the quasi-steady aerodynamic modeling and the . The variation of the second element of each blade describes the strain energy due to the gimbal restraint is rotation of the pitch bearing. GS to the gimbal spring. The maximum tilt angle. much like last degree of freedom at the left-hand node of the earlier droop stop modeling. which describe the elemental transformation matrix is given by flap deflection (w.0 sin yr / s II. two external used to express the motion of gimbal. These degrees of freedom are The degrees of freedom in the rotating system are distributed over five element nodes.w'). lag deflection (v. blade 3# —*■*! »2 • • • v 2 '1 ''2 »j *. and ß and ()> are the gimbal tilt degrees of freedom in the rotating system.v'). is Siven in rotating system by ySm„ = (/?cosy-0siny)cosym„ + ^Element© (25) (/?sin^ + 0cos^)sin^m„ Jemenf® | where \pmax is the azimuth where the maximum Gimbal Degrees of tilt angle located and given by ^4 Freedom.(ßm-flH )Sß^ (24) DOF.

-kk co^sir^ (28) M: =-EAe(u' + — + 2 2 )-EB Yä'9' * (33b) kßac&iriy+kß coly/ + EIt (v'cos(i9 + <j>). For a gimballed tiltrotor. w[m and twist. mass matrix. 1-13 The energy expressions are spatially where M. = £7 (v"sin(0 + <£). the blade root moments to the non-rotating hub- fixed coordinate system and summing over each w'1m=ßcosy/m+</>smy/m (29) blade."-F.= GJip' + EAK't{e + <f)\u+—+— where N is the number of finite elements for each + EBß'Y . damping M" = M" cos 0 . inertial and into the elemental virtual energy expressions.15)-Kpw[m (j)"'. The stiffness matrix. pitch moment and torque are expressed as <p"' = -ß sin ym + (j> cos iym (30) M" = 2^(M'cosy/-M' sinyf-M'ßr cosy/) (35a) where the subscript 1 indicates the left-hand node N.M" sin <f> (36a) matrix and force vector are assembled to form w'' their global counterparts. The flap bending moment. Hub Moments During gimbal restraint impacts. and F are for the entire rotor.f + £C.w'cos<p (36c) Mq+Cq+Kq-F=0 (32) .EBß'(y'cosd + w'sin ö) (33c) blade. +^A co^sir^ k]. =AT>'+(1 )(M. of the first element blade and \j)m is the azimuth M" =^i. (34) + etw{r-0. K. C. The roll moment. and twisting moment in the where [kG] is the stiffness of matrix due to deformed coordinate system are defined by" gimbal restraint and is expressed as M. The force vector The virtual energy expression for multi-bladed includes the constant structural. (29) and (30) are then substituted into Eqn.v'sin ö)j attached to the rotor shaft by a gimbal. lag bending moment and twist moment at the root of mth blade in the undeformed Eqns. the blade is fixed to the hub at the blade root. M™ and Mf are the flap bending wm = \m-\) (31) moment.cos0) (36b) equation of rotor motion is obtained as M" =M"(1 )-M'. The hub is . it is important to calculate blade and hub loads for design limits. The motion of the hub is expressed using two degrees where 9 is the pitch angle and defined as of freedom in this analysis. discretized by substituting the shape functions They include the linear structural.")a (27) IV. + C"q7 + Kpq. ß and <|>.M'siny/ + M'icosy/-M'ßrsiny/) (35b) of mth blade in rotating frame and can be expressed as Ml = 2JM'Jr +JW") (35c) where M'".w"sm(e + (*)) +L cosy/sir^ M. lag bending moment. which rotate 0 = OQ+ 0ic cos \jfm + eu cos \ffm with the rotor.w'cos(e + <ß)) (33a) -k^cosj/sin/r -EC.Wsin<j>-M'.(w'cos6> . coordinate system.f k^colys+k^sidys. aerodynamic contributions. The root slope. and are given by (28). ß and 0 .[EC. inertial and tiltrotor in terms of the elemental matrices and aerodynamic contributions as well as the load vector becomes nonlinear structural and aerodynamic contributions. The discretized M. of each blade are not restrained and can be expressed as a function of two gimbal degrees of Rotor hub moments are obtained by transforming freedom in the rotating frame.sin0+M.

Transient Response Analysis Using the analysis described in previous section.11R and 0.64 0. two lag and one torsion. The flap and lag a Experimental Data moments predicted by this analysis show good correlation with the experimental test data at moderate collective pitch. parametric studies are carried out to investigate the effects of wind speed and blade flexure stiffness on the rotor response and loads. The rotor collective pitch is set 67J = 0 ° with zero cyclic control input. 2 i In an effort to clearly understand £-200 engage/disengage behavior.00 0. Parametric Studies I. The maximum tilt indicates when the impacts between rotor and gimbal restraint occur.1-14 Model Validation I. ß and \y.88 Radius (r/R) (i = 0.00 0.52 0. The corresponding full-scale wind speed is equivalent to the 35 knots.40 0.75 0.64 0. 67S = 5°. The flap bending moment. 975 = 7.75 0. including three flap.5°.28 0.075. Figure 22 shows the time-history of transient response of gimbal tilt. The tiltrotor a Experimental Data operation was simulated for a starboard WOD of 0. A 25% upflow through the windward Figure 20a-b: Flap and Lag Bending Moment half of the rotor disk and a 25% downflow Comparisons (975 = 5°) through the leeward half were simulated.40 0. of tiltrotor.88 The normal rotor speed of the model is 888 RPM Radius (r/R) (helicopter mode).15R represents the coning-flexure of the blade. The blade transient response is calculated by integrating the discretized rotor — 200 c A -^ f equations of motion for a specified rotor speed I o run-up profile. Eighteen rotor modes are used in the transient analysis. WRATS Bending Moments in Hover Experimental data for blade bending moments and twisting moment in hover condition of a 1/5"1 scale model of V-22 WRATS gimballed rotor was used to validate the analysis.11 0. ßmax.5°) Note that the element between section 0. and twisting moment at six blade sections was measured. and fair Figure 21a-b: Flap and Lag Bending Moment correlation at high collective pitch. Comparisons ( 07S = 7. A comparison of these experimental results with the predictions of the present analysis is shown in Predicted Figures 20a-b and 21a-b. 0. .28 0. and the maximum tilt. The test was conducted for hover condition with varying collective pitch (zero cyclic pitch) at a rotor speed of 875 RPM.11 0. —ii i Predicted □ Experimental Data The rotor properties are similar to the Bell- NASA 1/5* scale aeroelastic model of the V-22.600 bending moments of the blades. an investigation was ^— Predicted focused on the motions of the tiltrotor gimbal and ■1.52 0. lag bending moment.

5 0.2 0.086. The total tp=0 sec flapwise stiffness is equal to Figure 23: The First Impact Event EIflap=EIyy cos2 6 +El„ sin2 6 (37) II.10) is just outboard of the pitch bearing and the second location (r/R = 0. where the gimbal impacts the hub is located at which is the maximum amount of gimbal tilt.4 0. 1-15 Azimuth (deg) 16 28 42 60 81 105 132 800 -JJ 600 400 200 I A I 0 '-■—■ t -200 -400 -600 -800 II Flap ■1000 Lag ■1200 0.7 0. Hub Moments The flap and lag bending moment time This would only be performed in the very histories at the inboard section of the coning low rotor speed region of a rotor engagement or flexure are shown in Fig 24. pitch moment.7 0. For articulated rotors.2 0. The t0.0 Time (sec) 0.9 1. moment vector sum is 18 times static hub load respectively.33 sec One manner of reducing the tip deflections is to %lad81=16° increase the flapwise stiffness of the blade by ^iadei=°° t.5 0.4 0. < 20%NR.1 0. The hub and gimbal restraint. th and t2 refer to the time of pitch moment during an engage operation at a beginning rotating. for various collective pitch settings. and peak vector sum of roll and 23. first connection between rotor higher WOD speed of \i = 0.8 0.=0. Two radial locations are shown. The rotor keeps connection with gimbal restraint while the blade 1 rotates to 20°. Note that the flapwise .9 1.3 0.8 0. The first location (r/R = 0. The azimuth static hub loads for 11" gimbal tilt operations.0 0. 101°.3 0.28sec increasing the collective pitch setting. | 4° Tilt Angle | | 11°TiltAngle| \ l\ Figure 25: Hub Loads During Run-up Operation Ongoing Work Several areas of investigation are currently underway.0 Time (sec) Figure 24: Bending Moment Response on Figure 22: Transient Response for Gimballed Blade 1 at the Coning Flexure Tiltrotor Figure 25 shows the peak roll moment. The rotor first impacts gimbal for 4° gimbal tilt angle operations and 6 times restraint when blade 1 rotates 16°. while the azimuth of connection moves counter clockwise from 101° to 118° during this period. and disconnection. control of the tunnel strike phenomenon is being investigated.1 0.0 0. It can be seen that disengagement where Q. t^O. peak The first impact event is illustrated in Fig.25) is the beginning of the blade airfoil section. Figure 26 impacts between rotor and gimbal restraint cause shows the increase in flapwise stiffness achieved substantial increases in blade bending moments.

S. W. Smith.Newman.. Engage/Disengage Operations. 1990. pp." Ph. S.—i—r-—. 4." Journal of Naval Sciences. Helicopters and Dr. Dissertation.Geyer. E. lO. the increased angle of attack. pp. S.Geyer. 8-91-8- Center. 233-239.. No. Pat Purtell).D. Engage/Disengage Operations. J. No. is also underway. Vol. i i > 1 i ' ' ' 1 Aeronautical Journal of the Royal Aeronautical r/R =0..P. "Wind Tunnel Helicopter Society. Larry Trick. W. "The Problems of Rotor Q.-j» --^T -. This research was carried out with support 1995. and Mr..• c 3 4.. 1995. of Aerospace Engineering.Newman.C.Newman. United Kingdom.C. Vol. and Smith. pp.25 .P. PA.W. Thesis." 4 . "A Theoretical Model for the value at the standard collective pitch setting at Predicting the Blade Sailing Behaviour of a r/R = 0. 3.J.." M.J. 1995. W. a _—[ EE . "Aeroelastic Analysis of analyses of simple ship configurations. large chordwise blade bending moments may be a 3.1-16 stiffness can be increased over three times its 2. Maryland (technical 8. 20." Vertica... 99. pp.J. Vol. Bill Geyer). finite stiffness of the wing and pylon. Incorporation of the airwake results from CFD 7. Mar. CT)." Vertica. S. 1988. "The Verification of a 5 10 15 20 Theoretical Helicopter Rotor Blade Sailing Collective Pitch (deg) Method by Means of Windtunnel Testing. S. .. Dept.Newman. Smith E. 96. Aug. Vol. Alexandria. increasing the collective Semi-Rigid Rotor Helicopter. 12..Hurst. The authors also want to thank Mr. . 267-278.. (0 /.25. Pennsylvania Acknowledgments State University. pp. pitch may also increase blade deflections due to No. Mr.Geyer Jr. "An Investigation into the of the rotor is through active stiffening of the Phenomenon of Helicopter Blade Sailing..J.. VA..C. for valuable technical support Aeroelastic Analysis for Shipboard and rotor data for tiltrotor configurations. Dept. from the Naval Air Warfare Center Aircraft Division at Patuxent River. However.. University of Southampton.).J. No. In addition. E. Q.P." Figure 26: Variation of Flapwise Stiffness with Aeronautical Journal of the Royal Aeronautical Collective Pitch Setting Society. 1. 152- Measurements of Ship Induced Turbulence and 167. and the Office of Naval Research 114. Analysis of Transient Blade Dynamics During Mr. and Newman." Aviation Depot at Cherry Point. such as in Transient Blade Dynamics During Shipboard Ref. 982.Newman. rotor blade by means of piezoelectric actuators. Vol. D.P. "Aeroelastic Analysis of Transient Blade Dynamics During Shipboard Engage/Disengage . Ben Settle of Bell 9..J. U ——r/R * 0...Geyer. The other way to increase flapwise stiffness 6. American 1. 41-51. the Naval Shipboard Engage/Disengage Operations. 956. 531-544. 14. and Keller. "Aeroelastic monitors: LCDR Mark Whittle. (technical monitor Dr. are being added to the simulation to model the Southhampton. VA. Helicopter. pp.S. S. pylon degrees of freedom Astronautics. Kurt Long. University Park. pp. the Aeromechanics Specialists' Meeting (Fairfield NASA/Army National Rotorcraft Technology County. 1995.10 Society. of Aeronautics and For gimballed rotors." Proceedings of the 52nd Forum of the American Helicopter References Society (Washington. Alexandria.. the Prediction of Helicopter Rotor Blade Response. I . "The Application of a concern.A. J. Engagement and Disengagement of a Shipborne Stands rdSetting -' 2 . 56-64. D. Theoretical Blade Sailing Model to Predict the Behaviour of Articulated Rotor Helicopters. Mark Nixon from US Army "Validation and Application of a Transient at NASA LaRC. i i . W. • ■ ' i 1 . . c i No. and Keller. North Carolina Proceedings of the 2nd International AHS (technical monitor: MAJ Robert Hellar)..C.1994. 5.. 20.

"High Order Accurate 12. VA. Helicopter Theory. Inc.. VA. VA. 18. Alexandria. and Smith.Keller. University Park. J..G. 16.. Smith. American Helicopter Society. 445-453. December 1997.S. Structural Dynamics.. pp. "Transient Response Analysis of Gimballed Tiltrotors During Engage and Disengage Operations. 13.S..A. J.Johnson. 3-17. June 1986. 14.Hao.Beddoes. K. and Materials Conference (Kissimmee. T.. 1994. J. MD. Vol. and Liu. and Knarr. UM- AER094-18." AIAA Journal of Aircraft." M. 42nd Forum of the American Helicopter Society. 17. .Bir. Dover. FL). The Pennsylvania State University. W.Long. Inc.. Reston.Leishman.C. "Nonlinear No.H." NASA TM X- 62..C.. Vol. E.. I. T. "Analytical Mode for Tilting Proprotor Aircraft Dynamics..C. American Institute of Aeronautics and Astronautics. 35. University of Maryland Advanced Rotorcraft Code (UMARC) Theory Manual. 345-357. DC). L.. York. Structural Dynamics and Materials Conference (Long Beach.369." Proceedings of the (Washington.S. 19." Proceedings of the 54th Aerodynamic Behaviour and Dynamic Stall Forum of the American Helicopter Society Using the Indicial Method. Equations of Motion for the Elastic Bending and Torsion of Twisted Nonuniform Rotor Blades. 3. and Leishman. University of Maryland. pp." Journal of the American Helicopter Society. C. College Park. 3.N. PA. New NASA TND-7818. E. American Institute of Aeronautics and Astronautics. and Conversion Mode Operation. "Experimental/Theoretical Correlation of Analysis for Helicopter Rotor Blade/Droop Stop Contacts. CA). 1-17 Operations. "An Experimental and Theoretical Correlation of an Analysis for Helicopter Rotor Blade and Droop Stop Impacts. "A Semi- Empirical Model for Dynamic Stall. Thesis." Proceedings of the 38'h AIAA/ASME/ASCE/AHS/ASC Structures. Washington. Department of Aerospace Engineering. 34. 20.Keller. 1989. D.Johnson. Aug. "A Ship Airwake Predictions for the Helicopter/Ship Generalised Model for Airfoil Unsteady Interface Problem.. Reston. Including Blade Torsion and Coupled Bending Modes. W. and Chopra.. J. 15. G.A. July 1994.Hodges D. and Beddoes." 11.. and Dowell E... pp." Proceedings of the 39"' AIAA/ASME/ASCE/AHS/ASC Structures.G. No.. 1994.H. 1974. J.

exhibit very good correlation with established momentum and power principles. At present. this method stage. and published in RTO MP-15.5m and the mass is 5. A simple theoretical model of the helicopter main rotor was constructed within the computational domain. These studies also provide no several positions above a ship profile. 2-1 CFD PREDICTIONS OF THE INFLUENCE OF EXTERNAL AIRFLOW ON HELICOPTER OPERATIONS WHEN OPERATING FROM SHIP FLIGHT DECKS. The power exerted at the rotor is obtained and compared to the ideal hover condition and A CFD analysis has many advantages. PA. The once obtained.1. Abstract. The 2. The flow solutions for every ship/helicopter combination. Introduction method solves the governing Navier-Stokes equations as shown in Equation 3. is the cost and speed at which one can be obtained. The boundaries of to occur when a test pilot performs the takeoff and the computational domain are also modified in order to landing manoeuvres.2 in recent years. Department of Ship Science. The solution is complete and. A CFD model of a hovering helicopter main rotor is developed to examine airflow in the presence of ship The flexibility of helicopters is reflected in the increased structures and side winds. . Flow solutions in the work load the pilot experiences and varies from low both two and three dimensions are achieved using the to dangerous. typically a frigate.J. The thrust exerted by the rotor was evenly transport whilst requiring less storage space and take-off distributed across the disc. The extra terms added to the determining safe helicopter operating limits is a very governing equations apply a downforce to the fluid. helicopter fuselage. Department of Ship Science. 5-8 October 1998. environments such as ships and oil rigs has risen steadily that is the rotor radius is 6. University of Southampton. pressure and velocity components are results also show that the helicopter control margins are known at every location throughout the model. Modelling the Rotor. the highest quality. the established method for region of the disc. Obtaining such a SHOL. SO 17 1BJ. An experimental SHOL (Safe Helicopter Operating Limit) such as this has the advantage of being realistic In order to model helicopter operations from a ship's but it also requires the provision of vast resources both flight deck. calculated. A CFD study can also be performed on ship designs whilst in the conceptual Using the modified boundary conditions. N. The whole procedure has to be repeated commercial flow solver CFX 4. Thus the effectiveness of helideck and hangar demonstrates the viability of CFD for predicting the ship designs can be tested to reduce turbulence over the flight airwake and the reduced power margins a helicopter deck.H. The Netherlands. therefore. The main rotor was designed to The demand for use of helicopters in marine reflect the loading and geometry of a Westland Lynx. contains a degree of danger. Department of Aeronautics and Astronautics. time dependant boundary conditions and dynamic flight. held in Amsterdam. Newman. but future models will include 3. by rotor is held constant and the resulting flow solutions are definition. Naturally the fidelity of the predictions must be of experiences whilst operating in the vicinity of the ship. Southampton. encounters large cross winds and velocity gradients. facilities than fixed wing aircraft. The rotor is modelled by demand to operate in ever worsening environmental modifying the governing Navier-Stokes equations in the conditions. also developed specialised military uses such as submarine detection and air to surface targeting. The helicopter provides a fast method of tonnes. UK. The most evident computational flow solution. The thrust of the experienced difficulties. This study has been exploratory and limited by computing resources. Cross winds are systematic information about the areas in which the pilot applied to the computational domain. S. tail rotor. Results from such an analysis are repeatable and not The flow solutions show that the airflow accelerates over subject to either measurement error or personal the flight deck and a helicopter operating in this region subjection. The flow more likely to limit the safe operating limit than the data obtained can be used in simulators so that pilots can power margin. Wilson. costly series of full scale trials. Wakefield. The helicopter has 1. The landing is rated according to generate a physically correct solution.1. The force was exerted within Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". This usually involves these forces are independent of the flow around the rotor waiting at sea for appropriate environmental conditions and equal to the helicopter weight. practice landing on ships safely. the rotor is modelled at in finance and time.

0 Equation 3.1.1: Hover Flow Solution.50m. an inner and an outer. Secondly.p. solution in Figure 4.2-2 the computational domain of Figure 3. At the outer rotor itself. The figure constant. this problem is alleviated by making must have the pressure/velocity relationship defined in the rotor two cells deep. including the shown in Equation 3. and is defined as the cell deep. the boundary always remains constant and acting through the centre of conditions have been modified. It is assumed that at an grid cell dimensions are 0. In reality the The default boundary conditions available within CFX domain size makes negligible difference to the flow 4.1. This is achieved using a 'pressure velocity or pressure is unjustifiable and gives physically boundary'. which has two implications. the Figure 4.1. Existing Boundary velocity and pressure are solved in exactly the same way Conditions as they are throughout the rest of the domain. An example is the flow within most CFD solvers. The force applied to the fluid outer region of the rotor. 1 y v1 y y V V y y1* V x X X x X X X X X X X X X X X X 1 X x w * y x * x * * X X X X y x X X X )< X X X X X X VSVN\\\\\\\\\\\. Ideally the rotor would be one pressure is atmospheric pressure. applying any of the boundary.2. the resultant vertical velocity component determines the induced power exerted across the rotor whilst the thrust To restore a realistic flow pattern. momentum in the downwash does not equate to the thrust applied at the rotor.1 and other commercial CFD software packages are solution. The air flow around a hovering helicopter is In order to instigate the given relationship the fundamentally different because there is no free stream computational domain must be constructed with two velocity and all fluid flow is induced by the helicopter boundaries. The thrust is distributed evenly across the span of the rotor model. 0 Thrust / Area / H Only thrust is applied to the fluid to evince the rotor. For this reason.1 Rot 01 ]\>1<5d el p.50m*0. The rotor is infinite distance from the rotor the velocity is zero.1 shows that the fluid flows equations are modified by the addition of the vector B as upwards throughout most of the domain. the pressure is available boundary conditions such as an imposed defined as zero.1 p+-. generated by allowing the velocity to vary around the boundary whilst the pressure remains .(Ü®D)=B + V. therefore to exert the required thrust to the fluid. thus validating this assumption. applicable to external flows which comprise a free stream velocity and some body causing a perturbation Equation 4. but instead a relationship between the two is used based In all the flow solutions described within this paper the upon Bernoulli's equation. In these regions the Navier-Stokes and actuator discs. the viscous 4.2. Quantitative study yields that acts vertically downwards.(a-pU®U) Equation 3. The therefore 26 cells wide.Vz=0 such as a wing or a building.^+pV. B was defined as in Equation 3. At the modified boundaries neither a pressure nor a velocity is applied the rotor. There are many features of this flow that indicates that the rotor thrust is exerted within the conflict with established understanding of both rotors hatched areas. For this assumption to be strictly accurate. Firstly. Equation 4.1 dt The vector term B has units of force per unit volume. but the flow solver can not resolve the sharp datum or zero. Modified Boundary Conditions effects that act on the fluid between infinity and the domain boundary must be neglected. the extremity of the domain.1.u re 3 . Figure 4. the flow is in no way predetermined at the location of the rotor. F i£. a boundary condition readily accessible incorrect flow solutions. Therefore any fluid entering the domain pressure gradients.

range used in the helicopter industry.dA = 67.3 reflect the Westland Lynx. these values were chosen to PH =J"w.2. Figure 4. Figure 5. = pj"w2dA = 4980N dt 5. as rotor that are not predicted within actuator disc theory. a test case was run to conserved within the system and Newton's equation is model the rotor in hover with no cross wind.Ap. A further study of a three dimensional rotor was performed.Q. The contours represent pressure variations of 25Pa.6. The 7% discrepancy between the two values can be attributed to variations in the induced velocity across the The momentum in the downwash was obtained.8*10"5Ns/m2.V Forces Applied to Fluid Here.A) = 63. Both two satisfied. with the thrust per unit area kept constant. The ideal power as found from standard actuator disc theory is shown in Equation 5.1. across the downwash. In each case the fluid used was air with a density 1.2. 13m and unit thickness. The thrust per unit area used is described in Equation 5. This integral is performed isothermal. and thus is not comparable to the power The two dimensional case modelled a disc of diameter exerted over the whole rotor plane. incompressible and turbulent. The difference between the predicted ideal power and the measured power is 7%. To verify that both the boundaries employed and the This indicates that momentum has been effectively rotor model work as intended.2 corresponds to an azimuth . The flow solutions is shown in Figures 5.2 surrounded by the boundary outlined above. In the region between the outer and inner boundaries additional terms were added to the Navier-Stokes equations enabling forces to be applied to the fluid to alter the fluid pressure.2N.3%. Outer Inner Boundary Boundary Computational Domain > > P=0 1/2. Figure 5.4.4kW is carried forward to studies of the flow around the ship helideck. /I7(2.2kW Equation 5. The power derived is per unit depth. Rotor In Hover The discrepancy between the momentum in the downwash and the thrust exerted at the rotor is 0.81 / (TT.4kW Equation 5. Half of the flow solution is shown in Figure 5. described in Equation 5.3. The solution is axisymmetric.1 : 2D Hover Flow Solution.2 : Modified Boundary Conditions. as shown in Figure 4. Other than at the walls.1.4 In order to gauge the validity of the flow solution.52) = 384. across the plane of the rotor itself. The total thrust exerted was 4996. The calculated power 67.3Pa Equation 5.p. The fluid was considered described in Equation 5.1 P1D = T.2. Thrust / Area = 5200 * 9. momentum and energy principles from existing actuator disc theory were compared to the computational disc.2kg/m3 and The power exerted by the rotor was evaluated as viscosity 1. the entire fluid domain is dM Equation 5. This integral was evaluated Such a difference is realistic since 10-15% is a typical 20m below the plane of the disc. and three dimensional cases were obtained. 2-3 The inner boundary was placed a small distance from the outer boundary within the computational domain.

and the rotor is placed over the centreline of the ship in Position 4 shown in Figure 7. 14m to port of the centreline of the helideck. This flow pattern agrees with experimentally measured data such as reference [3].2% discrepancy. A horizontal free stream velocity of 30 knots was imposed. . the contours are pressure contours at 25Pa intervals. USA.1 : Ship Geometry The flow solution is shown in Figure 6. Ship Airwake Apart from operational and performance calculations. The vectors represent speed and the incompressible and isothermal. 6. A flow solution was obtained for a 2D model of a helideck. Comparison of momentum in the downwash and thrust applied yielded a 0. Canada and Australia. Figure 5. The first position is shown in Figure imposed at these surfaces. Vo=30knots 14. The second position is 7m to port of the centreline as shown in Figure 7.00m Figure 6.2-4 angle of 0°. UK.2. downwind and above the centre of the ship.2.5m above sea level and was used and the flow was assumed to be turbulent. the airflow around the ship in the absence of the helicopter thrust is of interest for other reasons. such as helicopter blade strike predictions during rotor engagement and disengagement. The geometry of the ship is shown in Figure 6. the rotor is at a height of 10.3. The power exerted at the plane of the disc was 3% greater than the ideal value predicted by actuator disc. The figure shows the flow separating at the windward edge of the helideck and a large region of recirculation both above and downwind of the ship. to determine the turbulent kinetic energy imposed at the windward edge of the domain. national defence research organisations from. Ship/Helicopter Interaction their research into the Helicopter Ship Dynamic Interface1. zero flow rotor were chosen. The geometry of the ship used is the same as that described in Section 6. epsilon. The arrows represent the velocity vectors. 1 Tripartite Technical Co-operation Program.1 The domain extended 75m upwind. The k-s turbulence model 7. Three positions for the main Both the sea and ship were modelled as walls. The dimensions of the helideck are those used by the TTCP Nations for 7.2 : 3D Hover Flow Solution.1. Reference [1] was used contours pressure variations of 25Pa. Reference [2] was used to determine a realistic mixing length for the initial turbulence dissipation constant.

1 xxl The lateral wind that the rotor experiences was also quantified.4kW.00m with the given cross winds was calculated using ^ ^j Equation 7.1.1.) These values 7._ 4 . position is important. «~> represent velocity vectors and the contours are spaced at 1 lOOPa in the first two flow solutions and 40Pa in the Position 3 S>^^ 1 -T final flow figure. This has not been attempted to date Figure 7. Vertical velocity components would provide an estimate of cyclic and collective pitch variations that achieve the required thrust magnitude and distribution. The power to hover in a stationary position . Whilst the power required to maintain a certain hover Four wind velocities were chosen relative to the ship. Flow solutions were have been recorded near the extremities of the rotor. Figure 7. The results presented in the following sections are therefore instantaneous and represent a snapshot of the flow solution. The k-s turbulence model was phases of the model. Position 3.3 Rotor. Figure 7.00m not only indicate the velocity gradients across the rotor but provide a measure of control demanded. The arrows .1 Position 1 L-^^--. The flow solution was solved as incompressible. The pilot must have adequate quantities of both in order to perform manoeuvres. The computed power requirements are shown in Table 7. ! ! P= fw. WP obtained for all of these wind velocities.dA Equation 7. As outlined in Section 5. the power required to hover out of ground effect in still air is . the turbulent kinetic energy and energy dissipation parameters were determined as described in Section 6. Position 1. 30 knots port to starboard. 2-5 67.1. Position 2. For this reason the vertical flow velocities port and 60 knots port to starboard. the solutions were obtained using a time- stepping approach. 5m to starboard of the rotor centre respectively. used.2 : Rotor.1-3. namely power and control. because these flow solutions are two dimensional and provide a qualitative understanding of the flow regimes. (5m is approximately 75% of the rotor radius. The techniques are under development and 2D cases are much less demanding of time during the validation isothermal and turbulent. For future development. a three dimensional analysis could be extended to include blade element theory and Position 2 inverse simulation. The solver could not achieve a steady state solution for any of the problems due to the turbulent unsteady nature of the flows. 30 knots starboard to significant. VAV is the average lateral velocity component across the entire span of the rotor.1 : Rotor. There are two factors that limit helicopter operations in adverse weather conditions. For each of the flow solutions generated the corresponding power was calculated accordingly.. with the rotor at and Ws correspond to the vertical flow 5m to port and each of these positions.1. 7. AW is difference between the vertical components WP and Ws. 14.Ap. after the rotor has been stationary in the given location for at least 30 seconds.r ' """eT positions are shown in Figures 7. Zero Wind The flow solutions for hover in each of the three '.1. control requirements are as zero.

0 0. the ship generates. _——— i zzzzz^~L~~ ^^= I^^I^^^ — ^^^^ter^rrrr-™" >> == "'^^2^>t^r ij. Position 1 2 3 P(kW) 85.2: Rotor in Position 2... there are vortices The only areas of recirculation are present at the present on both sides of the rotor. There is a high pressure region directly beneath These three flow solutions were determined with a 30 the rotor and on the deck."'-''.1. Figure 7. The pressure significantly less than the hover.'.1.3: Rotor in Position 3.0 Table 7.1.2 -1. shown in Figure 7.1.'. The favourable ground effect knot free steam velocity acting from port to starboard.3 displays the flow solution for the rotor over 7.'.8 -13.1.1. The model shows the balancing of the rotor downwash with the upflow generated by the . These cause intersection of the ship's sides and the sea surface.'. which in turn increases the These are small in comparison to the vortices generated power requirements to a value greater than the simple with the rotor in the other positions. Wind 30 Knots. hover case. Zero Wind. The values VAV are the average lateral velocities across The rotor in Position 2 has a vortex present at the the rotor span.'. The other two ship.1 66.1 -3.0 WP -20.5 -14.3 0. contours are spaced at 40Pa.8 -13. The rotor in position 3 has VAV of zero outboard edge of the rotor but there is no vortex over the consistent with symmetrical flow./) \\ f \ \„ZZZZ"Zlll"A. The flow solution for the outboard position is displayed in Figure 7. The flow through the rotor itself is approximately vertical."1 . '■'-'*'■ '■ '■ '■ Figure 7. The ground effect give a power requirement comparable to difference in vertical flow across the span is also small in the hover case. --^ >~. the ship's centreline. gives a power requirement which appears left to right in the figures. The rotor has no recirculation at the tips.C*^-.5 49. Port to Starboard. — -.1 : Rotor in Position 1. This figure shows a large low pressure region above the rotor and a weaker high pressure region Figure 7.2.'. Figure 7.1 : Power and Control Requirements. However there is little vertical flow through the rotor itself.9 AW -2.'.'.6 -17. all three positions. since the rotor is effectively in climb.9 ws -18. significant recirculation.'. The combination of the vortex and the beneficial positions have small lateral flows across the rotor. below the rotor.1.2.9 %PH 126 99 74 VAvOns"1) 1.2-6 In Position 1.2.

2.2.6m/s or 9 15.9 10.9m/s helicopter. The fluid is passing upwards is more than double the other two positions caused by through the rotor on the port side and downwards at the only a small change in position of the rotor.3. the side of the ship. The pressure contours are spaced at 75Pa. As with the starboard side. This is reflected in the power exerted.1 ws -7. 10.2.2 6. This represents increased control other two positions the fluid is flowing upwards through requirements.0 -6. In reality this stream velocity of 15 knots.2 shows the flow solution with the rotor over Port to Starboard.6 10. However the rotor is experiencing widely knots. Figure 7.1.9m/s or 21 knots.2.7 Table 7. Port to Starboard These three flow solutions were generated using a free stream velocity of 60 knots acting from port to starboard.0 %PH 0 -1 0 VAvtms"1) 3.3 : Rotor in Position 3. Similar to the other positions the induced power Position 1 is largely influenced by the free stream and is small due to the upflow across the rotor.1 : Power and Control Requirements.5 4. with correspondingly high resulting control demands. The significantly less lateral wind than the free stream of lateral velocity the rotor experiences is 4. There is however a large vertical velocity velocity the rotor experiences. This indicates an autorotative state.6 -7.2 -1.1 shows that the flow around the rotor in 7.1. The greatest . The rotor over the ship's centreline is shown Figure Figure 7.8 13.9 4. This flow also exhibits a would necessitate large cyclic pitch variations to trim the large vertical velocity variation across the span.2. This is in fact less than the undisturbed free varying flow vertical flow across its span.0 0. The most the presence of the ship's helideck. This gradient across the rotor. 7. The power requirements are negative as shown in exerted at the rotor and the rotor is operating in Table 7. Position 1 2 3 P(kW) -0.2. 30 Knots Figure 7. Wind 60 Knots. as shown in noticeable feature of the flow solution is the lateral Table 7. 2-7 ship.4m/s.9 WP 3.6 AW 10. At this position the helicopter is resting in the upflow generated by the blockage of the Table 7. 10. the rotor on the port side and downwards on the starboard side.3.2.3.1 shows that in these cases there is little power ship.

to the rotor in the outboard position. These flow solutions indicate that although power requirements are small or negative the helicopter has to The flow solution for the rotor above the side of the ship contend with accelerated side winds and large velocity is shown in Figure 7. and finally up towards the rotor.3 WP 5. The rotor is operating in a strong than the free stream velocity. resulting in a larger 7. There is another recirculation region downstream of the ship.4. The rotor is operating within an upflow greater than the outboard position.4 0. The rotor This vortex appears to be small and does not contribute is operating in an upflow generated from the ship. as through the rotor varies by 13. In many respects this is similar gradients across the span of the rotor. The lateral wind speed across the starboard reference points. The vertical flow of fluid upflow resulting in a negative power requirement.1 : Power and Control Requirements. rotor is 23. The airflow and pressure and upwind to the ship. Starboard to Port negative power at the rotor.2m/s cross wind flow. Port to Starboard. Wind 30 Knots. Wind 60 Knots.3 Table 7.2-8 pressure gradients occur at the windward edge of the wind speed across the rotor is 65 knots.9 -2.3.8 28.1 %PH -32 -39 -30 VAvCms"1) 23.9 10. a free stream velocity of 30 knots.1 -20.5 ws 3.1. where the The flow solution for the rotor over the centreline of the downwash of the rotor travels down to the sea surface ship is shown in Figure 7. Position 1 2 3 P(kW) -21.3m/s between the port and shown in Table 7. not at the rotor.4.2 33.3. this is similar to the free stream The flow solutions within this section were obtained with velocity of 60 knots. which appears right to left in the figures.1.8m/s which is less than the undisturbed free stream velocity.4 7. The rotor operating in the outboard position is shown in Figure 7.0 13. The lateral .8 AW 2.2.8 -26. The solution shows that the flow separates at the windward edge of the helideck and the deck is covered in an area of recirculation.3.3. this is greater deck. shown in Figure 7. adversely to the power requirements as given in Table causing a large negative power requirement. variations are clearly dominated by the ship.1. The rotor is operating in 28.3.3.8 7.

The net effect of the up and down flow is a velocity.0 -7.3 shows the flow solution for the rotor over the rotor experiences the most dramatic vertical velocity the ship centreline.5 -9. shown in Section 7.8 -12.4.1 AW -16. Starboard to Port. The 60 knot wind cases indicated that the ship airwake was predominant and the helicopter thrust caused small disturbances by comparison. 2-9 7.0 -10.7 6. Position 1 2 3 P(kW) 8.0 %PH 13 10 0 VAvCms"1) -8. this is less than the free stream downwards. of the power exerted in the simple hover case. Regarding wind from the starboard side of the ship Figure 7. The lateral wind speed across component at the port and starboard end of the rotors is the rotor. indications of power and control requirements are evident.8kW.3. than the free stream velocity. which is only 14% of helicopter induced flow and from the results clear the power required in hover. The vertical flow is approximately constant across the span of the disc so the control requirements are limited.1. Wind 30 Knots.4. The variation between the vertical velocity zero power requirement. shown in Table 7. 7. The power exerted by the rotor These results combine the ship airwake and the within the solution is only 8. The flow pattern corresponding to the rotor in Position 2 is shown in Figure 7. The helicopter over the ship's centreline experienced local wind speeds greater than the free stream velocity.3 -13. separation at the windward edge of the deck is greater than that found in the previous flow solution. The angle of the gradients across the rotor span. which is an exact reflection of Figure gradients across the rotor in these cases.4.8m/s.3 6. In these cases the power requirements were negative.0 -6. In these cases the variation of vertical velocity flow across the span was actually less than those in the 30 knot cases.4.4. The rotor has a large These three flow solutions exhibit limited power region of low pressure above the rotor whereas there is requirements and lateral flow speeds but large velocity little pressure increase below the rotor. The three cases that considered the 30 knot wind from the port side all exhibited power requirements of around zero.4 3. With no side winds the helicopter rotor experiences slight power increases.1.9 WP -11.4.7 Table 7.2 above. The flow At the leeward end of the rotor the air is flowing across the rotor is a 8m/s. The upflow of the air over the ship causes the air . Conclusions the rotor downwash being 'fed' into the separation region above the deck.8 0.6 ws 5. indicating 8.9m/s. but there are negligible lateral flows to contend with.1: Power and Control Requirements.4.2. 16. once again the power requirements were minimal compared to that of the hover.3 : Rotor Position 3. is 10. The power exerted at the rotor is only a fraction to flow upwards through the rotor at the windward end. However the large velocity gradients across the rotor indicate high demands on pilot and rotor control. However Figure 7. this is less large.

1982. such as hover or forward flight. However this study demonstrates the viability of CFD in order to predict accurate flow solutions and resultant power and control requirements. This Behaviour of a Shipborne Helicopter During Rotor will. PH =power required to hover in still air. the Method'. In a three dimensional solution. lateral. T = total thrust of rotor. Ap = thrust/area at rotor. The geometries considered have only dealt with a lateral AW=WP-WS. p = pressure... [1] Plate E. and vertical velocity components.. 'An Introduction to two dimensional and some of the features exhibited in Computational Fluid Dynamics. Berlin. P = power exerted at rotor.J.v. For example. wind relative to the ship and therefore do not include any downflows due to the influence of the ship's 10. 'A Theoretical and Experimental longitudinally around the rotor. 'Engineering Meteorology'. . 9. 75th Fluid Dynamics Symposium. The Netherlands. AGARD Conference predicted from induced velocities at the rotor plane. A = area of rotor. The vertical velocity variations evident in the flow solutions are not found in any other Ws = vertical velocity component at 5m starboard of normal operations. in turn. PH PID = ideal power of actuator disc in hover. the air can travel [3] Newman S.13. Nomenclature All Units S.(ed). %PH= power exerted at rotor as percent of hover. Proceedings No. rotor centreline. The computational resources for a three dimensional study far exceed those of the two dimensional study undertaken presently. 1994. u. unless otherwise stated.J. The boundary conditions and rotor model employed for a three dimensional case are identical to those used for the two dimensional study. Oct. 552. variations around the ship and helicopter because of the two dimensional nature. References supers tructure. affect the power predictions which are Engagement and Braking'. Ch.K. VAV= average lateral velocity component at rotor. the power limitations. These results provide a qualitative measure of the flow Amsterdam. rather than remaining in Investigation Into The Rotor Blade Aeroelastic the same plane and being forced through the rotor. Elsevier. V = velocity scalar. 1995.2-10 This study indicates that the control requirements of the WP = vertical velocity component at 5m port of rotor helicopter are more likely to limit safe operations than centreline. Longman. In reality the flow is far from [2] Versteeg H.I.. The Finite Volume these solutions will be less prominent.w = longitudinal. upflow through the rotor when the rotor is upwind of the ship. Malalasekera W.

7 hazardous operation. The frigates typically carry one 'Professor or two SH-2G Seasprites or SH-60B Seahawks. 3-1 Higher Order Accurate Solutions of Ship Airwake Flow Fields Using Parallel Computers Lyle N. standard deviations. dating numerical models. Introduction Most wind tunnel tests include measurements made in the wake of a model ship exposed to a uniform veloc- Sharp-edged box-like ship super-structures create nu. predict.7'8'18 There have also been shape. Time average. and published in RTO MP-15.11 which gives the unstructured grid approach is most useful. structure of most modern ships is very complicated. PA 16802 lnl@psu. It is also difficult to measure the sity for ship air wake problems and MPI is used in the flow field accurately in the harsh ocean environment and NLDE solver. radar dishes. Un. . ity profile and almost zero turbulence level. antennae. held in Amsterdam.23 A wind tunnel investigation of the characteris- have been numerous incidences where the helicopter tics of the airwake behind a model of a generic frigate was blades have actually impacted the helicopter fuselage. On the other hand. many numerical simulation at. ing. Considering the complex geometry of real ship. In etc. All these experimental tests are crucial for vali- the flow (which may cause numerical stiffness). The flow around these obstacles is very difficult to this method we first solve for the steady state flow field. spectral densities and time correlations are presented for tempts have not been successful due to the inherently both velocity components for various position in the air- unsteady nature of flow and the low-speed character of wake. Modi * Department of Aerospace Engineering The Pennsylvania State University University Park. There CNRC. On the ^Research Associate frigate we are mainly interested in studying the hangar t Graduate Research Assistant deck area (aft portion of the ship). 1 and 2 show a frigate and an LHA. The Netherlands. tively. admittedly This paper presents a new method for simulating ship rather crude. Parallel computational methods are a neces. conducted. The parallel performance on various com. The mean some important information on real ship airwakes. in- and very difficult to predict. realistic test was conducted at NASA Ames in the "Ship- steady separated flow from sharp edges (and excessive board Simulator" with a neutrally buoyant atmospheric ship motions) make landing helicopters on ships a very boundary layer. velocity were made. and their airwakes vant research is building aerodynamics which shows the are very different also.9'10 In order to avoid correctly-scaled atmospheric boundary layer. The super- airwake flow fields. 5-8 October 1998. respec- several different approaches. ments on real ships.edu Summary general features of flow about blunt bodies of different aspect ratios. Measure- this and other engage/disengage problems. Long * Jingmei Liu t and Anirudh V. (NLDE) is fourth-order accurate in space and time. Research on ship airwakes has been conducted using Figs. course it is difficult to perform very controlled experi- tographs. These flows are inherently unsteady. These are very different ships. In addition. The simplest model of a ship. determining ments of streamwise and vertical components of airwake safe operating envelopes is very costly and time consum. The wind tunnel simulation incorporated a which is called a "tunnel strike". and the separated Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". is a sharp edged blunt body. Steady Geometrically precise studies are needed and have and unsteady results are presented for a generic frigate been done in wind tunnels.5 One of the sources of rele. the strong unsteady Another reference for simulations is that by NRC- flows can cause severe rotor blade deformations. The method presented cluding towers. then we solve for the unsteady fluctuations. full scale tests performed by the US Navy. One more merous aerodynamic and fluid dynamic problems. most ships. exhaust stacks. Of flow results are compared with oil flow visualization pho. in the presence of the strong electromagnetic fields on puters is presented also.

(2) inherently unsteady flow. In the structured grid and low-order method. preliminary attempts at high order accurate ship airwake predictions have been made by solving a steady flow field with a well-developed CFD method (CFL3D19) and a perturbation field with a high-order method. Very strong vortex ing ship airwakes. Saipan) and solver based on the 3D multi-zone. E are the viscous terms. helicopters The need for numerical simulations comes from the can land on many different locations on the deck. _j dF i 8G u dH — dR dS . We found in Stokes method. DDG51 was some cases the flow velocity ranged from 40 knots 12 feet chosen to validate an unsteady inviscid solver with an un- off the deck to zero velocity 3 feet off the deck. The methods that could accurately simulate ship airwakes. dE (1) w Figure 2: Tarawa Class LHA (length 820 feet. the at- mospheric turbulent boundary layer and the effect of the wind/ship speed ratio on the turbulence intensity. Nonlinear Disturbance Equations (NLDE) The methodology used here is based on the nonlinear disturbance equations. The large separated re- gions from superstructure sharp edges are quite difficult to capture accurately. The wind direction can vary a great deal.12'13 No method mid-section of the ship the very large island has a strong to-date has been entirely satisfactory for predicting these effect on the flow and tunnels the flow tangential to the flow fields. where F. it has been shown that the key features of ship airwakes are (1) a low Mach num- ber (about 0.) ers very attractive. On the LHA. which is a newly developed nu- merical method.S.05). beam dt 8x 8y dz dx dy dz = 132 feet). the wind conditions over rough seas have to be considered. such as. The results presented here will .3-2 flow that effects this region. thin-layer Navier- helped Kurt Long measure ship airwakes. and high cost of determining the safe operating envelopes for each of these can experience quite different flow fields. helicopters in a ship environment (and the huge test- The LHA's can carry 9 CH-53D Sea Stallion or 12 CH. and 6 AV-8B Harriers. forward portion of the deck is primarily influenced by There have been other attempts at numerically simulat- the separated flow off the deck edge. When this ratio is increased. ing backlog). It would be very useful to have numerical 46D Sea Knight helicopters. The complex ship geome- Figure 1: Oliver Hazard Perry Class Guided Missile try makes unstructured grid solvers and parallel comput- Frigate (length=445 feet. Flow Nature of Ship Airwake Simulation From previous studies. and (3) large regions of separated flow. island. beam=45 feet. The result is high-order-accurate 3D simulations.20 A US navy destroyer. and H are the inviscid terms and R.15 The general Navier-Stokes equations in a Cartesian coordinate system are: dgH. G. The airwake about a DD-963 ship sheets emanate from these edges.S. One of the authors configuration was simulated using a steady-state flow (Long) spent three days on an LHA (U. In addition. We also have some preliminary results for steady flow predictions from an unstructured solver (PUMA). the turbulence intensity will decrease and its spectrum will shift to a high value in the streamwise direction. S. since the air flow can impact the ship at any yaw angle (even 180 degrees). In this paper.

w'(e0+p0) + (w0+w')(e'+p') p0u' + p'u0 + p'v! p0v' + p'v0 + p'v' (6) The mean flow source term Q is time independent: p0w' + p'w0 + p'w' e' (dF0 dG0 ^dH0\ dR0 dS0 8E0 Q + On the left hand side of the NLDE are terms related --{~dx..+ -^ + -dz. The flow field is then split into a mean p0v' + p'vo + p'v' and a fluctuating part: pov0u' + p0u0v' + p'u0v0 q = q0 + q' (2) +p'v'u0 + p'u'v0 + (p0 + p') u'v' where p'v20 + 2p0v0v' + p' G'=( (8) +2pv'v0 + (p + p') v'v' P pu p0v0w' + p0w0v' + p'v0w0 9 = S pv (3) pw +p'w'v0 + p'v'w0 + (p0 + p') v'w' e v'{e0+p0) + +(v0 + v')(e'+p') and p0w' + p'wo + p'w' I rto+T «0 = Äryto q{t)dt (4) p0w0u' + p0u0w' + p'u0w0 +p'w'u0 + p'u'w0 + (p0 + p') u'w' Substitution of equation (2) into (1) and rearranging results in the nonlinear disturbance equations (NLDE): poW0v' + PoVoW1 + p'w0V0 H'=l 0) +p'w'v0 + p'v'w0 + (p0 + p') v'w' <V dF_ dO_ dH' _ dt dx dy dz (5) p'wl + 2p0w0w' + p' Where +2pw'w0 + (pop')w'w' . for tities F'. It also minimizes round-off error since we F' = < (7) +p'v'u0 + p'u'v0 + (p0 + p') u'v' are only computing perturbations. More p0u0w' + p0w0u' + p'u0w0 discussion on this new method is in the reference. CFL3D. p0u0v' + p0v0u' + p'u0v0 respectively. +2p'u'u0 + (p0 + p') u'u' This methodology allows us to use the most effective algorithms for the steady and unsteady portions of field. -dy.19 p'ul + 2p0u0u' + p' INS3D. We seek a solution of the perturbation variables q' p0u' + p'uo + p'u' with a known mean flow field which can be obtained from existing well-developed CFD codes (e.14 This type of boundary condition treatment allows one to easily introduce a disturbance at the incoming bound- ary by deriving an expression for one of the incoming .) for steady flow. Thus.15 +p'w'u0 + p'u'w0 + (p0 + p') u'w' Characteristic Boundary u' (e0 +Po) + (u0+u') (e' + p') Conditions for NLDE The boundary conditions for the NLDE are developed by applying Thompson's characteristic method22 to the nonlinear disturbance equations.. G' and H' are given as a laminar flow Q = 0. OVERFLOW.1 .+ -dz-)+-dx.g. 3-3 all be inviscid. Reynolds-averaged Navier-Stokes equation. PUMA. whereas the right hand side contains If the NLDE is time averaged. it becomes the strictly mean flow terms. where the The convective fluxes involving the perturbation quan. Reynolds stresses are on the left hand side. (10) to the perturbation properties and the cross terms (lin- ear and nonlinear). We can even use dif- ferent grids for the steady and unsteady solution.

It is 240 feet long. portant data for coupling the airwake solution to the this is a generic frigate model and it is shown in Fig. The whole computational domain is divided into many three di- mensional zones. This makes multi- block grid simulations and domain decomposition very difficult. parallel computers. So far we have been concen. On 8 Pentium II (266 MHz) cluster - 1R0 at incoming boundaries. from an aerodynamic point of view the craft carriers and LHA's with several helicopter landing airwake should still be representative of that for an ac- .4 times faster than eight 266 MHz Pentium IPs networked together. uration. 2. the SP2 costs roughly 28 times more than the PC cluster. On 8 R10000 nodes (195 MHz) 4 GB shared memory ship superstructure and ocean surface are both treated -.21 The time integra. 4. 5 dynamics analysis of the main rotor blades of helicopter in a computational mesh. Fig. For frigates the flow separation area sors. tail. and its dominant frequencies. 1EB with 100 Mb/sec networii 256 MB each node " boundary conditions are essentially non-reflecting. I 3MSPZ ä$U j : 60 The NLDE are cast in a generalized coordinate system i 40 and solved numerically using a finite difference based .. 140 128 MB each node : . a three dimensional single-block grid is used. Figure 4: Timings for a ship air wake case on SP2. ing helicopter. In this paper some preliminary simulations have been In the helicopter/ship interface problem. In order to make the code scalable and flexible.million grid points) even for a generic frigate test model. It was chosen because there are tions. the intensity of flow perturba. Such an approach some experimental investigations using the same config- is presented in this paper. Efficient computing performance is achieved by using a three dimensional domain decomposition strategy. such as. A 64-processor SP2 is roughly 2. MHPCC).3-4 characteristics with a source term. SGI Power Challenge and Pentium Number of processor II Cluster. tion is a fourth order accurate Runge-Kutta method. IBM SP2 1RM c-p. The Q. While a 24-processor IBM SP2 is 8. However. the ship geometry is very complicated. On 24 Power2 thin nodes (66MHz) .160 -. while on aircraft carriers and LHA's the deck leading edge vortex and separation are the key flow Results and Discussions phenomena. In fact. The discretized equations are solved in a time ■ \ t 2 3 4 5 accurate manner by taking advantage of computational PSU PSU PSU MHPCC NPACI aeroacoustics (CAA) methods. It is acknowledged that the ship superstructure trating on two types of ships: (1) frigates with helicopter does not resemble a typical frigate superstructure in de- landing pads on the deck behind the hangar and (2) air. wide and 55 feet high. done for a generic ship shape (TTCP ship). 4 spots on the deck around the control tower.86 million grid points using various number of proces. 3. As mentioned early. Penn State. the most im. (160 MHz. The spatial flux deriva. Methodology | r>T C hallenge 73 I. 45 feet are the mean flow field. Atmospheric bound- ary layer conditions can be incorporated in this manner : PC Cluster 1.6 time faster than behind the hangar cube has a strong effect on a land- a 16-processor SP2 (when problem size is kept fixed). The grid points are evenly distributed across each processor. Computer Site tives are calculated using seven point stencils of the fourth order optimized Dispersion Relation Preserving Figure 3: Timings for a ship air wake case on several (DRP) scheme of Tam and Webb. the . the IBM SP2 (e. Timing of NLÖE code at SP2 at NPACI (1. 5. At the outflow boundaries. SGI Power (160 MHz.86. The NLDE solver is implemented portably on paral- lel computers. thin node) 256 MB each node . : 5 20 25 *n scheme. On 24 Power2 Super Chips as hard wall boundary conditions.g. these two cases. thin node) 512 MB each node -. The code is written in Fortran 77 plus Message Passing In- terface (MPI)16 and is scalable in three dimensions. On 24 Power2 Super Chips Numerical Method and Parallel . 3. A comparison of code performance for the ship airwake run on various machines is shown in Fig. The air- gives the wall clock time for a ship air wake case with wake influences on the helicopter are quite different in 1. 30 40 50 Npaci.

7 show the contour plots of velocity magnitude Figure 6: Flow speed contours on the surface of TTCP on the surface of the TTCP ship for the zero and 40 ship with 0 degree yaw angle wind (CFL3D). clearly shown after each block. NASA Langley and Ames research centers have de. Fig. there is massive flow separation. the physical flow features are well captured fewer grid points than CFL3D. Comparing fig. we are concerned primarily with the inviscid phenomenon and used the Euler solver of CFL3D. In fact. aft section of the ship. structure. It is Special attention is given to the helicopter landing shown that the large region of recirculating flow extends area which is the square.75 feet above the flight deck. The TTCP ship computational domain is di- yaw ingle wind (m/s) vided into 10 blocks. grid stretching flow pattern from experiments. 10. The air wake is greatly influenced by both of them. the vortex pair in the higher plane is much close to the center line Mean flow simulation based on and the hangar. 11 the velocity vector on the flight deck floor is voted significant resources in the past decades to devel.24 The figure (a) is the results from CFL3D us- age from NASA Langley is used here to simulate the ing structured grids and the figure (c) is from another mean flow which will be given as a background flow solver (PUMA) using unstructured grids. of velocity magnitude at the ship's center plane. and flight-deck are compared with oil vi- walls close to each corner. Fig. 14. degree yaw angle wind cases from CFL3D results. proach. 10. ulated. In this mean flow simulation. The symmetry property of zero yaw angle flow is captured A case with 40 degree yaw angle wind was also sim- very well. 13 and fig. The former is heavily influenced by the Reynolds number.0 pack. The upwind-biased-differencing using the flux-difference- splitting technique is employed. There is over the flight deck and rises higher than the hangar. The latter is purely an inviscid phenomenon. 3-5 tual frigate since this study is concerned with the macro. where four distinct flow was used to enlarge the domain. which will lead to vortex shedding over the Fig. 12 (a) and (b) the flow patterns on the . This indicates that there is a horse shoe structured grids vortex as shown by the topological drawing in fig. The code discussed in the next section. by the simulation. the separation line is The three dimensional separation lines are clearly shown. The wa- ter surface is assumed to be a hard wall boundary. So far both mean flow regions are behind the hangar. with free vortices originating from the sharp corners. 4. fig. Our numer- The computational grid for this problem is 201 x 109 ical results are compared with a flow pattern obtained x 85 which results in a grid resolution of two feet or from an experimental study18 in fig. the NLDE needs much experiment. Comparing to the mensional interpolation. In the numerical plots. compared with flow visualization results for the TTCP oping modern CFD technology. 8 depicts the contour hangar wake. 6 and fig. A finite volume algorithm ferences in the attachment line and the position of the with a spatial-factored diagonalised. In fig. The numerical results on the front and middle ners and there are several reverse flow regions near the bridge-deck. The Mach number chosen for the simulation is a high wind case. independent of Reynolds num- ber. The CFL3D 5. 12. It shows the less in each direction around the ship. From the experimental results. a 20 feet drop down to the landing deck from the hangar This flow region is in the landing path. There are two types of separa- tion: one due to viscosity and the other due to sharp corners of the blocked structures. is very unsteady and yaw-dependent. ship. it is known that the flow is mostly separated. The photograph in the is a Reynolds-Averaged thin-layer Navier-Stokes flow middle is from oil flow visualization. The incoming flow speed is 41 knots. 9 shows velocity vectors in two horizontal planes deck causing landing approach hazards. After the blocked structures sualization photographs in fig.75 feet and 8. This three dimensional based on the structured grids and NLDE simulations vortex and reverse flow has very low speed but generally were based on the same grid in order to avoid three di. implicit scheme is vortex center. Of importance to the landing operation is the flow scopic flow properties and large scale phenomena in the condition over the flight deck.24 There are dif- solver for structured grids. which will be to the unsteady flow computation of NLDE. The flow is accelerated around the sharp cor. This is probably due to our inviscid ap- used in discretization of the partial differential equations.

ally have an abundance of memory. ther explicit or implicit time integration may be used. 15 the performance of CFL3D (a serial supports mixed. code may be run so as to preserve time accuracy for un- For this TTCP ship configuration it takes days to get steady problems.000 time iterations. In the next section.86 million grid points and of tetrahedra. ternal non-reacting compressible flows over arbitrary ge- ture is partially captured since the viscous effects become ometries. velocity. front part of bridge deck are very similar and close to For the unstructured grid flow field predictions we are each other since this part of the flow is less influenced using the PUMA code1 from Dr. little effort has been tion. which usu- results using unstructured grids are discussed in this sec. Primitive flow quantities (density. Chris Bruner (NAWC). PUMA is based on FVM (Finite Volume Method) and Finally in fig. MPI (Message Passing Interface) to ensure high porta- parisons are made for a 30 degree yaw angle case using bility and good performance. Comparing is a computer program for the analysis of internal and ex- fig. wedges. or may be run using a pseudo-unsteady the steady state results using an SGI Power Challenge formulation to enhance convergence to steady-state. PUMA is written entirely in ANSI C and uses stronger downstream. by the viscosity and our nonviscous simulation is more PUMA (Parallel Unstructured Maritime Aerodynamics) capable of accurately capturing the flow. The for a typical run it needs about 10. Ei- (Single Processor).3-6 100 400 Figure 5: Configuration of TTCP ship and computational mesh (1. Preliminary for distributed-memory parallel computers. the unstructured grid solver PUMA. 14 (a) and (b). the same com. We use 1. . the flow na. unstructured grids Because PUMA uses DMA (Dynamic Memory Alloca- While we have simulated relatively simple geometries tion) . pyramids and hexahedra. Since PUMA is targeted control tower on the LHA) are important. problem size is limited only by the amount of mem- here. more complex geometries (such as the 'island' or ory available on the machine.86 million grid points).topology unstructured grids composed code) is shown. 13 (a) and (b) and fig. and pressure) Mean flow simulation based on the are computed at the cell centers and saved on exit.

not including message passing overhead. . The numerical results on the front and mid- main decomposition parallel technique. significantly by compiling the code using single precision floating point variables for which an option is provided. For tetrahedral grids. 3-7 B. 16. in fact. They The high wind speeds relevant to the ship/helicopter are intended only to give approximate envelopes for the interface problem arise from storm centers far from the different regions and provide the background flow for actual ship and are called neutrally stratified. study. such as the vortex location after the surfaces. The flow field is generally very unsteady. PUMA needs 582 bytes/cell and 624 bytes/face. Perturbation simulation We ran the TTCP ship in PUMA by converting the As mentioned before.080 cells and 1. this mean flow.769. the flow pattern update the value at wall grid points. Characteristic boundary conditions are used at experimental results. The TTCP ship is divided into 88 solid The flow pattern at all three locations are similar to the boxes. At each the island in fig.837 nodes used in PUMA. Cur. The . Figure 10: Topological drawings from experimental placed on reducing PUMA's memory requirements. This wind NLDE simulations. and flight-deck are compared with this difficulty. There are large fluctuations about condition is considered at our inflow boundary. 17 and the three dimensional separa- time step. shown represents a very rough mean of the flow.. There are ship. I 3.18 rently. NLDE has been developed for structured grids (used in CFL3D and NLDE code) to solving more complex geometries.J flow speed ato dag. with double precision floating point variables used throughout the code. which allow multi-solid-boxes inside the compu- 1.3 ^BHI Figure 8: Flow speed contours on center plane of the TTCP ship (CFL3D). tion line is very similar in fig.7S feet tram flight deck !fiow direction! Hangar. such as the TTCP unstructured tetrahedral grids for PUMA. im Figure 9: Flow velocity vectors at two horizontal planes Figure 7: Flow speed contours on the surface of TTCP over flight deck (CFL3D). NLDE code. the results from NLDE simula- (or 250 words/ cell). after the single block computation is finished. This makes the boundary condition A case with 30 degree yaw angle wind is simulated us- implementation difficult. This requirement can be reduced tions are discussed based on this mean flow. 18 the solid box wall boundary conditions are applied to Based on the discussion above. especially combined with a do- ing PUMA. 17 and 18. this amounts to 2002 bytes/cell In the following section. y»w sngfe wk*$ (roi) A •^ n: I 9-9 - ■ 46 ^TVfl ^^■■feAftk. ship with 40 degree yaw angle wind (CFL3D). tational domain. edges and corners of these boxes.718. To overcome dle part bridge-deck. a single block domain was chosen for the oil flow visualization photographs in fig.

/ («0 Figure 11: Predicted surface flow velocities compared to Figure 12: Predicted surface flow velocities on the front surface oil flow images on the flight deck. Since this is a zero yaw angle case. it is shown that large perturbations occur around ations. cc -03 (64 bit) 500 577 179. the transverse pertur- ity perturbations u' in the center plane at two different . From these unsteady re- (or integral) length scale of the turbulent velocity fluctu. v/y//^////y/W/V///J/ 7/ }j/oo</oooy/////f///i'/ii it ■V <>•>•/"/■>/■ v (■' y // </ Vj // ij i (a) Flow Direction (b) ry^'. In the field far from function of the mean wind speed. 24 incoming disturbance is determined by the turbulence the vertical perturbation intensity Vw^ is given in the intensity. (2) the turbulence intensity. High instantaneous vertical perturbations trum by using a random walk (random phase) technique. elevation and rough. 19 and 20 give contour plots of longitudinal veloc. (3) the longitudinal dinal velocity intensity VV . contour plots of vertical and trans- duced from the inflow boundary. Its spectrum is obtained from the wind spec. In fig.21 SP2 IBM Power II mpcc -03 (pwr2) 500 736 140. time-averaged over an appropriate time steps.3 the hangar can be clearly observed. In fig.3-8 Computer Architecture Nodes Compile option Iter See MFLOPs grimm P-II 266 cluster mpicc -04 500 983 105. sults. The magnitude of the verse perturbation velocity v'. 21 gives contour plots of the longitu- scale. (c) PUMA with 40 degree yaw angle wind. are found in the region just after the hangar trailing edge.52 power SGI Power Chall. center plane. Incoming characteristics with source terms are intro. Vortex shedding from ness length scale. (b) oil flow visualization principal parameters of the free stream airflow are (1) the mean wind speed. especially in the area after the items 74030.74031) for the above four parameters as a hangar and after the leading edge. w' are shown.24 (a) CFL3D. the ship the flow is quite steady.24 Table 1: Some Performance Comparison of PUMA on Different Computers 24 m/s 24 m/s ::'. Fig. part of bridge-deck compared to surface oil flow images (b) oil flow visualization. Empirical relationships are available (ESDU data the TTCP ship structure. Fig.24 (a) CFL3D. 22 and 23.

detailed experimental data is not yet available. "Ship to experimental results. By comparing [2] Carico. and show that the key flow phenomena bation is quite weak over the flight deck. "Parallelization of the Euler Equations on Unstructured Grids". From those preliminary results. We gratefully acknowledge ONR Grant No. Future work will concentrate on more detailed com- steadiness is important around the helicopter landing parisons to experiment. the inclusion of more geomet- deck.W. [1] Bruner. 3-9 «— 24 m/s _ 24 m/s ^ \ \ \\\\ I \ \\\'V 11 \ '/'/'/'/'/'/'l'l'l l X \ 1 s f 's -s v v •!/ '!■' v ■•/ // // (a) Flow Direction Flow Direction j BK«j»:*-. bance equation solver has been developed using parallel AIAA 97-1894. throughout the domain but in particular the flow un.W. D. and the inclusion of viscous intensity in a horizontal plane 17 feet above the deck.. The parallel performance of the code has been compared on various computers. our present results are qualita.j-. By comparing can be captured using a steady-state code followed by the instantaneous perturbations in fig. N00014-97- In the meantime.. effects. where the helicopter rotor would be. Reddy. Canada) and Kurt Long (USN Pax River). the the NLDE code.-i. B.. (b) oil flow visualization flow visualization tively correct. Fig. the unsteady features Acknowledgments of a TTCP ship air wake are captured qualitatively. Figure 13: Predicted surface flow velocities on the middle Figure 14: Predicted surface flow velocities on the flight- part of bridge-deck compared to surface oil flow images deck compared to surface oil flow images with 40 degree with 40 degree yaw angle wind. such as a The unsteady three-dimensional flow is of interest real ship shape. Airwake Measurements and Modeling Options for .. Walters. Concluding Remarks References This paper presents steady and unsteady flow field predictions for frigate class ships.. computers.'. and Dimarzio. complex geometry and practical simulations. C.S. 19 and 23. and R. A nonlinear distur. We would also like to thank Steve Zan (NRC prepared for quantitative evaluation and analysis. (b) oil yaw angle wind. C.24 (a) CFL3D. the NLDE code is being improved and 1-0530.24 (a) CFL3D. However. 25 presents a contour plot of perturbation rical features of the ships. perturbation length scale of longitudinal perturbations The unstructured grid approach can be used for more is different from that of vertical perturbations.

Washington. March 1987. Keller.R. "Dynamic Interface Testing and the Pilots Rating Scale. 12. for Shipboard Engage/Disengage Operations. Bangalore A.M. [12] Landsberg." 13. pp. J. American Helicopter Society Theoretical Correlation of Analysis for Heli. 54th Annual Forum... 1992. M. W. Val Healey. Or- [6] Johns. Vertica.K. 1996 ceedings of the 52nd American Helicopter Society Forum. July-Aug. 1988. and Newman. and E. AGARD CP-509.M. [11] Williams." NRL/MR/6410-96-7898.. A.. "Establishing a Database for Flight Forum. Sandberg. VA. 74031.N. Young. D.C.3-10 z 140 Tlmtnn fnr TFP SP - a." Pro. 559-564.N. [14] Liu." Naval Engineers Jour- nal. J. T.J." Journal of Aircraft. Washington. "An Efficient. May. Smith.P. copter Blade-Droop Stop Impacts..J.. and Boris.. and [9] Geyer. 267-278. [3] ESDU Data Items 74030. Rotorcraft Applications.R.. and J. Boris. Val. May. No. Parallel Method for Solving Flows [8] Hurst.3. E HE» S Q. DC.. pp. [4] Funk." Naval Engineers Jour- nal. Washington. and Wang Q.G. (b) oil [5] Healey. Engineering Sci- ences Data Unit International. Long L. 45-63. "DDG-51 Flt-IIA Airwake Study.C. 1998. and E. "Validation J. = | U 20 E a II c 0. S.C. K. W." submitted for publication in the Proceedings of 38th [15] Morris P. A. D." In: Aircraft Ship Opera- tions. Vol. Computer Name Figure 15: Timing for CFL3D on several computers. January 10- Prediction of Helicopter Rotor Blade Response.. with 30 degree yaw angle wind. Val. Struc- tural Dynamics. NV. L. AIAA/ASME/ASCE/AHS/ASC Structure. Smith. April. 1994. J. Part and Application of a Transient Response Analysis 2: Hangar Interior Flow. T..Jr.. 1994. 1996. Vol. "The Prospects for Simulating the flow visualization Helicopter/Ship Interface. 120 132 ~: «a> J= 100 IBMSP2 - 5 SGI 9Z Power Challenge 3 »« SI fiSt B2 > - ii 60 67 E S 111 S [ £ 40 o_ ~ 3 ■ox ." The 53rd AHS [7] Healey. 29. Egolf.C.24 (a) PUMA. pp. "A Comparison of Heli- copter Rotor Inflow Predictions in Turbulence using Blade Element and Vortex Lattice Models.C. Reno. No. and Materials Conference.4. "Wind Tunnel in Complex Three-Dimensional Geometries". McLean. W." Pro- Figure 16: Predicted surface flow velocities on the front ceedings of the 50th American Helicopter Society part of bridge-deck compared to surface oil flow images Forum. November 1991. J. S.P.^ . June.. April. in the Wake of Structures. A. and T. Young.. DD-963 Class Destroyter. 1997. J..R. 32nd Measurements of Ship Induced Turbulence and the Aerospace Sciences Meeting & Exhibit. Nov.. J. D. "The Airwake of a lando.. J. "Experimental / Interface Problem".D. 1997. "Higher Order Accurate Ship Airwake Predictions for the Helicopter/Ship [10] Keller. 1989. [13] Landsberg.. "A Parallel Three-Dimensional Computational Ae- . May. and Long. and Long.

L. Vol. 1998. of Computa- 1997 and AIAA 97-1598. and Healey. J. July-August 1992.V. Phase I: Methodology Development.R... .C." NASA Lan- gley. [20] Tai.W. and Webb. ». Institute for Aerospace Research. [23] Zan. J.C. 626." Research Report sub. and C. tional Research Council of Canada. (b) oil flow visualization flow visualization [21] Tarn C.. School of ments of the Airwake Behind a Model of a Generic Aerospace Engineering. "Ship Airwake Modeling and Simulation.." J. tional Physics. J. "Parallel Programming with MPI. \ \ \ S.A. 29. "CFL3D User's Manual (Version 5.0). J." Morgan Kaufmann. "Time-Dependent Boundary Conditions for Hyperbolic System". Geogia Institute of Tech. tational Physics 89.N.J. Na- Aircraft. S. Frigate. P.. pp. Computational Aeroacoustics.. nology.S. (b) oil yaw angle wind. M.24 (a) PUMA. 1990.v \ \ \ V N. "Flight Deck [24] Zan.J. Unpublished data from the Aerodynamics Aerodynamics of a Nonaviation Ship. Biedron. 1994. Vol. San Fransisco. J. Val.4.. and E. R. P. 1997 [22] Thompson K. M. 133. 1996.. [19] Krist...1.. Mavris. D.. Computational Physics." Journal of Laboratory.L.K.W. Nov. and Carico. "Wind Tunnel Measure- mitted to Naval Air Warfare Center. T. and Schräge. 3-11 ** \ v \ •.T.. Vol. June. "Dispersion- rocoustics Method Using Non-Linear Disturbance Relation-Preserving Finite Difference Schemes for Equations". 1993. 619." AIAA 93- 3002. August 1992. Atlanta. Rumsey. Garry. D. >. of Compu- [17] Prasad." NRC-CNRC Report LTR-AA-13. S. S. [16] Pacheco. No. [18] Rhoades.24 (a) PUMA. D. "Simulation of DD-963 Ship Airwake by Navier-Stokes Method. •* \ \ N (a) \%mk Figure 17: Predicted surface flow velocities on the middle Figure 18: Predicted surface flow velocities on the flight- part of bridge-deck compared to surface oil flow images deck compared to surface oil flow images with 30 degree with 30 degree yaw angle wind.

>"*"' VWfgTVSF? Figure 21: Contour of longitudinal velocity intensity.4 1. -4. .2 -2.6 2.3-12 u'(m/s) Figure 19: Contour of instantaneous longitudinal velocity perturbations at t = h (NLDE).8 3. 6 u'(m/s) 4.4 -3 6 •.8 JBiillllii -6 Figure 20: Contour of instantaneous longitudinal velocity perturbations at t = t2 (NLDE). .2 0 -1.

3-13 6 v' (m/s) 4.6 *».2 . 6 w' (m/s) 4.8 -6 Figure 23: Contour of instantaneous vertical velocity perturbations (NLDE). Figure 24: Contour of vertical velocity intensity (NLDE).2 0 -1.6 2.4 1.6 -4.4 1.2 0 -1.2 ■I -2.4 -3.8 3. .8 Figure 22: Contour of instantaneous transverse velocity perturbations (NLDE).8 1 3.4 -3.6 2.-2. j^ -4.

.. ürms Cm/si Figure 25: Contour plots of perturbation velocity intensity in the rotor plane of helicopter (NLDE).'. ..3-14 i. ..

Tai and Carico [4] applied three- controllability in the final approach and landing is dimensional. and heave). the Stokes equations. Tai [6] simulated the airwake flowfield Problems of Vehicles Operating near or in the Air-Sea Interface. namely an air burble. turbulent airwake from the ship's dynamics to provide some analysis of the problem. the flow is largely separated viscous-inviscid interactions. USA 1. usually of high frequency and diminish fairly Typical results in the form of particle traces and the quickly. including gusts and fluctuations due to It is well known that the interface between the ship motion (pitch. implicit finite-volume type Reynolds- greatly influenced by the ship airwake. At present. Major flow features sharp corners or edges are generally referred to as including viscous-vortex interactions observed unsteady flow. Previously. superstructure [1. Carderock Division Code 5300. surrounding the A multi-zone. and published in RTO MP-15. Tai* Naval Surface Warfare Center. The interface environment Nevertheless. options are open to model the airwake flow either by the unsteady Euler type inviscid The technical issues involve aspects concerning solution or by the steady-state solution to the Navier- both the air vehicle and the ship. Current unsteady airwake. Marine and Aviation configuration. wind speed of 12. held in Amsterdam. simulation. Later. and the final approach and landing represent the actual geometry. Associate Fellow AIAA. their mean values can be captured velocity field over and aft of the ship are presented. including recent unsteady computational need to often alter the ship's course into the wind to schemes that also neglect these unsteady sources. 9500 McArthur Boulevard West Bethesda. These aircraft and ship poses problems in operations and sources of unsteadiness are real causes of an survivability of naval aircraft. 5-8 October 1998. reasonably well by the steady-state solution of a vortical flow. The critical unsteady flow in a ship 2. In case of unmanned air vehicles (UAV). Amsterdam. October 1998. in fact. For aircraft. INTRODUCTION airwake is. Either option ignores the lift characteristics become very sensitive to the unsteadiness due to atmospheric turbulence and ship surrounding environment due to relatively low motion. SUMMARY The ship airwake is defined as an arbitrary volume of air. Free vortices with a Reynolds number of 221 million based on from the ship top-side geometry generated by the ship length.. the airwake consists of mainly the steady zero. roll. and vortices due to behind the superstructure. The Netherlands.87 m/s (25 knots) at a direction of Physically. the steady-state Navier-Stokes freestream velocity and the use of high lift solution has offered useful data in ship airwake devices. Computed viscous flow results have Department. The flow is fully turbulent flow with some unsteady components. demonstrated good correlation with measurements Presented at the NATO RTO Symposium on Fluid Dynamics [5]. the complexity of the problem becomes more complicated by the presence of requires the use of modern computational fluid massive. 2]. about an LPD ship with rather complex surface Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". establish necessary wind-over-the-deck conditions for a successful landing. thin-layer Navier-Stokes method is ship. 4-1 Simulation and Analysis of LUD Ship Airwake by Navier-Stokes Method Tsze C. These unsteady components are experimentally are captured in the simulation. The ship's superstructure is disturbances caused by the ship that are perceptible modelled with blocked structures that closely to the pilot. 10 and 20 degrees. Bldg 4E Carderock Div. The effect of airwake on the aircraft/ship utilized to investigate the airwake about an LHD interface operation is determined by the airflow ship configuration. but are unfortunately often aircraft/helicopter operations aboard ship are ignored in both theoretical and experimental characterized by restrictions due to weather and the treatments. averaged Navier-Stokes (RANS) scheme to consider the airwake problem about a simplified DD-963 ship * Senior Research Scientist. The freestream has a patterns required for shipboard operations [3]. MD 20817-5700. In general. attributed to atmospheric turbulence. . The Netherlands.

Details are given by Thomas et al [10]. the one-equation models. 1 . the CFL3D code [10] with multi-zone capability. development.5 ship lengths from the ship centerline. Reynolds- averaged Navier-Stokes equations. SIMULATION METHOD bow of the ship. The completeness. 3. basic grid generations. The former determines proper length scales for separated flows while the latter uses curvilinear arc length in Fig. and O-type in multi-zone. A cylindrical grid topology is adopted for its capability to treat a body with a sharp The present effort is an extension of previous nose. body-conforming grid freestream involved in the present work. an increased mesh scheme called the area-volume rule for representing density is needed. three directions. along with the standard two-equation models. the Baldwin and Lomax model [11] with the Degani-Schiff modification [12].The numerical LHD ship model place of normal distance. along with the extension by Tai [13] is used. including the basic Baldwin and Lomax algebraic model. The code is upgraded with a variety of turbulence models. The code is based on a finite volume algorithm with a spatially factored diagonalized. 4 and 6 is the crossflow plane.1 Grid Generation and Grid Topology direction. Figures 1 and and advantage of the grid topology adapted are 2 show the actual ship and resulting numerical presented in Ref. the . set at 2. is used as the basic flow solver. This number is larger than those used for analysis of aircraft because of the very low speed A structured. The numerical model for the surface work [9] indicates that when using a compressible grid is derived by using the Tai simplification flow code at low Mach numbers. Although simple. 7. or the computational An overall coarse grid is generated first by method employed has been described previously. and 3. work on the DD-963 and LPD ships. among others. The most forward plane is set at 1.0 ship lengths for the wake. Appropriate modifications to the code for applying specific boundary conditions are implemented. The method is globally second-order accurate and well suited for patched grids in a multizone domain. In the latter. radial distances are then clustered near the surface It contains two main elements: 1) the grid generation and stretched in the outer region for shear layer and 2) the flow solving. The details of the grid generation the surface components of a ship [8].0 ship length from the 3. curvilinear. Previous is generated. implicit scheme for discretizing the three-dimensional. The upwind- Fig.LHD ship configuration. In the present work. the cargo door and elevators enhanced by increasing the number of points in all are closed and the antennas are neglected. The outer cylindrical surface is employed with an enhanced grid in the simulation. For 3DGRAPE using a multi-block procedure. The grid resolution was models. RANS scheme used in Refs. The complete grid has a total of 187 x 123 x 57 points with 57 points in the radial 3. 2 . The topology is basically an H-0 mixed type.4-2 geometry representation and obtained reasonably The NASA Ames 3DGRAPE code [8] is used for useful steady-state flow results. The same with H-type in the longitudinal plane. the method is briefly described here. biased differencing technique is used for the inviscid terms and central differencing for all viscous terms.2 Flow Solver The NASA Langley thin-layer Navier-Stokes code. The simulation method.

Of course the the ship length for the full scale ship. atmospheric boundary layer is approximated by a power-law profile: 4. The large number of iterations is due to upstream. and 20 degrees with a speed of 12. on the rear flight deck. The height above Reynolds number would have to be reduced can be twice as much as shown here in about three approximately by a factor of 8. The waveless. 4 and 5 give the profile At the downstream boundary condition. Whichever way the ship lengths away. the static even keel. = (z/h)n (1) Various views of the particle trace of the streamlines emanating from various stations on the where h is reference height and n varies from 0. This approach was reached after stern. RESULTS AND DISCUSSIONS airwake. This assumption independent of the Reynolds number. This is confirmed by the astern turbulent flow range. remain unmixed with the water.3 Boundary Conditions Center.10 to ship surface are shown in the following three pages 0. Both types of corresponds the water surface as neither viscous nor separation create free vortices as evidenced by the inviscid. The flow condition of the air burble (airwake) which might be yields a Reynolds number of 221 million based on misleading in the perspective view. Thus.87 m/s (25 knots).12. The use of the Navier-Stokes method ensures the basic physical features of the 4. the flow falls into the right aft of the stem. 3) characteristic form of inflow-outflow compressible flow solver. volume blocks representing the superstructure. 2) atmospheric pressure recovery the very low freestream Mach number used in the downstream. The required CPU time is at the cylindrical outer boundary. If the ship's height. instead and astern views of the same airwake over the ship at of imposing the usual freestream recovery. The use of characteristic form of inflow-outflow boundary condition at the cylindrical There are two types of separation: one due to outer boundary is known to improve convergence of viscosity and the other due to sharp corners of the solution at low speed. . namely wind at zero. of the air burble beam is used for the Reynolds number basis. The mechanism of the vortical flow involving some other attempts at treating the water boundary free vortices is well demonstrated by the present failed. with free with velocity components being extrapolated from vortices originating from virtually all sections. a reflective boundary is applied and streamlines rolling up forward and aft of the the vertical velocity component must vanish at the superstructure. Further.104 ship length. and aft of the water surface. The former is closely influenced by the types of flow The water surface is assumed to be flat and involved which is Reynolds number dependent. the grows as the flow proceeds downstream. Figure 3 shows the perspective view of the airwake particle trace. h = 0. The airwake experiences a dip Reynolds number is calculated. view shown in Fig. 4-3 model has been regarded as the best turbulence All the computations were performed on the model for flowfield dominated by vortical flow. namely the three-dimensional flow separation which occurs over all the geometric Numerical results of the flow over a LHD ship surfaces. deck above the water surface. and the flow properties in the airwake latter is purely an inviscid phenomenon.14 [14]. while Figs. as well as the overall size. are captured. configuration subject to an atmospheric wind speed of 12. 5. Cray facilities at the DoD High Performance Computing facility at the Naval Oceanographic 3. The h is set to be the height of flight for three wind directions.87 m/s (25 knots) and a wind direction of 0.000 iterations (time steps) The boundary conditions for the Navier-Stokes requiring approximately 8 to 9 hours of Cray C-90 flow solver are: 1) atmospheric boundary layer flow CPU time. It turns out to be a good approximation. Even at zero wind angle. 10.1 Particle Trace U/V«. steady-state solution. the figures all atmospheric pressure condition is satisfied along indicate that the flow is mostly separated. and 4) viscous reasonable and affordable in today's environment of nonslip flow at the surface of the ship. Converged solutions were obtained in about 8. and n is set to its mean value of 0. the interior. The astern view also depicts how the airwake grows in size and shape. The computer capability and resources. The profile view helps indicate the actual height 10. and 20 degrees are obtained.000 to 9.

4-4 Fig.Particle trace over LHD ship at even keel at 12.87m/s (25 knots) and zero wind angle. . Astern view. 5 .

Perspective view. Front view.Particle trace over LHD ship at even keel at 12. 7 . .87 m/s (25 knots) and wind angle of 10 degrees.Particle trace over LHD ship at even keel at 12.87 m/s (25 knots) and wind angle of 10 degrees. 6 . 4-5 Fig. TOUCH-DOWN POINT FLIGHT DECK Fig.

Particle trace over LHD ship at even keel at 12.Particle trace over LHD ship at even keel at 15. 8 . Perspective view.4-6 Fig. Fig.87 m/s (25 knots) and wind angle of 20 degrees. 9 . . Astern view.44 m/s (25 knots) and wind angle of 20 degrees.

The region of separated flow is generally restricted to the starboard side of the ship where the superstructure is located. Note that streamlines near the edge of the deck do not peel off from the deck at this 10-degree wind angle. the However.Particle trace over LHD ship at even keel at and FFG-7 [15]. see Fig. On the flight deck (in the rear of port side). see Fig. TOUCH-DOWN POINT As a consequence. and LPD [6].2 Velocity Distribution low pressure region created behind the large superstructure on these ships. Figures 8 and 9 show the perspective port-side plane 30 feet from the ship's centerplane. the flow seems to be unseparated as shown in Fig. depends on the ability to a low pressure region behind the superstructure provide the correct velocity distributions needed to which can influence the rear portion of the flight define the flight envelope for aircraft shipboard deck. 4-7 the airwake is highly asymmetric even at zero wind angle because of one-sided location of the superstructure. and the other about one-ship length aft of the touch-down leeside streamlines may in fact depart from the edge location. including the touch-down area. 7. 8 and VELOCITY LINES the leeside streamlines fall off the cliff. operation. indicated the leeside streamlines Front view. 10 . 11. This is in contrast to one's intuition that the touch-down area is in the shadow of the superstructure (due to wind angle) and therefore should be in the separated flow region. More vortex rolling up takes place aft of the stern as a result of more severe flow separation generated by viscous-inviscid interactions. The particle traces offer is an indication of whether the simulated airwake contains the right Although the flight deck of the LHD ship is not physics it should have. this phenomena occurs up to certain wind velocities U. see Fig. on the other hand. Rolling up of the vortex flow intensifies as depicted in Fig. The reason for swerving inward is because of a 4. The reason for this lies on the fact that the flow actually experiences mild acceleration due to the narrowed flow path between the superstructure and the streamlines along the edge of the deck. one directly at the touch-down location and the the separation becomes extremely massive. These lines lie in a of the deck. The asymmetry of the airwake becomes more evident at a wind angle of 10 degrees. It indicates that the characteristics of the airwake flowfield of the LHD configuration is very much wind-angle dependent. however. As the wind angle increases to 20 degrees. the flow on the rear portion of the flight deck including the touch-down area becomes separated. Previous work. see Fig. lines. 11 . 6. 7. W are examined at two vertical angles. both experimentally on DD-963 [5] Fig. The usefulness of the located right behind the superstructure. V. there is still resulting airwake. 10. 9.87 m/s (25 knots) and wind angle of 20 degrees. and computationally on DD-963 [4] 12.Geometry of velocity lines . Fig. would even swerve slightly inward in the flight deck area. and astern views of the particle trace over the LHD ship at a wind angle of 20 degrees. as defined in Fig. For flight envelope analysis.

The deficiencies at low altitude. touch-down. ß = 0.1 coarse to cover the clustered spacing inside the boundary layer. see Fig.Longitudinal velocity distribution above Fig. at the edge of the boundary layer) the U velocity has a magnitude of 41 ft/sec (12. These velocity components are taken from interpolated rectangular meshes that actually are too £-0. 13. due to the height of the superstructure involved. zero as the altitude increases. has Larger variations in velocity components are slightly negative values (from starboard to port) at observed along the line about one-ship length aft of low altitude. W. transverse. ß = 0. the longitudinal velocity U exhibits large these values are outside the boundary layer. then it touch-down. V.87 m/s (25 knots).87 m/s (25 knots). It recovers with a steep variation is rather small because of the zero wind ascent between 50 to 150 feet.5 m/s) -0. Again. for the case of 12. The vertical velocity.4-8 The longitudinal. and 0. Completely out of the 100 feet from touch-down. 500 1000 1500 500 1000 1500 Vertical Distance (ft) Vertical Distance (ft) Fig. The reason for changes in the (downward) in the entire altitude range before it velocity slope above 150 feet level is believed to be vanishes. 14.1 magnitude in feet per second. The flow at this point is -0. see Fig. V. 12. and then gradually angle. The coordinate gives the velocity 0. 15. = 12. 13 . see Fig. 13.36 m/s) Fig. The velocity in the transverse direction. It then diminishes to narrowed flow path and almost totally in the wake. 12.2 speaking. 14 .87 m/s). monotonically approaches the freestream value.4 500 1000 1500 apparently unseparated as discussed in the previous Vertical Distance (ft) section on particle traces. VM = 12. starts at a slightly approaches the freestream value as altitude positive value (upward) and then becomes negative increases.Vertical velocity distribution above touch-down. V.87 m/s wind with a zero wind angle.87 m/s (25 knots). see Fig. W) on the vertical plane directly at the touch-down point are plotted in Figs. Under the influence of nearby wake behind the superstructure. ß = 0. = 12. V«. At the touch-down point.Transverse velocity distribution above at a height 100 feet above the touch-down.3 which is very close to the freestream velocity of 42 ft/sec (12. while the abscissa measures the vertical distance from the touch-down Ü 0 point. the U velocity gradually decreases to 34 ft/sec (10.2 14 respectively. but gradually becomes positive above the touch-down point. (strictly >-0. and vertical velocities (U. 12 . ..

16 . VM = 12.87 m/s (25 knots).. the air burble will double surrounding. 18 .87 m/s (25 knots). In general. plane one-ship length aft touch-down. the flow has low velocities in the primary source for the growth of the air burble. the imposed. The area with velocity its size about two-ship lengths aft of the ship stern.87 m/s (25 knots). upward magnitude is considerably higher than those however. i. V and W.87 m/s (25 knots) at a wind from the wake.e. the aforementioned sections. i. see Fig. Vm = 12. 17 . 17. 4-9 o "3 1 > 500 1000 1500 0 500 1000 1500 Vertical Distance (ft) Vertical Distance (ft) Fig. The location where the change of angle of 10 degrees are given in Fig. sy 500 1000 1500 Vertical Distance (ft) Fig.. ß = 0. certain area exceeds the freestream wind speed For the vertical velocity shown in Fig. Similar to U and V. velocity slope takes place is consistent with the the flow in the windward side has high speed that in longitudinal velocity.Mach contours at four longitudinal stations plane one-ship length aft touch-down. there is a change of W velocity This confirms the growth of the airwake discussed in slope at about 150-feet height. ß = 10 deg. As central region and high velocities in the observed experimentally. VM = 12.Transverse velocity distribution in the Fig. ß = 0.87 m/s (25 knots). ß = 0. the Mach contours exhibit the wake in the touch-down plane.e. = 12. 18. 18. . at about 150-feet level. deficiency grows as the flow proceeds downstream. respectively. Figures 16 and 17 show the transverse and Mach Contours vertical velocity components. At some distance behind the ship. 15 ■ Longitudinal velocity distribution in the Fig. This upward velocity is the characteristics.Vertical velocity distribution in the plane one-ship length aft touch-down. Here the V values are mostly negative (starboard to The Mach contours four longitudinal stations for port) with increased magnitude due to influence the case of V„ = 12. V«.

4-10

**Lastly, the Mach contour is presented at a plane 6. ACKNOWLEDGMENT
**

that runs over the ship's superstructure for the case

The present work was supported by the Naval Air

of Vra = 12.87 m/s (25 knots), ß = 10 degrees, see Warfare Center, Aircraft Div. (Richard Huff, technical

Fig. 19. On the fore deck, the flow closely resembles monitor) and the Research Venture Program at NSWC,

that of a flat plate until it reaches the superstructure Carderock Div. Hie DoD High Performance Computing

represented by blocks. Over the superstructure, the facility at NAVO provided the Cray CPU time.

flow undergoes deceleration with very thick shear

layers. The local Mach number exhibits large 7. REFERENCES

deficiencies between the two large structured blocks,

behind the end structure, and immediately behind [I] Garnett, T.S. Jr., "Investifation to Study the

Aerodynamic Ship Wake Turbulence Generated by a

the stern that are usually observed in the backward

DD963 Destroyer," Report No. NADC-77214-30, Boeing

facing steps. Thus, the flow in these regions can be

Vertol Company, Philadelphia, Oct 1979.

characterized by that of a backward-facing step: [2] Healey, J.V., "The Prospects for Simulating the

having massive flow separation involving reversed Helicopter/Ship Interface," Naval Engineers Journal, Vol.

flow accompanied by circulation. 99, No. 2, 1987. pp.45-63.

[3] Carico, D., Reddy, W., DiMarzic, C, "Ship Airwake

Measurement and Modeling Options for Rotorcraft

Applications," AGARD Symposium on Aircraft Ship

Operations, Seville, Spain, May 1991.

[4] Tai, T.C. and Carico, D., "Simulation of DD-963 Ship

Airwake by Navier-Stokes Method," AIAA Paper 93-3002,

Jul 1993. Also Journal of'Aircraft, Vol. 32, No. 6, 1995.

pp. 1399-1401.

[5] Healey, J.V., "The Airwake of a DD-963 Class

Destroyer," Naval Engineers Journal, Vol. 101, No. 2,

1989. pp. 36-42.

[6] Tai, T.C, "Simulation of LPD Ship Airwake by

0.003 Navier-Stokes Method," Proceedings of the Sixth Asian

Congress ofFluid Mechanics, May 1995, Singapore

Fig. 19 - Mach contour in a longitudinal plane [7] Tai, T.C, "A Single Structured Grid for Complex

Ship Geometry," in Numerical Grid Generation in

VM = 12.87 m/s (25 knots), ß = 10 degrees,

Computational Fluid Dynamics and Related Fields, Edited

by N.P. Weatherill, Pineridge Press Limited, Swansea,

5. CONCLUDING REMARKS U.K. Apr 1994.

[8] Sorenson, R.L., "The 3DGRAPE Book: Theory, Users'

The airwake of an LHD ship configuration subject Manual, Examples," NASA Technical Memorandum

to atmospheric wind of 12.87 m/s (25 knots) at wind 102224, Jul 1989.

angles of zero, 10, and 20 degrees is simulated by [9] Volpe, G., "On the Use and Accuracy of Compressible

using a multi-zone, thin-layer Navier-Stokes method. Flow Codes at Low Mach Numbers," AIAA Paper 91-

The resulting flow contains regions of massive flow 1662, Jun 1991.

separation along with free vortices. Major flow [10] Thomas, J.L., Krist, ST., and Anderson, W.K.,

"Navier-Stokes Computations of Vortical Flows over Low-

features including viscous-vortex interactions are

Aspect-Ratio Wings," AIAA Journal, Vol. 28, No. 2,

captured. Some concluding remarks may be drawn 1990. pp. 205-212.

from the results of this study: [II] Baldwin, B.S., and Lomax, H, "Thin-Layer

Approximation and Algebraic Model for Separated

(1) The LHD airwake is asymmetric and wind- Turbulent Flows," AIAA Paper 78-0257, Jan 1978.

angle dependent; the flow over the rear flight deck is [12] Degani, D., and Schiff, L.B., "Computation of

influenced by the wake of the superstructure. Supersonic Viscous Flows Around Pointed Bodies at

Large Incidence," AIAA Paper 83-0034, 1983.

(2) The flow over and behind the superstructure [13] Tai, T.C, "Extension of Baldwin-Lomax Turbulence

Model to Three-Dimensional Flow," AIAA Paper 97-0209,

has characteristics of a backward facing step with

Jan 1997.

massive flow separation involving reversed flow [14] Plate, E.J., Engineering Meteorology, Elsevier

accompanied by circulation. Scientific Publishing Co., Amsterdam, 1982, pp. 527-569.

[15] Gilbert, N.E., "Helicopter/Ship Simulation Model

(3) The flow over the fore deck resembles that of for Seahawk/FFG-7," Paper presented at TTCP HTP-6

a flat plate and is unaffected by the superstructure. Dynamic Interface Workshop, Lakehurst, NJ, Apr 1995.

5-1

PREDICTION OF SHIP AIR WAKES OVER FLIGHT DECKS USING CFD

P.Tattersall

C.M.Albone

M.M.SoIiman

Aero/Structures Department, DERA,

Farnborough, GU14 OLX, UK

C.B.Allen

Department of Aerospace Engineering,

University of Bristol,

Bristol, BS8 1TR, UK

**SUMMARY modelling issues involved which can only come from
**

method development.

A Computational Fluid Dynamics (CFD) method is

presented for calculating the airflow over a ship There is no doubt that, to achieve a completely faithful

superstructure, with emphasis on the flow over aft- representation of the kind of flow field associated with

located helicopter decks on conventional naval ships. airflow over a conventional naval ship, a time-accurate,

The non-aligned grid generation and flow solution viscous flow calculation would be required. Given the

methods are described, including discussion of the current state of turbulence modelling for bluff body

modelling of time-accuracy and rotor downwash flows [1], it seems likely that nothing less than a

effects. Work on the coupling of the CFD results with a Large-Eddy Simulation (LES) would be sufficient.

rotor performance code (CRFM) is also described. Such a level of modelling is not practical at the present

Example solutions are shown to illustrate the current time. However, CFD tools relying on lower fidelity

capabilities of the method. than LES have been used in aircraft design for many

years [2]. This is because the ability to correctly predict

1. INTRODUCTION the trends in the flow, as a design is modified, is often

sufficient to help limit the envelope of possible

Helicopters returning to their parent ship frequently designs. Therefore, at the outset of the project, a

encounter a hostile aerodynamic environment, number of important questions were asked. What are

characterised by complex, unsteady interaction of the important geometric and flow features to be

vortices, in the vicinity of an aft-located flight deck. modelled and the underlying fluid mechanisms? What

This environment, which arises from the airflow over degree of and types of information are likely to be

the ship superstructure, can cause severe handling required? What is likely to be achievable from a CFD

problems during the transition over the flight deck and code intended for routine use within the next few

the landing itself, making helicopter recovery a years? The answers to these questions have a great

difficult and potentially hazardous task. In addition, bearing on the computational approach to be adopted

with certain ships, downdraughts occur, or are induced and none of the answers are obvious.

by the helicopter, at landing spots close to the ship's

hangar which significantly reduce thrust margins In the first place, it was assumed that representation of

during takeoff and landing. the geometry would need to be as accurate as possible,

though not necessarily precise in every detail. This

In order to reduce the hazard, it is necessary to be able would mean that CFD methods relying on structured,

to predict the main flow features that constitute the surface-aligned grids would become steadily more

hostile environment and to minimise them, either difficult to apply as the geometry was made more

through improved superstructure design or special complex. Unstructured grids are a viable alternative but

measures, such as alternative direction of helicopter they carry associated efficiency penalties and there are

approach or post design geometric fixes. One of the particular concerns over their accuracy for viscous

major approaches available for prediction of airflow flows. Thus, it was decided to adopt a non-aligned grid

around ships is the use of Computational Fluid approach. This renders grid generation relatively

Dynamics (CFD), which has a proven track record, in simple for even the most complicated geometry. The

the field of aircraft aerodynamics, for providing tools route taken to a practical method is described in section

of great value for practical design purposes. Whilst 2.

commercial software which could be applied to the

prediction of ship air wakes is available, the option In terms of flow modelling, the assumption was made

chosen at DERA Farnborough has been to develop an that, for conventional naval ships, the principal flow

in-house modelling capability. This offers the features in the vicinity of a flight deck arise from

advantages of a tool adapted specifically for the vortical separations originating at sharp edges. The

problem and the greater understanding of the critical generation and early development of these can be

**Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the
**

Air-Sea Interface", held in Amsterdam, The Netherlands, 5-8 October 1998, and published in RTO MP-15.

5-2

predicted with good accuracy using the (inviscid) Euler successively embedded, such that smaller grid cells

equations. Accurate modelling of the downstream occur near a solid surface. In order to apply solid

development of the vortices, however, requires the surface boundary conditions, a particular type of

inclusion of viscous effects. There are a number of curvilinear mesh is associated with each "feature" of

arguments in favour of neglecting viscous effects, as the geometry. For example, an O-type mesh is

was chosen to do here, at least as a first step. The most associated with an aircraft wing section, whereas a

important is that, since the area of biggest concern for cylindrical polar mesh is associated with the

flow prediction is in the immediate vicinity of the intersection line between a wing and a fuselage. These

flight deck and separation from sharp edges can be curvilinear meshes overlap each other and the

captured by an inviscid method, it is reasonable to Cartesian blocks. They also control the embedding of

investigate how accurately the resulting flow field can the blocks. Communication between the various

be modelled without viscous effects included. In this meshes is governed by a set of hierarchical rules.

context, the information required from a flow

prediction is important. If the complete, accurate For application to ship air wake prediction, the whole

interaction and downstream decay of vortices is curvilinear mesh philosophy was removed from

needed, then an inviscid model is unlikely to be FAME, partly because generation of these grids, for

adequate. If, on the other hand, a general indication of bodies as complex as ships, is very time-consuming.

flow severity, such as maximum downwash in the Control of the embedding of the Cartesian blocks,

plane of the helicopter rotor, is sufficient, an inviscid originally carried out by the curvilinear meshes, now

model may well prove effective. takes place with a very small set of user-specified

parameters. The finest grid which can be used to obtain

It was expected that the influence of the helicopter a flow solution, on an SG Indigo 2 workstation with 32

rotor on the air wake would be significant and the MW of RAM, takes less than ten minutes to generate.

capability to model this by the actuator disc approach This includes all the flagging of the various types of

was included. Methods for improving the rotor grid point (internal to body, normal field, actuator disc,

downwash representation have been investigated interpolation required, etc.), for any geometry.

recently and are described later. It was intended that a

CFD tool should be produced which could be used for A computationally convenient device is currently used

routine design purposes within the next few years. This to describe the ship geometry. The ship is defined by a

means that a time-accurate, viscous method is unlikely collection of hexahedral elements, because each

to be a viable candidate. Of the two aspects, it was hexahedron can be divided into six tetrahedra and there

expected that it would be more important to model the exists a very simple test to find whether a point lies

variation with time. The evidence to date suggests that within a tetrahedron. This is useful because, during the

this is a valid assumption and this enhancement is grid generation process, special points, which lie inside

currently being developed. Initial results are presented the body but have neighbours outside, must be

later' in this paper. identified, since these are required for the

implementation of solid-surface boundary conditions.

2. MESH GENERATION More general numerically defined geometry can be

handled by routines existing within the FAME suite.

In light of the considerations described in the However, this has not yet proved to be necessary. A

introduction, it was decided that a "non-aligned grid" section through a typical grid, for an AOR (Auxiliary

approach should be adopted. This makes the generation Oiler Refueller), plus a simplified helicopter body, is

of a computational grid very easy, but necessitates shown in fig.l, highlighting the embedding of the

some special procedure within the flow solution code blocks near the ship and the non-aligned nature of the

in order to apply the solid surface boundary condition grid.

accurately. It should be emphasised, at this point, that

the approach taken was not to develop a state-of-the- 3. FLOW SOLUTION

art, non-aligned grid capability [3]. This would have

entailed starting "from scratch," requiring more Two alternative numerical techniques have been used

development time than was available. Instead, an to obtain the flow solutions presented here. In both

existing CFD method [4], known as FAME (Feature cases, a compressible flow solution code already

Associated Mesh Embedding) was used as the starting existing within the FAME system was modified to

point. solve the incompressible Euler equations, using the

technique of artificial compressibility [5].

FAME was developed, originally, for civil and military

aircraft applications. It solves the problem of 3.1 Governing equations

generating surface-aligned grids for highly complex

geometry by using a series of overlapping and The approximation of incompressible flow is a valid

embedded, Cartesian and curvilinear meshes. The one for the prediction of ship air wakes. The equations

whole region of interest (away from solid surfaces) is can be solved by conventional CFD methods by adding

covered by a set of regular, Cartesian blocks, which do a fictitious time derivative of pressure term to the

not, in general, align with the geometry. They are continuity equation. This terms vanishes in the steady

5-3

**state, but its presence alters the mathematical nature of
**

the equation set, allowing time-marching techniques to if M>0, u=u, «=0,

be used. The equations to be solved are: if H<0, w+=0, u=u.

**Ut + E* +£y +_GZ = 0, (1) Second-order spatial accuracy is achieved by the use of
**

defect-correction. In short, this entails periodically

where taking central differences of the solution and adding

the result, as a source term, to the equations which are

U= (p, u, v, w) , then operated on by first-order upwinding again.

Although this method is relatively rarely used, the

E= (ßu, u +p, uv, uw)T, algorithm was attractive during the early development

F = (ßv, vu, v +p, vw) , of the FAME system because it requires only a seven

G = (ßw, wu, wv, w +p) . point stencil in 3D, reducing the memory requirements

significantly in comparison with more conventional

Here, x, y and z are the Cartesian co-ordinates, u, v and CFD techniques. This algorithm is used for the steady-

w are the associated velocity components andp is the state solutions shown later, including those with

pressure. These are the 3D incompressible Euler actuator disc representations of the helicopter rotor

equations, augmented by the pressure derivative in the effects.

mass continuity equation. The term ß is the artificial

compressibility parameter. Formally, it has the 3.3 Finite-volume method

dimensions of a velocity and its optimum value is

dependent on the non-dimensionalisation used, but, in Recently, a finite-volume method has been developed,

practice, it is generally set to one. within the FAME system, originally for inviscid flow

[7]. It is currently less efficient for steady flows than

3.2 Finite-difference method the finite-difference method, since it does not include

multigrid convergence acceleration. An incompressible

An incompressible version of a finite-difference version has been developed, for ship air wake

method [6] was developed first, since it was most predictions, which solves equations (1), using the

mature at the start of this work on ship air wake normal, explicit finite-volume approach. The equations

prediction. It cannot, however, be easily extended to are written in the conservative form:

time-varying or viscous flows. The solution progresses

towards the steady state by simple Euler time-stepping, d r r FV

— \UdV+ [ F .ndS = 0,

augmented by multigrid convergence acceleration. dt #- w- ~

Although the FAME mesh environment is less than

ideal for multigrid, the technique still improves where, again, U is as in equation (1) and Ffv is the

convergence by a factor of two or three. The total flux vector. V represents a computational cell

underlying spatial discretisation is based on first-order volume, dV\5 the boundary of V, dS is an element of

upwinding. For the incompressible Euler equations dKand n is the outward normal to dS. The spatial

with artificial compressibility, the decomposition used discretisation uses the TVD approach. Although this

is:

complex method is not necessary for ship air wake

flows, it has been retained to avoid diverging too much

P from the original method [7]. Temporal discretisation

2ax 0 0

uses an explicit, multi-stage Runge-Kutta scheme and

2ax 2a, 0 0 local time-stepping is adopted as an acceleration

U,+ ur technique.

-u V

2 ■uu / 0

"I

w (u+axf 3.4 Time-accurate method

■uu 0 u+

V", , 2 al I 2

More recently the finite-volume method has been

2ax 2ax 0 extended to time-varying flows [8]. If a fully explicit

time-stepping scheme is used for time-varying flows,

2a, 2ax 0 the whole domain must be integrated forward by the

+ (»-°J : Ur same time-step, which must be equal to the smallest

all 2

-—u 2 UU u~ time-step over the whole domain for stability reasons.

_W_ «-"x \«-a*Y ■uu 0 This results in excessive run times. A particularly

a2A 2 efficient alternative is to make the code implicit, using

the dual-time approach [9], which has been shown to

+ (similar) v and w terms = 0. be effective for upwind schemes [10,11]. The implicit

form of the differential equation for each

Here, U is as in equation 1, ax = V(u2+1), the arrows computational cell is considered,

indicate forward or backward differencing and

5-4

**across which a "jump" in one or more flow variables is
**

+l rr"+l>

prescribed. In accordance with simple momentum

d(V"«+1U"+') theory (see e.g. [12]), the rotor is modelled by a

+ R(U"+1)=O, constant pressure jump across the rotor disc. For

dt

convenience, the position of the disc is assumed to be

where V\s the time-dependent cell volume and R is the equivalent to the nearest grid line (this usually entails

upwinded flux integral. The superscript refers to the an error of a few centimetres on the ship scale) and the

time level, n+1. To date, in this application, only pressure jump is simply added in to the flux evaluation

stationary geometries have been attempted, so that the at that point, in a computationally convenient way.

cell volumes need not be recomputed at each time step.

The implicit temporal derivative is then approximated Whilst this simple actuator disc technique gives the

by a second-order backward difference, giving gross effect of the rotor downwash on the overall flow

field, it can clearly be improved upon quite easily by,

3

2M

vn+lu"+i "U"]+ for example, imposing a more representative pressure

jump across the disc. A real rotor generates lift, which

yn-\jjn-\ +1

1

2 At

+ R (u" )=o. varies in both radial and azimuthal directions. Hence, a

similarly varied pressure jump should represent the

influence of the rotor more accurately. Such a pressure

lf+' must now be iterated on to satisfy the equation. jump distribution can be calculated and supplied to the

To achieve this, a new residual R*(U) is defined as CFD calculations if the full interaction between the

ship and the rotor are well represented. The interaction

is influenced by the ship geometry, the operational

conditions and the rotor performance.

^[V^U'-1]+R(U)

The rotor analysis code CRFM (Combined Rotor-

and then a new differential equation can be written in Fuselage Model) is used in this study to perform the

terms of a fictitious time x, (called pseudo-time) rotor side of the calculations. The code was developed

jointly by DERA Farnborough and GKN-Westland

Helicopters to predict the aeroelastic behaviour of

v»^^L + R*(u)=o. helicopters [13]. The code is used here, initially, to

dr supply the distribution of pressure jumps as an input to

the CFD code. However, it is essential that the flow

This is simply time-marched to convergence in the field, surrounding the rotor, which is supplied to the

fictitious time x, for each real time step. As R* -» 0, CRFM code is accurately represented before the

U->U_"+I. Hence, although U"+l is an unsteady pressure jumps are calculated. Since this flow

solution, it satisfies a steady equation. Steady information is supplied by the CFD code, it follows

acceleration techniques can thus be implemented in that accurate modelling in both codes can only be

pseudo-time. As in the steady case, only local time- achieved through a process of iteration between the

stepping is used at present. There is now no limit to the two.

size of the real time step that can be taken, leading to a

large reduction in CPU times. The time step is now The process of iteration starts from the CRFM code,

limited by accuracy, rather than stability, which assumes that the helicopter is in free flight away

considerations. from the influence of the ship. The code calculates

spatial and azimuthal values of the pressure jumps

4. ROTOR INFLUENCE from the blade load calculations, then feeds this as an

input to the CFD code. An initial run of the CFD code,

The airflow over a flight deck during helicopter with these pressure jumps, produces a flow field at the

operations is undoubtedly influenced by the downwash rotor disc that reflects the influence of ship structure as

induced by the helicopter rotor and vice versa. It seems seen by the rotor. The CRFM code is then run again

obvious that some representation of the rotor effects but with the flow field from the CFD code imposed on

should be included in any CFD calculation of such a its own flow calculations. A new set of pressure jumps

flow field. Various levels of approximation can be is then fed to the CFD code, which returns a new set of

employed in order to model rotor effects. The ideal is flow velocities to CRFM ... and the iteration

to perform a fully integrated calculation in which the continues. This process is outlined in the flow diagram

geometry and rotating motion of the individual blades below.

is included, in a time-accurate sense, within a time-

varying ship air wake. However, even with state-of- The iterative process eventually converges when the

the-art computing facilities, such a calculation will be rotor control angles, as calculated by CRFM, do not

unfeasible for a number of years. On the other hand, change considerably between the last two iterations.

there are a number of approximations that can be made The control angles are used as a criterion for

in order to include rotor effects within CFD. Perhaps convergence detection because these are sensitive to

the simplest is to represent the disc as a discontinuity, changes in the flow field. The first test case with this

5-5

**This ship/helicopter combination is used for most of
**

Run CRFM for the results given here to illustrate the calculation of a

rotor free flight realistic situation, although no experimental data are

available for comparison in this case. A significant

amount of validation against both wind tunnel and full-

Calculate comrol angles and scale experimental data has taken place at DERA. A

pressure jump distribution particularly useful exercise, undertaken under the

auspices of TTCP AER-TP-6 KTA-2, involved the

comparison of results with those from other CFD codes

Run CFD tor Combined JTJItor

Run CFD and data from a wind tunnel experiment on a simple

"AOR/Rotor" model 'Rotor" model shape exhibiting the main features of a typical frigate.

This exercise is described elsewhere in this

Symposium [14]. The result of this and other exercises

Calculate velocity Calculate velocity was that the Euler code does correctly predict the main

distribution due to distribution due to flow features arising from sharp edges.

AOR and rotor (VAR) rotor only (VR)

The effect on the overall flow field of including rotor

influence is shown in fig.3. In this case, an actuator

Calculate velocity distribution disc is used with a constant jump in pressure across the

due to AOR (VA=VAR-VR) disc equivalent to that needed to keep a large helicopter

in a steady hover. As expected, the impact on the flow

field is dramatic, as can be seen by comparison with

Run CRFM with fig.2a. The importance of including some

imposed values of (VA) representation of the helicopter body (albeit a relatively

crude one in this case) is shown by the fact that a small

fountain effect is seen below the body, which does not

Calculate control angles arise if the actuator disc is included in the calculation

and pressure jump without a body. Another interesting feature is the way

distribution in which the downwash from the front part of the rotor

strikes the deck and is then deflected upwards by the

hangar face before being drawn back into the rotor.

Control This feedback interaction has been suggested as a

NO angles cause of problems observed during some trials, in

converged? which the proximity of the helicopter to a vertical

surface gives rise to handling difficulties entirely

YES independent of the natural airflow over the ship

superstructure.

**END In fig.4, a result is shown for a similar calculation.
**

Here, a spatially varying pressure jump is imposed

across the actuator disc. The distribution used here is

approach is described in the next section. It was found that which is calculated by CRFM for the helicopter in

that three or four iterations between the two codes were free flight, where the helicopter is assumed to be a

enough to achieve stable convergence. Merlin.

**5. RESULTS As an illustration of the interaction of the CFD code
**

with the rotor performance code CRFM, two CFD

An example of a ship air wake, calculated using the solutions are shown in fig.5a. These are the first and

steady, finite-difference, Euler code, is shown in fig.2. fourth calculations in a sequence as outlined in section

The flow around the AOR plus simple helicopter is 4. In the first, the imposed pressure distribution is that

shown, in fig.2a, by the use of velocity vectors in the which is predicted by CRFM for a Merlin in free air

vertical plane through the helicopter centre line. The with a 20 knot wind coming from 20 degrees to port.

rotor itself is assumed to be stationary in this case. The (This is a closer view of fig.4). In the fourth and final

flow appears quite mild since a scaling of the vectors calculation, the distribution imposed is that resulting

has been chosen to facilitate comparison with later from successive applications of CRFM with the

cases. However, the air over the deck is significantly velocity components from the CFD solution. Again,

disturbed by the presence of the superstructure. This is velocity vectors are shown in a plane through the

more obvious in fig.2b, where the generation of helicopter centreline. Under close examination, some

vortices is shown more clearly by the use of "stream differences can be seen between the two, although they

ribbons," which are linked pairs of streamlines. are not dramatic. In the absence of experimental data,

there is no reason to suppose that this is a particularly

severe case. Also, a relatively coarse CFD grid was

5-6

used to reduce turnaround times, since the main point deck will be separations from sharp edges, which can

of this exercise was to check that the CFD/CRFM be predicted by numerical solution of the inviscid

interaction process was feasible. Since much of the equations. It is essential to include some representation

detailed change in disc loading is concentrated in of helicopter rotor effects on the airflow. This can be

relatively small areas, particularly near the outer edge, done, as a first step, by a constant actuator disc

it may be that a finer CFD grid is necessary to capture approach. Improved modelling can be achieved by use

the resulting changes in the flow field satisfactorily. of a variable actuator disc. One way to decide on the

Fig.5b shows how the sequence of CFD/CRFM variation required across the disc is to use initial CFD

calculations produces a predicted change in collective results as boundary conditions within a helicopter

pitch and cyclic control settings and that the iterative performance code. Such a code can produce a new

sequence achieves convergence. It is hoped that the variation across the disc, which is then fed into an

veracity of the predicted changes, for this or a similar updated CFD calculation, before returning to the

case, can be checked against experimental data in the performance code, etc. This iterative process has two

near future. potentially important effects. The first is to produce a

more accurate CFD solution than that obtained with a

The air wake over a real ship superstructure is highly constant actuator disc model. The second is to provide

unsteady. Indeed, it is not obvious that a steady CFD predictions of the effect of the air wake on the

solution would be of any use whatsoever in predicting helicopter, hence aiding the understanding of the

this situation. The general conclusion from the testing handling problems suffered by the pilot during

carried out to date is that steady solutions can be of shipborne operations.

use, although more evidence is still needed to back this

up. However, it is undoubtedly the case that the effects The highly unsteady nature of a real ship air wake will

of variation of the flow field with time must be undoubtedly be an important consideration.

investigated. The route chosen towards time-accurate Modification of the existing code to include time-

CFD solutions was described earlier. An initial result varying effects has taken place, using an efficient

has been obtained for the geometry tested as part of the implicit scheme. Initial results are encouraging,

TTCP collaboration described in ref.14. A relatively although much more testing and potential development

coarse spatial grid was used to facilitate a more rapid is necessary.

turnaround of the solution. This particular test case

involves a port/starboard symmetric geometry and a Even with steady, inviscid flow solutions, much

yaw angle of zero degrees (wind from dead ahead). It is progress has been made on the prediction of realistic

to be expected that the real flow will vary with time, ship air wakes. Indications are that a useful predictive

but that a long time-averaged view of the flow pattern tool has been developed during this project.

will look symmetrical. Fig.6 shows two views of the

computed flow field, one corresponding to a particular

instant in time and one to an average of the solution

over a large number of real time steps. It is clear that REFERENCES

the instantaneous solution is highly asymmetric,

whereas the averaged view is almost perfectly 1. Breuer M, Lakehal D and Rodi W. Flow around a

symmetrical. This average solution agrees reasonably surface-mounted cuboidal obstacle: comparison of

well with that observed in the experiment [14], LES and RANS - results. In Notes on Numerical

although it was found, in the context of steady flow Fluid Mechanics, vol.53, Vieweg.

solutions, that a finer grid is necessary to really obtain 2. Albone C M, Catherall D C, Hall, M G, Joyce

a good representation of this situation. Thus, for a very Gaynor. An improved numerical methodfor

preliminary investigation, this result can be regarded as solving the transonic small-perturbation equation

highly encouraging. for the flow past a lifting aerofoil. RAE TR74056.

3. Powell K G. Solution of the Euler and

6. CONCLUSIONS magnetohydrodynamic equations on solution-

adaptive Cartesian grids. VKI Lecture series

A method has been presented for modelling the air 1996-06.

wake over a conventional naval ship, with particular 4. Albone C M. An approach to geometric andflow

attention paid to the region of an aft-located helicopter complexity usingfeature-associated mesh

landing deck. Grid generation is reduced to a very embedding (FAME): strategy andfirst results. In

simple problem, by use of the "non-aligned mesh" Numerical methods for fluid dynamics III, Oxford

technique, in which no attempt is made to align any University Press, 1988.

grid lines with solid geometrical surfaces. Accuracy is 5. Chorin A J. A numerical method ofsolving

enhanced in the region of solid surfaces by the use of incompressible viscous flow problems. Jnl. Comp.

grid embedding. Phys., Vol.2, pp. 12-26, 1967.

6. Albone CM. A second-order accurate scheme for

The airflow is modelled by solution of the Euler the Euler equations by deferred correction of a

equations. This relies on the assumption that the first-order upwind algorithm. RAE TR88061.

dominant flow features over an aft-located helicopter

Athens. Time dependent calculations using multigrid. 1998. ECCOMAS CFD Conference. Pitman. 9. Allen C B. Vol. Proceedings 5th ICFD Rotor-Fuselage Model. Vol 211. 1959. 1997. research objective 1. AIAA paper 91-1596. Funk J D. 1992..101. Wilkinson C H. 11. 10. 14. Payne P R. overlapping and embedded meshes. r-W'-i-ur^ Figure 1.. The development of the Coupled algorithm for FAME. Hansford R E. Pub. with applications to unsteady flows past ACKNOWLEDGEMENTS airfoils and wings. A collaborative venture. 1997. assignment 2. London. 1995.. IMechE Journal of Aero Eng. / i '>-. Aero. Jnl. Conference. A TVDflow 13. 15f.. Section through computational grid for AOR + helicopter . Allen C B. Onslow S H and Tattersall P. 5-7 7. ILLUSTRATIONS ••-*.. pp. Amsterdam. AGARD Symposium. Allen C B. Grid adaptation for unsteady flow ©British Crown Copyright 1998/DERA computations.pp7-16. Gilbert N E. The reduction of numerical entropy This research is funded by the UK MOD under ARP package generated by unsteady shock waves. aerodynamics.237-250. 8. Zan S J. Published with the permission of the Controller ofHer 12. Helicopter dynamics and Britannic Majesty's Stationery Office. Jameson A. No.1001. 48* AHS Forum. 9. Parallel Modelling and simulation ofship air wakes for implementation of an unsteady Euler solver using helicopter operations. Onslow S H and Peshkin D A. 1998. Oxford. 4 Paper No.

Air wake over AOR + helicopter with no rotor effects-velocity vectors Figure 2b.5-8 Figure 2a. Air wake over AOR + helicopter with no rotor effects-stream ribbons .

Air wake over AOR + helicopter with variable actuator disc . Air wake over AOR + helicopter with constant actuator disc Figure 4. 5-9 —9—9—f~P—F—$-~P-- Figure 3.

Results from sequence of iterations between CFD and helicopter performance code -B— Collective Pitch -X— Longitudinal cyclic -0— Lateral cyclic ' 5 Iteration No Figure 5b.5-10 First CFD calculation Fourth CFD calculation Figure 5a. Results from sequence of iterations between CFD and helicopter performance code .

Unsteady calculation . 5-11 -> > > > > > >_ -> > > 5> > >- / S S s» ->—>—>- t r / s -> s^ \ \ V \ \ \ \ t t / ^ _~ _ ^* \ \ \ \ \ \ \ \ \ i .instantaneous solution ""^ 't=* ~p*-- S ' I \ N. ^M -> 5> J \ \ / / 1 — •*.time-averaged solution . Unsteady calculation . / N \ \ \ 1 / y I v \ / s s* Figure 6b. s* ^ V \ \ \ \ \ \ \ \ \ I / s V \ \ \ \ \ t / / ^ V N \ \ \\\\\ \ \ 1 r / / *"* v N \ \ \ \ \ \ 1 ! t / / / ^ V \ \\\ f / ///// / Figure 6a.

its influence on the helicopter flight dynamics Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface".Wind tunnel tests. unsteady flow provides high mean speed gradients to which aerodynamic turbulence (fluctuations) is added. has developed an frigate moves. and published in RTO MP-15.Effects on the Helicopter Flight Dynamics A. Three global performance. dimensional measurements of the mean wind speed and velocity fluctuations were performed around the Under SPAe funding.Demonstration of the frigate wake effects (mean This paper describes the test results obtained and the velocities + fluctuations) on helicopter loads and work performed in modelling the La Fayette wake and flight dynamics.Air France ** ONERA-IMFL . 1. the helicopter has to face with the changing deck of a frigate La Fayette. . under SPAe funding (Service des pilots as highly risked operations. This on helicopter flight dynamics. 6-1 Aerodynamic Perturbations Encountered by a Helicopter Landing on a Ship .Lille cedex France 1.Definition and development of a mean air-wake Overall Simulation Tool (HOST). flights above the deck with this model demonstrated 4. A. INTRODUCTION This paper describes a French activity on helicopter Flying a helicopter above or around a frigate deck. The turbulence model is realised using power spectral densities of velocity fluctuations The main following topics are presented in this paper : measurements. ABSTRACT 2. 2. 6. Taghizad* . Ch.Departement Commande des Systemes et Dynamique du Vol BA 701 . held in Amsterdam. Wind tunnel tests were carried out in ONERA-IMFL These conditions have important effects on helicopter with a 1/50* scaled model of the La Fayette. landing on it or taking off from.Definition and development of a turbulence model. This model was then used in order to mean wake model [1] to which a turbulence model adds study the effects on flight mechanics. ship landing operations simulation improvement.Implementation in HOST (Helicopter Overall flight dynamics. ONERA has realised wind tunnel landing deck area. Verbeke** .Boulevard Paul Painleve 59045. The model was connected to the Eurocopter Helicopter 3. tests on a 1/50* scaled La Fayette frigate model. are considered by ONERA. velocity fluctuations. important effects of the ship air-wake on the helicopter 5. Indeed. and has developed an aerodynamic wake model of the ship The La Fayette aerodynamic wake model includes a landing area. and has tested its effects aerodynamic wake of the ship superstructure. but also in the neighbourhood of the aerodynamic disturbance model around the landing deck. Simulation Tool). Desopper* * ONERA . The Netherlands.Ecole de l'Air 13661 Salon . Simulations of model. and behaviour.Departement Aerodynamique Appliquee 5. not only the Programmes Aeronautiques). 5-8 October 1998.Analysis of the test results obtained.

z) 3. Because of the model and the test section sizes. The downward flow deviation can easily be seen In addition. z : space co-ordinates of the local point The different effects described above can be seen again t : time variable in the 3D flow visualisation of figure 8. Due to the aerial elements (missiles.2 Test configurations paper [1]. Thus. Therefore. y.. w) after a For the frequency. configured with its Crotale Missiles Du => pu: longitudinal turbulence PSD (figure 1). free stream velocity. Arrows represent the mean velocity projections on visualisation The fluctuations (u. while the longitudinal velocity decreases to a U mean : the longitudinal ° velocityJ mean value.). The maximum vertical velocity is reached on the centre U(x. 2 redundant of scale 1 frigate (wind tunnel frequency / 50).1 Test equipment description free stream Wind tunnel measurements were carried out in the u(x. the relative wind side-slip angles were limited to Dv=> pv: lateral turbulence PSD ±15° around the frigate longitudinal axis.2 Ship air-wake mean velocity measurements 2. Uw) were provided. funnel. The first 50 meters of the marine atmospheric boundary layer was also simulated. For example. the mean value and the fluctuations around air-wake decreases with the height above the deck and this mean value.y. minimum in the same area.t) U(x.4 m above the performed. v and w) can be considered as planes. hot film anemometer. The normalised velocity fluctuations are characterised by their Power Spectral Densities (PSD): Measurements were performed on a 1/50* La Fayette frigate model. due to the hangar wall cliff-effect (h=6.6% of the infinite upstream wind.60m). Vm. Two velocity components were simultaneously measured (u.6-2 3. denoting a general down wash on the deck The three velocity components are decomposed into and behind it..z. distribution This topic has already been discussed in a previous 3. At last. especially below the hangar height. the results will be presented function 90° rotation of the sensor. 15° and 180°).The mean vertical velocity is all the time 4. DATA ANALYSIS flow (Wm<0). whereas coloured areas show turbulence levels. t) line. turbulent terms.y.. a clear asymmetry of the turbulence distribution is obtained. y. measured velocities are normalised by the and a high turbulence level is noticeable on the deck.y. and then (u. Dw => pw: vertical turbulence PSD 3D unsteady velocities were measured using crossed This approach is similar to the one used in [2] and [3]. radar. the longitudinal velocity when moving to the rear. t) = Umem (x. U : fluctuations around Umean x. Wm) along the vertical axis. Figures 2 and 3 show the test volume and the selected test points positions inside. This last effect is due to the is decomposed into: second flow deviation on the back side of the deck.4 m in Vfree stream ~ V.t) ONERA-IMFL low speed wind tunnel (SH). same parameters evolution respectively along the Some tests at 25 kts and 0° of side-slip were also longitudinal symmetry axis (height 4. y.z) tunnel has a closed circuit and a test section of 2.1 Velocity decomposition negative. on the deck centre (point A).4 m). Figures 6 and 7 show the wind and three side-slip conditions (0°. . z) + u(x.y. The longitudinal component of the mean two parts . y.free stream diameter. Figure 4 gives a summary of these test deck) and the lateral axis which crosses the deck at configurations. These 3 figures show a clear downward deviation of the 4. 8 horizontal Figure 5 gives the evolution of the mean velocity planes above the deck were considered. components (Urn.z. Speed measurements error is estimated as 4. deck. we will confine our discussion to The 3 unsteady ship air-wake velocity components a brief presentation of some results and the synthesis of were measured in a volume surrounding the landing mean velocities measurements. WIND TUNNEL TESTS umean(x>y>z) Um(x. longitudinal velocity measurements (Uv. v). z. z. for a 50 kt wind speed Detailed data measurements were realised with a 50 kt without side-slip (ß = 0°). point A (height 4. This wind Du = -Um(x. Light colours corresponds to the highest turbulence level. the air-wake is deviated from both sides toward the centre.

Even an or extrapolated using the results at 25 and 50 kts. it was decided to extend. Above the PSDs and it is not possible to extrapolate these results deck. 15°. 1. Dv. kts. were selected in order to compare the results for the 2 wind conditions (25 and Figure 12 shows a 3D presentation of the vertical 50 kts). PSDs start getting flat. 15). fluctuations on the deck centre (point A). C and D. This on point A. spectral densities show an energy A clear wind speed effect on PSD amplitudes is visible concentration approximately around 0. A small asymmetry . This investigation shows that mean velocities [1]. the mean wind normalised angle is a quite important factor acting on turbulence components have similar evolution. (rearward) side-slip conditions. a width right side) at the same height (fig. for a 50 kt wind speed and 0° side-slip.3 Aerodynamic wake fluctuations c. considerably compared to the case of 0° side-slip. • A detailed database is also available at 50 kts with 2. measurements at 50 kts with respectively 15° and 180° This asymmetry is due to higher deck's aerial elements. along the vertical axis on point A. which is out of this main wake at 15° side- 15° of side-slip. The mean flow changes slip. of the velocity components fluctuations (Duv. As it was already concluded for the side-slip effect on without side-slip. the La Fayette A 50 kt rear wind database is also available. Side-slip ß = 0° a. These points are respectively located on the velocity (w) Power Spectral Density (PSD) evolution deck centre. Dw) on different test planes above these points. by interpolation. is to be noticed. On point C and D this effect seems to get reduced. 2 cases. Figure 13 presents the 3 velocity components spectral The first row of plots from the bottom corresponds to densities on different test planes above the deck centre. the lowest test plane and the last one to the highest test For planes below the hangar height (first 3 rows from plane. which remains in the frigate superstructure main wake for both cases. interpolation between 0 and 15° is critical. at the rear and behind the deck (figure 19). For the amplitudes on this point are higher at 25 kts than at 50 planes above the hangar height. Indeed. 14) and along the lateral axis crossing the centre point A (from one deck width left side to one deck Concerning the frequency of PSDs maximum (fmax). Figures 14 and 15 illustrate the vertical velocity (w) Additional investigations have shown that the wind PSD evolution respectively along the median axis (on speed effect on PSD amplitudes is much reduced when the deck and one deck length behind) at a height of 2. measurements have been done at (ß = 0°) and 3 side-slip conditions at 50 kts (ß = 0°. the bottom).Wind velocity = 25 kts. 50 knots with a 15 degree side-slip. Duw.point B.5 Hz. normalised power spectral densities maximum of energy decreases with the height. Side-slip ß = 15° Comparisons of mean velocities evolution for 2 wind In order to study the side-slip effect on air-wake speed conditions (50 kts and 25 kts) without side-slip velocity fluctuations. However. However. in order to extend the database to vertical axis. of the 3 velocity components PSDs along the "tolerated". it has been decided that interpolation Figures 17 and 18 are comparisons. these results show that the side-slip except above the deck.Wind velocity = 50 kts.4 the point considered is far from the deck. respectively on A between 15° and 0° of side-slip at 50 kts will be and B. b. 6-3 Figures 9 and 10 are 3D flow visualisations of around the median axis (longitudinal). in order to illustrate a comparison between the 2 tested side-slip • A detailed database is available at 50 kts of frontal configurations (figure 16): wind (ß=0°). at 2. These relationship between fmax and the wind speed figures show again that turbulence is much higher (Strouhal number similarity) can be expected: above the deck than beside it. 4. A. Three points. Large differences can be seen between the intermediate side-slips. m (fig.Wind velocity = 50 kts.point A. The effect of the lateral hangar wall can be seen for a side-slip angle of 15°. We can notice high turbulence rates concentrated below the hangar wall Figures 20 to 22 illustrate the power spectral densities height. A partial database has been generated at 25 kts.4 m above it. Side-slip ß = 0° Limited measurements were performed in 25 kt frontal Figure 11 gives an example of normalised velocity wind conditions (ß=0°). these components will have to be interpolated for side-slip configurations higher than 15°. aerodynamic turbulence model to intermediate side- slips (0 to 15°) at 50 knots of wind speed. 180°) [1] provide the following conclusions on the mean air-wake model usage domain : Two points have been selected.

Above the deck. PSDs will have to be distance to the considered point. a. Side-slip ß = 180° Measurements were also performed for a rear wind The approach consists in locating the helicopter centre condition at 50 kts. VF. the ratio of fmax at 50 kts on fmax at 25 (turbulence). further investigations were carried out to 5. in order to extend the b. air-wake components can also be v 50kts calculated at 50 kts.Calculation of measured velocity fluctuations PSD. Figure 24 illustrates this 15° of side-slip. the model will give an air-wake by C being a caracteristic length ' 25to interpolation. except above the deck. using a linear combination of measured mean velocities on the e. Fluctuations are then processed using a signal generation method ensuring the similarity between • A 50 kt rear wind database is also available. measurements: • A detailed database is available at 50 kts of frontal In order to avoid any velocity discontinuity when going wind (ß=0°). The approach consists in locating the helicopter centre without side-slip.Identification of a mathematical model (S) Measurements synthesis done above. to describe the model structure and to define its limits of validity: . At any assumption of Strouhal number similarity will be point (H) of this area the 3 mean air-wake components considered for the turbulence model realisation.3. LA FAYETTE FRIGATE AERODYNAMIC The method consists in: WAKE MODEL a. Fluctuations PSDs change method. 5. St.4 and 2. PSDs of measurements and of the generated signal.1 Model structure b. kts was found between 1. d. ' 25to CÜ 50to WF). In our model and a model of velocity fluctuations investigation.2 Model realisation consider this question. The mean air-wake on this point is defined 3D visualisation of the test measurements.Wind velocity = 50 kts. the a grid according to the test points definition.6-4 Cß 25kts In every frontal wind velocity condition (ß=0°) the St 25/ta model can provide air-wake components (UF. interpolated or extrapolated using the results at 25 and 50 kts. non comparison can be of gravity in the test area elementary parallelepiped done with forward wind configurations. makes it possible fitting the experimental PSD. components PSDs. using the mean air-wake measurements of neighbouring points. if Shskts ~ Shokts =* ** 25kts Q 50kls For rear wind. including mean velocity and turbulence. are interpolated. Velocity fluctuations are generated using the 3 velocity slip at 50 kts will be "tolerated". interpolation between 15° and 0° of side.Velocity fluctuations model (turbulence) database. in respect with their similarity. Figure 10 is a (figure 23). It has been decided that. considerably compared to the case of 0° side-slip. in respect with their The following conclusions come out from the distance to the considered point.50kls At 50 kts of wind.Mean air-wake model Due to the fact that the determination of the frequency The test area above and around the ship deck is actually of the maximum amplitude is not always very easy. the respect of Strouhal number elementary parallelepiped tops. for any side-slip between 0° and K 50to 15°. in/out of the test area. the normalised velocity Fluctuations PSDs on this point are defined using a components fluctuations have similar evolutions linear combination of measured PSDs on the with however. via its components in the frigate axes. 5. Obviously. Thereby. a transition region has been defined.Synthesis on velocity fluctuations measurements elementary parallelepiped tops. This investigation shows that of gravity in the test area elementary parallelepiped. • A partial database has been generated at 25 kts. where velocities are interpolated between the • A detailed database is also available at 50 kts with test area and the free stream. Comparisons of PSDs at 50 kts The La Fayette air-wake model includes a mean wake and 25 kts were done on several test points.

turbulence conditions. -11° in roll assuming a constant speed.t)=2ZyjS(x. Figures 30 and 31 illustrate this case. . demonstrated the opposite effect. and attitudes have changed by -7° in pitch.180°). The ship stops after 8 seconds when helicopter altitude is zero trajectory is then calculated in the air-wake model.Velocity fluctuations computation from the parameters (helicopter attitudes. helicopter controls keep their trim values (figure 27). longitudinal symmetry axis (y-0). DVF-FGW. In these cases. Blades flapping and the local turbulence terms (DUF-FGW. Figure 25 is an example of vertical turbulence generation on the deck centre (height = 2. VF-FGW. these turbulence effects occur around a general tendency given by the mean velocities and will Typical results are presented on figures 27 to 29 for a certainly have an effect on piloted simulations.Temporal frequency for f sample of the simulation and a quick fall in altitude with a large lateral deviation (20 m to the left).5 t 7. Other simulations have • locating the helicopter in the test area grid. 15°. WF-FGW). Figure 30 shows that the longitudinal component of the air-wake (UMF- FGW) decreases during the approach because the 6. 9 m above the deck and 23 m behind its rear extremity (X=25 m in the The first phase of this activity started with wind tunnel frigate axes). The comparison of these 2 trajectories shows that for the first simulation (mean air-wake + turbulence). In order to demonstrate the frigate aerodynamic wake effects on helicopter flight dynamics. 'non favourable' initial turbulence generates a lift loss and a quicker altitude fall. 6-5 c. the The air-wake model fulfils the following tasks: altitude loss has been delayed because of an initial lift raise.y. The connection was done The combined effects of these two air speed assuming the helicopter as a mass point.y. using the following method components. The helicopter. components result in a general decrease of the main rotor lift (FZG-RP). Figure 26 illustrates this and -25° in heading. CONCLUSION Dauphin. The air-wake model uses some general data coming The helicopter pitches down and accelerates because of from HOST (wind velocity and direction). However.z. 1=1 N\ Number of samples in S These plots show a short deceleration at the beginning ff. with a higher altitude • calculating the 3 velocity components at the loss compared to the simulation with the only mean air- helicopter CG. fuselage and horizontal Example of u generation: stabiliser). a 3. out with HOST.. wake.co^2jfit+(fi(x. The data analysis showed FGW.y.fi)-^. EFFECTS ON HELICOPTER LOADS AND helicopter comes closer to the deck. descending flight on the deck at a speed of 10 kts and a slope angle of -1°. the 50 kts for 3 wind side-slip configurations (0°. this last being caused by 'favourable' initial • generating the ship trajectory. Figure 29 [4]: shows the aerodynamic loads on different helicopter elements (main rotor.z)] presented on this figure. the helicopter flight . with those coming from the same simulation but in mean velocities effect only. Simulation Tool) [5]. DWF-FGW) are also u(x. The frigate is still and faces with a 50 kt forward wind (ß=0°).) and trajectory coordinates. on a l/50th model of the French frigate La Fayette. A detailed database was provided at The simulation being realised in open loop. The vertical air- FLIGHT DYNAMICS wake (WMF-FGW) is all the time negative (down The frigate air-wake model has been connected to the wash) and changes not only with height but also with Eurocopter simulation code HOST (Helicopter Overall helicopter X and Y positions above the deck. ground speed identified PSD. and consequently a loss of altitude. and specific the pitch moment decrease (MYG-RP). is trimmed with the local mean air-wake This paper presents an ONERA activity on helicopter conditions at the starting point located on the ship landing operations simulation improvement. implementation. (p>: Random phasebetweenOand TjZMithaunform probability density It is interesting to compare these results obtained with turbulence and mean aerodynamic velocities.. A partial database was also generated at 25 Figure 28 illustrates the local wind components (UF- kts of wind with 0° side-slip. off time open It comes out from these open loop simulations that loop simulations of flights above the deck were carried turbulence can have a positive or a negative influence.4m) at 50 knots of wind speed. tests in ONERA-IMFL.z. Simulation ship state data (velocity and heading).

-M. Taghizad.P.V.V. [4] M. Follow-on activities will consist in : • completion of the data base with new wind tunnel tests for a 15° side-slip and a 10° roll angle of the ship . Marvis.6-6 important aerodynamic effects due to the hangar cliff. Modifications of the model will certainly be needed for real time applications. Shinozuka (Columbia University. 1-5. 1994. Ch. [2] H. J. Prasad. Schräge (School of Aerospace Engineering-Georgia Institute of Technology) A Simulation Methodology for Modelling Ship-Air wake Turbulence. . REFERENCES [1] A.R.R. Germany. A. New York). [3] D. velocity fluctuations (turbulence). New York) Digital simulation of random processes and its applications. connected to the Eurocopter HOST code (Helicopter Overall Simulation Tool) demonstrated important effects of the ship mean and turbulence aerodynamic wake on helicopter loads. Washington Aug. AIAA. D. Prasad. J. D. [5] P. (Eurocopter) Aerodynamic Design of the NH90 Helicopter These data were used in order to define and to develop Stabilizer. a ship air-wake model of the La Fayette deck area. Desopper (ONERA) Modelling thefregate La Fayette aerodynamic wake for helicopter landing operations To be published in Aerospace Science and Technology (AST). 25th Annual International Symposium of the Society of Flight Test Engineers.N. September 1997. Marvis (School of Aerospace Engineering-Georgia Institute of Technology) Ship air wake effects on helicopter rotor aerodynamic loads. Zhang. includes a mean air-wake model and a model of Dresden. C. It 23rd European Rotorcraft Forum. 111-128. Verbeke. This model. Journal of Sound and Vibration (1972) 25 (1). moments and generally on its flight dynamics. • transferring the model to Eurocopter for implementation on Eurocopter real time simulator. Eglin effect and ship lateral wall effects.N. Jan (Gibbs and Hill Inc. 1994.

9 m/13.9 ml 18.4 ml 4.i ^ I "H Figure 1: Frigate La Fayette 1/50 model Figure 2 : Test planes position Test Configurations .4 m/15.5 et 6 ISJ=ä| 1824 points .FrwdWindlS* (50kts) : Areas 2.4 m/ 8.4 m Figure 3 : Test points position Figure 4 : Summary of test configurations VITESSES MOVENNES NORMEES n. 6-7 FREGATE LEGERE «Gs=£T. ß = 0° .Wind = 50 kts.4 ml10. version crotale ED FT F^ m \mtt-G*. Z<m>: Niveau plancher Figure 5 : Mean velocities evolution along the vertical axis on the deck centre (pt A) .4 m/6.3.HearWndO" (50kts) :Area 1 et 1/2Area2 720 points I 1 WHDPipKTWrlS Planes height to the deck FuelS" Rearo* I I *""F 2.FrwdWindO* (50and25Ids):Areas 1 to6 2*2584 points .

5- -20 -10 10 20 30 | 0. ON THE DECK CENTRE -20 10 20 30 SYMETRY LONGITUDINAL AXIS §0.4m) Wind = 50 kts.6-8 VITESSES MOYENNES NORMEES g0.5- 3 -20 10 20 30 -0.2 30 -10 0 10 20 Yf(m) Figure 7 : Mean velocities evolution along the lateral axis on the deck centre (pt A. height=4. ß = 0° Figure 8 : 3D visualisation of tests at 50 kts .2 -20 -10 10 20 30 0.4m) Wind = 50 kts.5-.2 -20 -10 0 10 20 30 Xf(m) Figure 6 : Mean velocities evolution along the longitudinal axis on the deck centre line (height=4.2 -0.2 30 -20 -10 10 20 0.2 -0.5 Z> -20 -10 20 30 0. ß = 0° VITESSES MOYENNES NORMEES LATERAL AXIS 0. ß = 0° .2 -0.

4 m Figure 10 : 3D visualisation of tests at 50 kts . 6-9 height = 6. ß = 180° TURBULENCES NORMEES 150 200 S50 300 350 400 1OO 1SO E50 300 ■wtn*v^# 1 SO EOO 2SO 300 350 400 M O 1 50 200 WMpNfli2SO 300 350 400 Figure 11 : Example of normalised velocity fluctuations.ß = 0° . on the deck centre Wind = 50kts.4 m Figure 9 : 3D visualisation of tests at 50 kts . ß = 15° frtrr n ^ nm ww height = 2.

w PSDs evolution above the deck centre Wind = 50kts.01 5 0.005 0 0 0 3 5 3 5 0 5 3 0.01 0.01 03 Q.01 U.005 5 0 0 0 D 3 5 0 5 3 5 0.. £0. 3 5 ) 5 0 5 0 5 Z(m) 0.005 1.01 0.U 1 0. to 0.UI | 0.005 1.005 ^_ : 0 C £ 0 5 0 5 0 5 0.005 1.01 ■ V.005 1.oiC 0.ß = 0° freq(Hz) X(m) Figure 14 : Vertical velocity PSD evolution along the deck median axis (height 2.005 1.005 1.005 1.005 3.01 0. ß = 0° .01 Q.005 1.005 1.005 1.005 -1.01 0.005 1.005 3.4m) .01 5 0.01 0.UI Jreq(Hz) 5 0.005 .6-10 puv=f(z) puw=f(z) pv=f(z) pw=f(z) £ 0.01 V/.005 1.005 r% E o n 'V—.01 ""' 0.005 .01 —| 0.005 1.Wind = 50 kts .005 0 ^ 0 . o^. v.01 0.. 0 ~-—- 0 5 0 5 0 5 "0 5 Hz Hz Hz Hz Figure 13 : u.4m) .005 0 /v_ 0 A^_ 0 ^ E c 5 0 5 C J f o.005 Figure 12 : Vertical velocity PSD evolution with altitude above the deck centre Wind = 50kts:ß = 0° 0 0 ) 5 0 5 0 5 3 5 0.Wind = 50 kts .01 S 0..01 "'"■ 0.W1 5 0.005 ^ 0 5 0 5 0 5 0 5 W. ß = 0° Figure 15 : Vertical velocity PSD evolution along the lateral axis on the deck centre (height 2.005 1.005 1.

005 . ß = 0° / 50 kt wind ß = 15° puv=f(z) puw=f(z) puv=f(z) puw=f(z) Pv=)(z) pw=f(z) £ 0.005 S o 'V.01 ) 5 0 5 0.005 1..005 ■ .005 5.005 1.005 f<— 0 f^_ 0 r~~±^ 0 .005 0 5 0 5 0 5 u 0 5 5 0.005 1.01 3 5 0.005 1..01 0. w PSDs evolution in altitude on the deck centre Figure 18 : Comparison of u.005 1.005 1.005 1.01 ) 5 0. side-slip 0715° wind = 50 kts .005 1. i °005 1.01 J 5 0. side-slip 0715° ~P=0P figure 19 : selected points for comparison of tests at 50 kts and 25 kts (ß=0°) . tS 0.01 ) 5 0. 6-11 Figure 16 : Selected points for comparison of tests at 50 kt wind.005 0 5 0 5 0 5 0 5 5 0. t 0.005 I.005 0 0 0 0. wind = 50 kts .01 ) 5 0. (1 ^^_. 1. v.01 0.005 D.01 ) 5 | 0.005 1.005 ^^_ 0 0 J~±^*_ 0 0.„.01 ) 5 0.01 ) 5 0.005 0 5 0 5 0 5 0 5 5 0.005 1.005 1.005 1.01 0.01 ) 5 0.01 ) 5 5 0.005 1.01 ) 5 0. "0 5 0 5 "0 5 0 5 "0 5 0 5 0 5 "0 5 Hz Hz Hz Hz Hz Hz Hz Hz Figure 17 : Comparison of u.01 fflQ.01 £ 0. v.. w PSDs evolution in altitude on point B (pt A).\ 1.005 1..005 1.005 1. 0 /4^ 0 ~_.005 1.005 0 0 0.005 1.01 J 5 "Q.01 ) 5 ! 5 0.—.01 ) 5 0.

005 1.01 0.005 8 0.005 ot 0 0.01 a.005 5 o 0 ^ Figure 20 : Comparison of u.01 5 0.005 1.01 5 0. w PSDs evolution in altitude on the deck centre (ptA) wind = 50 kts and 25 kts .01 r 0.005 1.005 1.01 0.01 ( 5 0.005 0 0.005 0 0. 6-12 pv=f(z) pw=f(z) 1 °'01 £ 0.005 0 C 0 0.01 3j 0. v.01 ( 0.01 5 0.01 5 0. v.005 0 0.005 0.01 r 0. side-slip 0° puw=f(z) c 0. side-slip 0° wind = 50 kts and 25 kts . w PSDs evolution in altitude on point C Figure 22 : Comparison of u.01 0. side-slip 0° . w PSDs evolution in altitude on point D wind = 50 kts and 25 kts .005 0 5 0 5 0 5 0 5 Hz Hz Hz Hz Hz Hz Hz Hz Figure 21: Comparison of u.005 5 °- 5 o 0 t *.005 1.01 ( o 5 0 0.01 0. 0 0.01 ( 0. » 0.005 0 I 0.005 0 0.01 1 0.01 S 0.01 0.005 1.01 ( 0.005 1.01 005 1.01 0.01 E 0.01 5 0.01 S 0.01 0.005 / 0 V__ £ 0.01 0.005 1. v.005 1.005 CO 5 o *.005 1. 0.01 5 0.005 1.

identified . Figure 24 : Model limit conditions Spectre 1 h2 -T 1 r- 0 20 40 60 80 190 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200 . generated 0 1 2 3 4 5 6 Figure 25 : Example of vertical velocity turbulence generation on the deck centre .i .. 2. 6-13 Figure 23 : Test area elementary parallelepiped i Objective: j Avoiding velocity discontinuity when entering the test area i Method: 1 ..Creation of a transition area. ■r/A /f lt. 30 m 25 m n littiillli * —t-4-j— 30 m A IT' 1 . i E»i u n xf k 1 (1 (Vue laterale du na» .Interpolation of air-wake mean velocity between the test area and infinite free stream. !*„> ----Z-- 1 K=n f> .

..~\. slope angle = -1° relative wind 50 kts . SIMULATION NON LINEAIRE .. slope angle = -1° flight above the deck at 10 kts.:... frigate still (mean velocities and turbulence effect) 10:JS.. frigate still (mean velocities and turbulence effect) (mean velocities and turbulence effect) . N TETA DES PSI DEG k : I \ ■ X.6-14 Wind Frigate Speed Frigate Wake Aerodynamic velocities in the ship air-wake Helicopter Model Frigate Heading t Helicopter positions Figure 26 : Ship mean air-wake model implementation in HOST T SIMULATION NON LINEAIRE ..FICHIER 365N1_W0 HOST D W0!/)7 H:54 SIMULATION NON LINEAIRE ..! .. ■■ Ml.. ß = 0° .. frigate still relative wind 50 kts ... . ß = 0° . •-■:■■■■} ^V'J I. . ß = 0° .....FICHIER 365N1_W0 HOST D "I"/" CONDITIONS: equiU COKEITICSIS: equi!2 DV?-_FEWM/S.„!..Helicopter commands flight above the deck at 10 kts. i...!.... ..\sC\ YFH-F3H H . ZFH-FGW H \ ' f i •..\..Helicopter parameters flight above the deck at 10 kts....„. 1 1 Figure 29 : HOST simulation .'J : "^-L : X ■ ... equil2 • 5B-5 '.FICHIER 365N1_W0 HOST D ia/°V« ±S:St CONDITIONS.„.l... TEMPS S LK...Helicopter forces & moments Figure 28 : HOST simulation .... slope angle = -1° relative wind SO kts .. Figure 27 : HOST simulation ..:..

DWF-FGW M/S DKj.-.'.M... frigate still (mean velocities isolated effect) .. slope angle = -1° relative wind 50 kts . . 2520 248" 244..:. ß = 0° .:iSsj^:/:::...'.:::::::.. frigate still (mean velocities isolated effect) SIMULATION NON LINEAIRE . slope angle = -1° relative wind SO kts . .PW 2560 JK. \Z\Z'{Z'S^it . . :::::::::.RP . .-. H/S VMF-P3I M/S ..:::.. .M. ß = 0°. TEMPS S FY-CEL DAN 0 -20 ZZ\ZZ"^". :^f|^|\^^ 2400 2360 MYG RP H. Figure 30 : HOST simulation .FICHIER 365N1JI0 HOST D W09/97 »:l CONDITIONS: equil2 UMF-B™.."'r'.28: CONDITIONS: equil2 DOF-R3W H/S.. KMF-FGW M/S .Helicopter forces & moments flight above the deck at 10 kts..PJH MZG-.Helicopter parameters flight above the deck at 10 kts.. 11 „FY-EM? DAS Figure 31 : HOST simulation . 1 ..'.FICHIER 365N1_W0 HOST D "Z09"7 10.. 6-15 SIMULATION NON LINEAIRE .RP.DAN.f DVF-rcw M/S.MN. ..

In addition these limitations are due. It contains an experimental component and a increase in rotor collective. limits or SHOLs) is almost always reduced when compared to This paper deals with the analysis and characterization of the the normal operating envelope for the helicopter. the pilot region measured included the boundaries of the recirculation may choose to pitch. SUMMARY and decreased mean wind velocities in the lee of the ship The operation of helicopters from a ship-based platform is a superstructure. Cheney Aerodynamics Laboratory Institute for Aerospace Research National Research Council Canada Ottawa. performance of an immersed helicopter rotor. The Netherlands.T. which gives rise to increased difficulty placed in the wind tunnel for validation of the numerical in tracking the flight deck prior to landing. which generated a properly scaled characteristics of the helicopter when compared to the same variation of mean velocity with height. The operational envelope of National Defence (DND) to investigate the air wake and for such manoeuvres (known as ship-helicopter operating rotor loads in and around the Canadian Patrol Frigate (CPF). the experimental element maps out pilot's vision. In the case of smaller non-aviation ships wherein the asymmetries in the flow field caused by the 20mm Phalanx helicopter must land in close proximity to the ship CIWS found on the starboard roof of the hangar. In the Canadian context. B. It is evident that of the above-discussed factors leading to The results of the wind tunnel experiments show the reduced SHOLs. some of which are: inadequate yaw control. pressure-based Navier. "flapping" shear layers. several of the smaller structures around the flight deck were characterized by intensities and length scales. Over the past five years the 2. strong vortical gradients in the mean wind velocities as well as increased flows and turbulence with length scales of the order of turbulence levels. attitude away from a ship because of the non-uniformity of the air wake. or because the existing windshields on the the flow field (14x13x16 matrix) in the vicinity of the flight helicopter do not permit adequate viewing of the flight deck deck of a 1:50 scale CPF model at 0° yaw and 12° yaw. These motion cues are likely to occur modifications exist on the real ships. Increased turbulence arising from separated and/or To complement this work. to the air wake.F. A reduction in the usable cue computational one. tail rotor is reduced because of the air wake. The during the final stages of landing. . The atmospheric boundary layer was included in the fixed location over the flight deck will be increased. a reduction in usable cue environment thus The current Canadian efforts are focussed on obtaining wind requiring changes in helicopter attitude. This varying changes in rotor output and be sensed by the pilot's modification exists for these experiments only. and published in RTO MP-15. Large- numerical simulation. the effort required to maintain trim at a field. The air wake turbulence. all at-sea helicopter landings are made on small ships. as Stokes flow solver was used to compute the steady-state flow turbulence increases. limitations. A program is under way in the National Research Council of In the latter case the rotor is immersed in the reduced- Canada Aerodynamics Laboratory (NRC/AL) to analyze the momentum flow of the air wake thereby causing a decrease in air wake formed behind the Halifax-Class Patrol Frigate thrust and requiring the pilot to initiate a corresponding (CPF). to the operational envelope arise for a number of reasons. To address this issue completely. unsteady flow fields containing generates an air wake which contains significant spatial recirculation regions. Syms. response of the ship to excitation from the wave spectrum. engine-torque Determination of air wake characteristics is not a trivial task. 7-1 ANALYSIS OF PATROL FRIGATE AIR WAKES S. the flow field around a modified vortical flow in the lee of the superstructure will effect time- CPF was computed using a Navier-Stokes flow solver. Using hot-film anemometers and the AL environment can occur because sea spray is obscuring the 5m Vertical Wind Tunnel. so-called "non-aviation" ships presents the Air Vehicle Research Section of the Canadian Department considerable challenges to a pilot. unsteady flow field which exists over Air wakes. to manoeuvring near a moving platform. simplifications were made to the bridge and corrective action (McRuer. No such inertial cueing system. An atmospheric boundary improve the visual cueing. held in Amsterdam. A structured.J. to track large vertical deck motions in high sea-states and as the rotor translates toward the flight deck on the final descent. both may significantly alter the helicopter rotor diameter. many are due to the influence of the ship air recirculation zone behind the hangar as well as the wake. Many of challenging procedure from a pilot's perspective. in whole or in part. one must deal with a fully-coupled problem Inadequate yaw control will occur when the flow through the including both the helicopter rotor and the ship air wake. There Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". such as that in the lee of a ship superstructure are the flight deck. Zan. the pilot must the region of complex flow around the ship superstructure. the air wake will assume a greater importance. increased turbulence tunnel and computational fluid dynamics (CFD) data. This modified ship geometry was also amplitude ship motion. The flow over the ship superstructure characterized as complex. 1994). with the pilot's "aerodynamic workload. numerical results agree favourably with the modified CPF experimental results and give further details of the air wake. multi-block. bank or yaw the helicopter in order to zone that exists behind the hangar." That is. Reductions air wakes and comments briefly on the effects on rotor loads. INTRODUCTION Aerodynamics Laboratory of the Institute for Aerospace The launching of helicopters from and recovery of helicopters Research (IAR) has developed a collaborative program with onto small. The superstructure. as a first step. will correlate removed. In the latter case. Such changes in helicopter layer profile was created by placing a set of horizontal rods attitude can result in changes to the stability and control upstream of the model. is caused by the solutions. the ship air wake can be mapped out limitations can become important when the pilot is attempting to estimate what inflow the rotor might experience. 5-8 October 1998. In order to facilitate grid over a frequency bandwidth wherein the pilot will attempt generation. Canada K1A 0R6 1. G. respond to the complex. Torque However. and large-amplitude ship motion.

The boundary layer The turbulence intensity components are expressed as the root profile represented wind flow over rough water encountering mean square values of the wind gusts divided by the local the ship steaming into the wind. In practice. A turbulent. modifications were described by means of a least-squares fit to a quartic function. the air wake differences between the wind tunnel results beam of 0. The wind tunnel freestream velocity v'(x. The Reynolds used in the CFD calculations. CFD calculations were made on a measured at 24. 1975). based on the ship plan. encountering the ship steaming "Modified" CPF configuration.328 meters.7-2 exists some full-scale airwake data acquired from a CPF wind represented wind velocity relative to the ship at a height of over deck survey (Kowal. the aft fire control radar from the front of the hangar will be in error. Specifically. the data set that provides the second orthogonal funnel and mechanical units were also not included in the direction also provides a redundant set of data for the model. the errors associated with turbulence intensity values are small. constructed from applicable drawings (see Figure lb). The solid wall provided a flat.15m above sea level (ASL). wire starboard were investigated.v. 3. this experiment simulated a 46-knot wind.z) V(x. EXPERIMENT DETAILS A 1:50 scale model of the above-water form of the CPF was Wind tunnel measurements were made of the longitudinal. made to the CPF wind tunnel model in order to match While the hot-film probe has good resolution and fast precisely the "Modified" CPF model (MCPF) used in the response. However. and a controlled environment when compared the ship's anemometer height of 24. as determined from the wind tunnel freestream. With the length scaling chosen to U(x. u'fx. and Chevray.z) is the local rms streamwise velocity The wind tunnel velocity was held constant with a freestream V(x.z) v'fx. By rolling the probe axis 90°.6m/s(40 kts) relative to the ship at a height of 10 meters possible. well in excess of the minimum 1 lxlO3 recommended for the wind tunnel modelling of ships (Healey. lateral and vertical winds over the CPF helicopter flight deck. From an anemometer is capable of measuring flow in the nominal aerodynamic point of view however. The structures reversing flows. the results yielded roof. Although these are not equal. wind tunnels have the advantages of reduced for the purpose of this paper. a ship at sea has only air wake data. When acquiring meteorological practice for a reference wind speed.y. especially when simulation is the end-product an anemometer with which to measure the relative wind of the air wake data.z) U(x.z) W(x.y.z) For an experiment such as this one that is insensitive to Vv anem Reynolds number effects and conducted in incompressible flow. A minor intensity and a higher than actual velocity component (Tutu modification was also effected to the bridge. The 15' circular test section of the tunnel was a turbulence intensity of 70% (Keffer.y.z) be 1:50 as previously mentioned. This boundary layer profile streamwise wind component. air wake data. The measurements were made antennas. the wind speed as measured at cost and time. a small lattice radar-mast and 57mm with a TSI cross hot-film anemometer. CFD results must also be validated. The which to mount the ship model.5 meters ASL will be used to at-sea trials. the films individual wake signatures of small structures will blend into are then sensitive to the second orthogonal direction.v.z) is the local rms lateral velocity . the longitudinal flow Vulcan Phalanx CIWS from the starboard rear of the hangar directions determined from the mean velocity components roof. in reality. and the time scaling chosen to be 10:1.z) time and velocity scalings. 1991) and in this paper the current 40. inserted at the test section entrance. the data were is representative only as. U(x. the velocity scaling defaults to 1:5. streamwise flow component. since it is possible to include at 10 knots into wind. where U(x.y. More explicitly. In speed. modified/removed from the CPF model were: the 20mm such as in the near wake of the hangar.z) is the local mean streamwise velocity u'(x. the for the two configurations can be examined to determine the model had sharp edges that reduced the sensitivity to sensitivity of the air wake to superstructure configuration Reynolds number effects. it cannot differentiate between forward and CFD analysis of the same problem. This comparison is intended to "validate" the wind ASL. the Reynolds number is changes.v. In highly turbulent flows. This represents a wind speed wind tunnel data is compared to those measurements where of 20. In the flow field as one moves aft.y.v. graded sea measured wind speeds were normalized by the reference wind boundary layer profile was generated by a grid structure speed. This height is consistent with the standard tunnel as a tool for airwake measurements.z) U(x. which had the identical geometry to the model 30xl06.5 meters ASL.. Vane. a second experimental configuration was wind The Reynolds number for the full scale CPF is approximately tunnel tested.z) w'fx. the wind profiles would reduced to the following non-dimensional quantities: vary according to the varying sea states and ship speeds. modified with a test section insert to obtain a rectangular 10' x 7' open jet section with one solid wall representing the sea Each measurement consisted of 2048 data samples acquired at surface. Thus. Thus. based on the ship beam of 16. 1986). To show 56 knots on the anemometer reading.z) is the local mean lateral velocity velocity of 5 m/s. handrails. The moving ship in this case would only limited structural details in a numerical model. overcome a possible limitation with respect to validation of the CFD data. Several of the smaller structures on the ship and ahead of the Approach wind directions of 0° (head wind) and 12° from helicopter hangar such as missile launch installations. and the by the probe will show a lower than actual turbulence large lattice radar-mast from atop the bridge.v. 1992).y.v.4 meters. the experimenter is free to choose any two of length. the present context. under conditions of reversing flow. Exhaust flows from the ship's addition. As a by-product of this test number for the experiment was 107xl03. A cross hot-film cannon were not included in the model. the landing officer's shelter from the flight deck.y. the relationship between cooling velocity and output voltage was In the second part of the experiment. In addition. For the film calibrations. CFD offers an attractive option for producing as the reference wind speed. vertical surface on a rate of 100 Hertz for each of the two hot-film sensors. being about 2% for The CPF model was installed in the NRC-IAR Vertical Wind turbulence intensity less than about 50% and reaching 12% at Tunnel. the air wake should still streamwise direction and in an orthogonal direction that is in be highly representative of a fully-detailed CPF since the plane of the films.

3m x 41. 24.9 ship lengths (L) ahead of the bow and the domain was 14.9m x 14. solution. For this replaced by an inward lateral component on the starboard work. The starboard side exhibits far greater lateral solver stability. the intensities are still a factor of two to (6. 7-3 W(x.51 m2/s2) than that in the referred to analysis (3.0m above the surface of the eye on the flight deck as shown in Figure lb. The results suggest flow the MCPF. This Figure lc.51x10 3 intensities exceed about 35%. The figure helideck. Flow recirculation exists when turbulence coefficient of 2.z) is the local rms vertical velocity while the downstream boundary was treated as an "outlet" or and Van. The asymmetry in removed and the bridge shape was modified to introduce the air wake due to the presence of the Phalanx gun located on more 90° angles.53x10 3 and a roughness length of 3. Grid spacing near solid over most of the flight deck. RESULTS AND DISCUSSION relative to the ship. a topology based (Figure 2). solved or the geometry being analysed. The fact that the air wake is angled The multi-block grid created for this geometry was "H". the CPF model.9m aft of the hangar and from 2. free stream (Figure 4) depict the air wake directed inward based solver employing a SIMPLEC pressure-correction over most of the flight deck.3 L long. The upstream boundary was anemometer height. The hot-film anemometer was installed on a traversing mechanism that was free to move in the vertical direction 5. but for the vertical velocity is presented in Thus with a fair degree of confidence. The ocean with a wind velocity of 20m/s at 10 meters off the levels of turbulence increase monotonically as one approaches surface. The code is a finite volume. multi-block Navier-Stokes The projection of the flow vectors onto planes normal to the flow solver using body-fitted coordinates.y. In this and sub-sequent similar plots. it could be expected that the non-uniform atmospheric boundary layer giving rise to a steady-state inflow will have an effect on the helicopter trim. These conditions generate a surface skin friction the flight deck. These turbulence will no doubt contribute to pilot workload. streamwise and vertical air wake velocity components. As mentioned previously. it is as though the downward vertical component of the port side is largely CFD-ACE offers a range of turbulence models. A similar plot. Further work is being carried out into refinement of this grid to Significant gradients exist in the time-averaged value of the address grid sensitivity issues. This said. Thirty-five blocks containing nose-down pitching moment will be generated and this is approximately 903. the advantages of using Figure 5.z) is the local mean vertical velocity profiles were applied to the upstream. Once more. a cut-out is indicated reflecting a region of flow around the frigate. The choice made here The flying probe was not used here. The origin of the coordinate system is the bull's 3. 7. the lateral component of velocity reportedly and Spalding. functions will produce answers similar to those of the k-u model integrated to the wall on simplified frigate shapes. For the slices closest to A CFD simulation will provide a more detailed picture of the the hangar (x < 0). CFD DETAILS widens as it progresses downstream. It is a pressure.1 0° Yaw Wind Tunnel Data longitudinally in the wind tunnel. and as demonstrated in Syms and must be simplified by either simplifying the equations that are Zan (1995). Previous has the least effect on helicopter operations for the 0° wind work (Syms 1997) has shown that the k-e model with wall direction. Vertical turbulence intensities of up to 20% suggest local flow . reflecting the fact that the boundaries was kept so that y+ values ranged between 50 and streamlines are bending toward the region of low pressure 150. 1974) was used with wall functions. behind the hangar face. depicts a classical bluff-body wake in that the momentum deficit is lessened as one moves downstream and the wake 4. Again the asymmetry due to the Phalanx gun is evident. asymmetry about the ship centreline is scheme with a blending of first-order upwinding to maintain evident. asymmetry is lessened as x becomes more positive.was moved laterally or 5. again reflecting the streamline algorithm. It is clear that the air wake is directed downward wall functions can be realised.5 meters ASL. available software package CFD-ACE (CFDRC.5 L wide and 3.5m. The details of the geometry are found in the aft starboard corner of the hangar is evident. In fact. The turbulent kinetic energy was set slightly higher recirculation zone. 1998). The convective terms in the governing equations deformation toward the region of low pressure behind the were discretised using a second-order accurate upwind hangar. The 14X x 13Y x 16Z The variation in streamwise velocity component in the matrix of surveyed points defines a full-scale rectangular vicinity of the flight deck as a function of location is presented volume measuring 41. to get such an answer recirculating flow. the flow considered reliable here.000 grid points were sufficient to model corroborated by pilot testimony.7 L high. considering the complexity of the MCPF model. the increased in the CFD model which exist in the real atmosphere. side and top boundaries w'(x. Even in the region outside the meters. 1986). Streamwise turbulence was to simplify the CPF shape.26 m2/s2) to three greater than that found in the atmosphere. valid data can be obtained with a flying probe. velocities than does the port side. The surface condition was taken to be an velocities are shown in Figures 6 and 7 respectively. placed 3. the x-direction has been The full-scale MCPF was studied using the commercially stretched to aid in visualization of the data. As these gradients are manifested over distances of the order of a The ship was considered to be stationary in a steady rotor diameter. Profiles in streamwise velocity and length scale (Figure 3) were based on a logarithmic profile (Simiu and The turbulence intensities for the streamwise and vertical Scanlan. extending from in Figure 4. Since the compensate for the lack of turbulence sustaining mechanisms length scales are of the order of a rotor radius. downward suggests that as the rotor enters the air wake. Stationary hot-film data cannot be using a reasonable amount of computing resources. This could be considered a medium-size grid angles of up to six degrees in the absence of rotor downwash.y. In order to survey the air wake in three dimensions.™ is the velocity of the relative wind at the constant pressure boundary. a intensities of thirty-five percent or more are used to define the "Modified" CPF was designed in which smaller pieces were boundary between valid and suspect data. However. the k-e model based on Launder and Spalding (Launder side. structured.

the The CPF and MCPF configurations were both tested in the velocities at Location 12 are higher than at the other locations wind tunnel. There is some fluctuations of about 2.2/5. of one of the sea-trial data points. It can be is generally good agreement between the wind tunnel and sea.0/0.2/-0. Location X Y (arbitrary (full-scale in (full scale in For the vertical velocity component.5 meters above components.5 meter height. The wind tunnel 5. Without forward water locations identical to those for the sea-trial data. it is also accepted that the wind tunnel cannot experiences intermittent flow reversal (Zan and Garry.1 XJ/Vmem. one sea-trial data.6 5.2/-0.8 0. or falls within location of the measurement at 9 meters above deck. These values suggest a region that measurements. the wind tunnel data were interpolated using a surmises. UV) wind intensity normalized by the ship anemometer velocity.5/-5. for the following (x. data features consistent trends with the sea-trial data. increase the mean velocity and decrease. There 5.0 horizontal resultant (i. classic bluff-body air wake over the helideck. 5. 20 Figures 4 and 5.8. being pushed off heading by both the wind and 9 meters. 1994).5 location in compensation for the error. The consistent positive values for the data were averaged for a full-scale equivalent of 204. the 20-knot sea-trial data was achieved with a bi-linear interpolation algorithm to arrive at values for stationary ship facing into wind. u'/U would be some value (<30%) and the of the statistical uncertainty in the measurement and not of statistical skew of the data measured by the cross-film different air wake characteristics. offsets (. anemometer would be near zero. Accepting the presence of error. Where the data is not an exact match. This likelihood is further increased when one considers that the sea trial data was taken In order to facilitate comparison of wind tunnel data with the under conditions of a prevailing 20-knot wind. the ship would be unable to maintain a heading for data were acquired at two heights above the flight deck. the practical limit in the situation. the wind tunnel data for the 5. As previously explained.y) locations: and the waves.6 -5. due to high turbulence levels recorded in the knots and 30 knots. the numbers preceding the slashes refer to as opposed to separating out the streamwise and lateral the (x. generates a for both heights. The 30-knot data would thus be achieved by steaming the ship at 10 knots into the prevailing wind. The fact that different values presented here for comparison purposes. test. The wind tunnel data falls within this scatter range for every location at the 5. pitching moment and rolling moment. for example. by small amount. an even more meter height while seventy percent match at the 9 meter favourable agreement might be obtained with the sea-trial height. The latter speed. 13 -1. which is a data are in reasonable agreement. The sea-trial data is more limited in that considerably smaller and more difficult to measure in both measurements were acquired at fewer locations. Figures 9 and 10. In particular.3°. and allowing much greater control of alignment with the Table 1: prevailing wind.0 averaging times. certainly due to a "red" wind. although no Reynolds number effects hot-film measurement. For a blade tip speed of 250 m/s in a Figure 8 compares the normalized streamwise velocity 20 m/s wind. Streamlines follow downward and inward trajectories as described below. However. Eighty percent of the data points match at the 5.2 The turbulence intensities. Every wind tunnel data point matches. Therefore. Figure 11. the turbulence intensity. the Despite these limitations to the comparison of data sets.7-4 angles approaching 30°. noted. these wind tunnel data are compared 9 meter heights. this suggests local blade angle of attack components of the wind tunnel and at-sea data. but probably an insufficient sea-trial data will be compromised by small changes in length of time to ensure the results are statistically significant approach wind angle. 1991) acquired using a standard three-component vane-type The lateral and vertical velocities. are anemometer.e. For these comparisons. reflecting the short 8 -6. the value is nevertheless would be expected for this flow. while in the wind tunnel no such for a stationary process.5/-6. a symmetric model. times. The actual measured u'/U value at Location 8 is 47% and the statistical skew of the In addition to the difficulties associated with obtaining at-sea U/Van. the wind tunnel data were acquired for a period of 30 seconds. but only by about 0. For example.5 extended periods. .5 meter height at Location 8 was omitted in the air wake plots. are plotted as the 18 5. The data are presented in this way. Again. The sea-trial results present data at two wind velocities.0/0. there wind tunnel value should be used with caution. The 1% turbulence intensity. In non-reversing low were obtained for each of the velocities is probably indicative turbulence flow. there is again number) meters) meters) considerable scatter in the sea-trial data. however.2/-0. that by reducing the U/Vanem value at this trial data. the wind tunnel data. enough to invoke significant scatter in the sea-trial data.0/0. In the above table.5 trends in the sea-trial data. Therefore. The wind tunnel data are consistent with the 12 -1. duplicate faithfully all details of the actual CPF. Thus data at location 8 are marginal in terms of acceptance effects of stack gas efflux are not modelled in the wind tunnel based on previously defined criteria. On a point of clarification.61m) are due to specific anemometer positions on the towers on which the anemometers were mounted. By way of contrast the wind tunnel concerns arise. The sea-trial cases than the streamwise component. it is believed that the sea-trial 30- Locations of Measurements for Sea-Trial Data knot data would be more accurate than the 20-knot data. The MCPF. since the latter are unavailable for the sea-trial deck and the numbers following the slashes refer to the data.y) location of the measurement at 5.0 0.2/5.™ time history is 0.3 MCPF and CPF Air Wakes at 0° Yaw Comparison are also consistent trends in location. though this is not a difficult aspect to incorporate in an the result of such a flow condition on the interpreted data is to experiment.6 0.2 Wind Tunnel and Sea-Trial Data Comparison underestimated the streamwise components at three of the five For validation purposes.8 normalized sea-trial lateral velocity data at 20 knots is almost seconds (3 min 25 sec).0 14 -1. likely due to the short measuring changes in rotor thrust. to full-scale at-sea data taken on a CPF (Kowal. local U/Vancm velocities increase with height.

but remains a conjecture in the show the rather benign flow around and over the hangar onto absence of experimental and/or numerical analysis. Either could be edge of the funnel and not being dissipated by the time it used for CFD validation. 1991). (It must be remembered that the high wind speeds as the landing is effected. side of the ship over the flight deck. surface shows that the flow re-attaches just aft of the centre of At some point a control limit would be reached. the actual angle was 11. CONCLUSIONS ship. The magnitudes of vertical velocity for the 12° yawed wind Montinuzzi and Tropea. analysing the flow ahead of the hangar would horizontal component of the air wake contains a gradient in provide only a small amount of information pertinent to the direction as a function of vertical distance from the deck. In particular. The topology of this air wake is similar to wakes seen behind three-dimensional bluff bodies (Larousse.4 0° Yaw CFD Results While the predominant wake deficit is aligned with the free The results of a numerical simulation provide details of the stream at higher elevations.8°). the separation zone immediately flow around the complete MCPF geometry. the CFD results were indicates a downwash on this side as the streamlines curve interpolated onto the experimental grid. This • Good correlation between wind tunnel data and full-scale is a consequence of the flow being energized by the front sea-trial data has been demonstrated. suggesting The first point to observe is that both flow fields show the the rotor downwash dominates in this particular case. The presence of the CIWS is thus seen to cause velocity is shown in the upper half of Figure 14. or lateral velocities difference between the two air wakes is generated by the CPF (not shown) can be characterized in much the same way as for large lattice radar mast. To facilitate comparison of angle. (Because of the vertical flow above the port side of the helideck is almost installation restrictions. For study of the air wake. The flow on the starboard side possesses a the air wake. results show a much more diffusive flow. wind tunnel to effect a "green 12" wind. the flow on the starboard side of to further validate the CFD results. Another significant not the wind axes. The major difference between the measured and This paper has examined the air wake of a CPF using several computed velocities exists in the flow gradients. The V-component. 5.5 Wind Tunnel and CFD Data Comparison by the onset flow rising over the flight deck and probably The MCPF geometry was also examined in the wind tunnel rolling up into a tight vortex with a relatively shallow helix for purposes of CFD validation. There is a wake behind the funnel and recirculation zones behind the hangar and the stern of the 6. the experimental sea-trial data and computational fluid dynamic simulation. and associated turbulence. Figure 13 shows a toward the low pressure region aft of and beside the ship comparison of the two sets of data. Since changes aft of the hangar and closer in elevation to the flight deck is were made to the bridge and all smaller add-ons were more closely aligned to the ship axis. the CPF model was yawed 12° to port in the greater lateral inward component than the port side. zero streamwise velocity). Other horizontal velocity contours are approximated by the U- differences between the ship models were not seen to have component data presented. reversing flow is suspected. This is borne out in practice. The remaining traces SHOL limit is plausible. that 5. the following conclusions can be computed solution exhibiting higher values in each of the drawn: three velocity components. Again data are "blanked" where appreciable impact on the airwake. reaches the flight deck. are depicted in the lower half of Figure 14. The focuses on the area behind the hangar. SHOLs defined by cyclic-input limitations are uncommon. the mast and extends well beyond the stern of the ship. the present discussion rotor and the yawing moment acting on the fuselage. Thus. at the height of the velocity magnitudes are appreciably greater. The U- a substantially different air wake structure over the helideck component is defined with respect to the ship axes system and than might otherwise be expected. The transparent surface marks the region of the angles would result in larger gradients and vortical flows recirculation zone (i. is somewhat different. this type of flow changes to the bridge were assumed to have minimal impact could be expected to affect the thrust generated by the tail on the flow over the flight deck). Although a reasonable agreement is obtained between the computed and measured data.e. Conversely. the CFD sources of information: wind tunnel measurements. a more comprehensive . This which would require a larger corrective input from the pilot. although the U-component flow deficit. Displayed are a set of above the sea surface. foot of the vertical horseshoe vortex standing in the The connection between air wake directional gradients and the recirculation zone behind the hangar. This results in the From this investigation. change in tail rotor output and fuselage moment will induce a yaw rotation of the helicopter. These changes in direction of vertical crossflow planes which are coloured according to the flow with lateral position could lead to rapid changes in rotor- nondimensional streamwise velocity with the vectors generated rolling moments as the pilot traverses from the lee representing the in-plane velocity components. While experiment obtaining more detailed information (in particular the flow on the port side of the CPF centreline is akin to a about the size and shape of the recirculation zone) is required classic bluff-body air wake. This mast contributes to a streamwise the U-component velocities. same flow topology. The CIWS generates a streamwise flow deficit. It is evident that the windward side of the deck experiences an upwash caused 5. The completely deleted on the starboard side due to the presence variation in the time-averaged non-dimensional U-component of the CIWS. thereby the flight deck. and associated turbulence. however. So. However. thereby requiring corrective Figure 12 shows the general character of the flow over the action by the pilot. the streamwise component shows a larger value at the higher elevations. In general. centreline is decidedly not. It would be expected that larger wind flight deck. 7-5 The CPF air wake. full-scale showing much higher gradients. This suggests the removed. data for the leeward side of the flight deck the CFD to experimental data. Further.6 12° Yaw Wind Tunnel Data extends 15m aft of the hangar until smoothing into the larger In order to assess the effects of a non-zero incident wind on air wake structure. the flight deck. The black particle trace shows the outline and defining one part of the SHOL.

T. Huntsville. 559-564. Three-Dimensional Obstacles. E. B. 785- 800. Syms (1997) Unpublished work. Control. S. (1992) "Establishing a Database for Flight in the Wakes of Structures". S. Lecture 3: A Short Course on Hot-Wire and Laser-Doppler Anemometry. nl. • Based on examination of velocity and directional gradients in the air wake.F. Montinuzzi. August 1995. Department of National Defence. Department of Mechanical Engineering. Ottawa. A. (1994) "Interdisciplinary Interactions and Dynamic Systems Integration". and Chevray. 7. (1974) "Computational Methods: Applied Mechanical Engineering". (1991) " Wind Over Deck Survey & CH- 124A/CPF Flight Deck Qualification Trials".. Applied Aerodynamics Laboratory. .J. R. (1995) "Analysis of Rotor Forces in a Ship Air Wake". Aerospace Engineering Test Establishment. J. Further study is required to quantitatively correlate air wake characteristics and helicopter operations. J. Larousse.7-6 • The air wake is substantially affected by the addition of the Phalanx CIWS atop the starboard aft corner of the hangar. National Research Council of Canada. December 1991.V. (1975) "Cross-Wire Anemometry in High Turbulence Intensity". G. v59." Proceedings 8th Symposium on Turbulent Shear Flows. Keffer..B. R. H. Syms.E. 71 pp. N. E.A. REFERENCES CFDRC (1998) "CFD-ACE Theory Manual" CFD Research Corporation. Canada NRC- LTR-AA-13. v. Launder. Paper 31.J. Healey. Canada. James R.. Alabama. Munich. Simiu.H. Waterloo. Jul- Aug 1992. v29n4. D. and Zan.K. (1991) "Flow Around Surface-Mounted. R. J. p269. pp. and Tropea C. (1986) "Introduction to Hot-Wire Anemometry". D.J. • The current CFD solution produces a flow field with the correct topology but with higher velocity gradients indicating an under-diffusive flow in the numerical simulations. Zan. Kowal. J. 1998. and Spalding. inferences can be draw about the qualitative effects on helicopter operations. 1994. Int. Ontario. Aircraft. Flight Dynamics Technical Memorandum FD TM 91/05. FRG. (1986) "Wind Effects on Structures". Aerodynamics and Aeroacoustics of Rotorcraft. 1986. (1994) "Wind Tunnel Measurements of the Air Wake Behind a Model of a Generic Frigate". McRuer. Ontario May 1986. 1991. and Garry. pp3- 12. University of Waterloo. Tutu. 1975. 1974. Volume 3. Fluid Mech. John Wiley & Sons. and Scanlan. AGARD-CP-552. June 1994.

A.i . . .ft H i r \\ it b) Canadian Patrol Frigate (CPF) Wind Tunnel Geometry c) Modified Canadian Patrol Frigate (MCPF) Wind Tunnel and CFD Geometry Figure 1: Wind Tunnel Models and Measured Air Wake Matrix . 7-7 a) Canadian Patrol Frigate: HMCS Toronto (FFH 333) Displacement: 5235 Tons Full Load Dimensions: 135.15 meters I:: BifS.5 x 16.4 x 6. s.

7-8 Figure 2: Sample portions of computational grid 500 5 10 15 20 0 50 100 150 200 250 300 350 400 450 500 Streamwise Velocity (m/s) Length Scale (m) Figure 3: Numerical Boundary layer profile .

50m U/Vanem ÖÖÖ ÖÖOÖÖÖÖO© I I I Figure 4: Variation of U/Vanemometer in YZ Planes (Note: X dimension has been visually scaled by 2.96m to 20-96m z is from 2.5 for clarity) .07m CPF Helideck y is from -2096m to 20- z is from 4-50m to 16-50] Figure 5: Variation of W/Vanemometer in YZ Planes (Note: X dimension has been usually scaled by 2. 7-9 x=2707m CPF Helideck y is from -20.00m to 16.5 for clarity) x=27.

67m x=-468m x=-1103m Figure 6: Variation of % U Intensities in YZ Planes (Note: X dimension has been visually scaled by 2.02m x=1.96m z is from 450m to 1650m x = 802m x = -468m x = -1103m Figure 7: Variations in % W Intensities in YZ Planes (Note: X dimension has been visually scaled by 2.72m y is from -20.5 for clarity) x = 27.5 for clarity) .96m to 20.00m to 1650m x = 8.96m z is from 2.07m CPF Helideck x = 20.7-10 x = 27.07m CPF Helideck y is from -20-96m to 20.

03 ]I -0.05 ▼ A A !♦.2 A 0.05 ♦ Sea-Trial: 20 Knots ♦ 0.09 ■ -0.0 1.20 10 14 10 14 18 Tower Position Tower Position Figure 9: Comparison of Wind Tunnel and Full-Scale V/Van(.05 A A ■ iL A II -0.8 1 1 ♦ ♦ T E 0.3 A 1i 0.0 Meters Above Helideck 1.1 0.13 -0.7 E 0-7 A A ▲ A s > 0.00 -0.♦ Sea-Trial: 20 Knots -0.30 ♦ Sea-Trial: 20 Knots 0.9 ■ 0.05 ■ - A 0.5 Meters Above Helideck 9.6 g 0.15 < -0.10 ■ .4 4► s i * iA I 0.1 ■ Sea-Trial: 30 Knots A Wind Tunnel 0.11 -0.20 -0.20 ♦ ♦ ♦ <► 0.0 10 14 18 10 14 18 Tower Position Tower Position Figure 8: Comparison of Wind Tunnel and Full-Scale U/Vanem Velocities Relative Wind .07 -0.0 Meters Above Helideck 0.01 i ■ ■■ -0.03 ■ Sea-Trial' 30 Knots 0.10 0.0 Degrees Relative Wind .0 Degrees 5.15 10 14 10 14 18 Tower Position Tower Position Figure 10: Comparison of Wind Tunnel and Full-Scale W/Vanl:m Velocities .07 -0.15 -I -0.0 Degrees 5.5 ca 0.05 A 4k -0.11 ♦ Sea-1 rial: I J Knots -0.13 .3 0.15 ♦ 0. ■ .5 | 0.01 w i ■■ A^i ►k -0.0 ■ 0.5 Meters Above Helideck 9.15 0.03 A Wind Tunnel 0. 7-11 Relative Wind .■ -0.15 1 ■ ■ Sea-Trial: 30 Knots A Wind Tunnel -0.5 Meters Above Helideck 9.05 0.25 0.6 A tt A T 3 0.01 -0.09 -0.4 S 0.0 Degrees Relative Wind .05 ▲ 0.0 Meters Above Helideck 0.10 ■■ -0.30 i 0.8 A Wind Tunnel 0.0 ♦ Sea-Trial: 20 Knots 0.9 i 0.0 Degrees 5. M Sea-Trial: 3 3 Knots A WindTunni:1 -0.20 A Wind Tunnel 0.03 ■ jL -0.2 ♦ Sea-Trial: 20 Knots ♦ 0.05 ■ ft a -0.25 0.10 ■ ♦ <► <► 0.ra Velocities Relative Wind .01 0.0 Degrees Relative Wind .00 A 0.

.7-12 Relative Wind .5 Meters Above Helideck 9.0 Meters Above Helideck 25 25 ♦ Sea-Trial: 20 Knots ■ Sea-Trial: 30 Knots 20 A ' A Wind Tunnel 20 ♦ ♦ <► 15 15 £ A A :« it > ♦ ...0 Degrees 5.0 Degrees Relative Wind . ■■ 10 f ♦ A 10 ■ X i ■■ 1 ■ T ♦ Sea-Trial: 20 Knots ■ Sea-Trial: 30 Knots A Wind Tunnel 10 14 18 10 14 18 Tower Position Tower Position Figure 11: Comparison of Wind Tunnel and Full-Scale Turbulence Intensities Figure 12: Boundary and Topology of Flight Deck Recirculation Zone .

3J O. CFD Solution .2J Oil 0-) -Oil -024 -0-3-3 CFD Figure 13: "Modified" CPF Wind Tunnel Results VS. 7-13 Wind Tunnel U/Vanem 11- 1-j 09-j 084 0-74 064 O-4J O.

7-14 c •8 o tt a.o o a c~i ■o u CO o. 2 co l II N c > C a .

This is particularly true for the military. held in Amsterdam. As the of the most significant technical challenges. light level and visibility. The activity under an international collaborative panel of unsteady air wake generated by the ship as the wind The Technical Co-operation Programme since 1990. and Research Agency. under the co. 506 Lorimer Street. The Netherlands. . The pilot. With simulation. and finally to the position over the and training.A COLLABORATIVE VENTURE C H Wilkinson SJZan N E Gilbert J D Funk Flight Management Institute for Aerospace Aeronautical and Maritime Naval Air Warfare and Control Research. It summarises the work Clearly. The difficulties are frequently compounded by requirement to dedicate costly ships or aircraft to this low visibility obscuring the horizon and high sea non-operational role. At night the used to test new ship designs for aircraft visual cues available to the pilot are degraded or lost compatibility before the ship is built.2 Role of modelling and simulation ordination of the authors. result that there is insufficient manoeuvre power to compensate for the air wake disturbance. development of procedures. MK41 6AE. putting into an operational importance to civilian operators. and the value of air wake through improved training. Ottawa. analyses to vital visual references. can be maintained with none 2. MD 20670. Research Laboratory. as well as the simulation in research and training. Bedford.1 Ship air wakes . The paper simulation offers substantial opportunities to increase demonstrates the applications of air wake prediction the operational availability of maritime helicopters in ship and aircraft design. altogether and lighting is often limited due to tactical Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". new technologies and qualification techniques. including the air wake. 2. K1A 0R6. aircraft dangerously close to the flight deck and A number of approaches are being pursued superstructure. qualification aircraft flight path. Maryland. and published in RTO MP-15. the pilot improved understanding of the key features of the air may reach the limits of control authority with the wake topology and to develop a validation database. In extreme wind conditions. sea state. Unexpected gusts may force the correct for 'troublespots' in the design of new ships. to modelling and simulation of the air wake problem but an all-weather capability is also of prime at the dynamic interface. he has less spare capacity the collaboration are also actively involved with to consider his next move and the situation becomes conducting full. Flight Dynamics. for example those context the more technical aspects to be discussed serving off-shore oil installations. more empirical. initially to the effectiveness of simulation in research. is still expected The development of simulation for application to the to land on the deck within a tightly defined area helicopter-ship dynamic interface has been an without overstressing the undercarriage. helicopter enters the ship air wake. This paper details the collaborative efforts being pursued by the member nations. however. National Research Council. or may move the helicopter away concurrently aimed at applying Computational Fluid from the ship into a position where the pilot loses Dynamics and other. the operators of maritime helicopters would being conducted and draws together the various aim for maximum aircraft availability and the ability research aspects and validation tests. Modelling and elsewhere in the Symposium. While the pilot is fighting to achieve a modelling capability. the pilot is forced Increasing the fidelity in this area can enhance the to compensate for disturbances. Technology Organisation. flows over the superstructure exacerbates the Modelling the ship air wake and its effect on problem still further and can impact severely on the helicopter behaviour is regarded by the panel as one operational availability of the helicopter. and open up opportunities to predict and landing spot. Center Aircraft Department. The aim is to to operate in the most severe environmental present a comprehensive and co-ordinated approach conditions. United Kingdom Canada VIC 3207. All participants in maintain accurate position. Landing a helicopter on the flight deck of a moving Piloted flight simulation enables aircrew training to ship is one of the most demanding of all piloting be conducted safely and significantly reduces the tasks. 5-8 October 1998. while off-line simulation can be states resulting in severe ship motion. Fishermens Bend. Patuxent River. Ontario. Building 2187 MS #3. Defence Evaluation Building M2.the problem of the flight safety issues associated with live flying. Defence Science and Division. Australia USA 1 SUMMARY considerations. control of 2 INTRODUCTION the environment. 8-1 MODELLING AND SIMULATION OF SHIP AIR WAKES FOR HELICOPTER OPERATIONS .and model-scale testing to gain an unpredictable.

ship motion. for ships such as destroyers and under the auspices of The Technical Co-operation frigates the area is frequently not much larger than Programme (TTCP) since 1990.2 Ship helicopter operating limits (SHOLs) features of an air wake topology and to develop a In order to assist pilots. in certain wind conditions. for Aerospace Research (NRC . interface (DI) simulation community. The paper discusses the piloting problems ship motion and the influence of the helicopter rotor associated with ship air wakes and the necessity for combine to produce a highly complex and dynamic air wake modelling. The development of comprehensive and co-ordinated approach to these SHOLs is an expensive and exacting task modelling and simulation of the air wake problem at currently determined through intensive and costly the DI. wake models.1 Effect on pilot workload and performance wind tunnel. The goals of the work are to progress the aerodynamic interactions of the helicopter with the application of simulation to maritime helicopters in turbulent air flow near the ship. the pilot may have addressed under the topic. these limits This paper is intended to provide an overview of the are not considered in this paper. design.3 TTCP collaboration and involvement them to safely land on a moving deck. is regarded as one of the most symposium. to maintain position relative have been actively involved with improving their air to the ship. However. The working-level the helicopter. Validation of an air wake addressed.8-2 Modelling the ship air wake.and model-scale testing to gain an improved understanding of the key 3. Other factors limiting launch collaboration. air wake modelling activities within the TTCP collaboration. helicopter pilots need to consider the conditions that enable 2. prediction is complicated by the difficulties associated with gathering quantitative measurements 3 OPERATIONAL PROBLEMS OF AIR WAKES of the flow field around a ship. The paper demonstrates the applications significant technical challenges for the dynamic of air wake prediction in ship-aircraft qualification. the Naval Air Significant factors affecting such operations include Warfare Center (NAWC) in the USA and Australia's visibility (day. all participants authority in a crosswind. being pursued concurrently aimed at applying These are usually represented in diagrammatic form. or degraded through poor Defence Science and Technology Organisation weather conditions or spray). which may not allow the pilot to with increased pace and effectiveness through safely take-off or land. with and without securing devices such as lashings or a 2. When departing or returning to the ship. an example of which is at Figure 1. which limit . the behaviour of and recovery include exceeding torque and control helicopters operating in the vicinity of a ship is limits. It summarises the generally operates its own unique combination of work being conducted by each of the nations and ship and helicopter.IAR). Pilots must often position the participants consist of representatives from the UK's helicopter with sufficient accuracy to engage an Defence Evaluation and Research Agency (DERA). night. The aim is to present a usually determines their own. at full-scale or in a 3. The paper details the collaborative The proliferation of ship and helicopter type efforts being pursued by the member nations. conducting full. if they are not. The envelope shown on the diagram for the helicopter to high potential pay-off from shared learning has land or take-off safely. the effect of training. These all contribute requirements capture. and its effect on technical aspects to be discussed elsewhere in the helicopter behaviour. Within the group.Institute mechanism. Computational Fluid Dynamics (CFD) and other. The irregular and the value of air wake simulation in research and nature of a typical ship superstructure. which may be as small as 2 metres . aircraft-mounted probe in a ship-mounted capture the National Research Council of Canada . on deck limits to prevent toppling and sliding. The wind speed more empirical. vortical flows are discussed and model validation issues are and recirculation. A number of approaches are develop SHOLs to define safe operating parameters. and test and evaluation to pilot workload. Through this topic. putting into an operational context the more First of Class Flying Trials (FOCFTs). SHOLs may also include present and discuss the different facets of their work structural limits of both ship and aircraft. the ship may proved to be a strong motivation for the different be forced to change course and speed to achieve a nations to align activities where possible and to safe wind condition.3 Aims of the paper ship-mounted capture system. under variations means that each individual country the co-ordination of the authors. maritime helicopter operators validation database. Although the The development of simulation for application to the size of the deck area varies considerably for different helicopter-ship DI has been a collaborative activity classes of ship. analyses to achieve a predictive and direction must be within the safe operating capability and to enhance piloted simulation. 'Ship air wake effects on insufficient control authority. and (DSTO). in particular tail rotor rotor behaviour'. as well as openly with other TTCP members. There are no agreed draws together the various research aspects and international standards for SHOLs and each country validation tests. Methods of air wake prediction flow field dominated by separations.

that it can be replicated in a simulator. suitably normalised and characterised so obvious advantages in terms of safety and cost. It is the hostile environment of the dynamic interface is thought that movement of the air wake could be when he first flies to the ship and. coupled with an improved understanding of the dynamic characteristics of Before an air wake can be modelled with acceptable helicopters in this environment. it is important that the student replicates above the flight deck. The trainee pilot's first introduction to assume greater importance in higher sea states. accurately predicted. and to account possible in a simulator due to insufficient fidelity. It is difficult to collect such data for regions other than directly In training. The nature of the air wake is such that there are sharp gradients in all three components of the mean flow velocity. structures such as vortices and oscillating shear layers failure to do so will negate the advantages of emanating from the hangar corners and sides. This is not a ship operations is now believed to be possible. Validation data refer to acquired air wake data. from which the model can be derived. have reached a stage fidelity in a simulator. representative levels of workload in piloted if the air wake is too severe. in the vicinity of the also affected by a moving ship. while his initial treated as a quasi-steady phenomenon. test data must be acquired where comprehensive aircrew training in helicopter. deck landing training has not been locations not measured or calculated. Figure 1. each of which has its the full range of operating conditions which he is advantages and limitations. This trivial task. Standard engineering practice relies on obtaining . with a specific deck landings will be conducted under the watchful time-lag relative to ship motion but this has not been gaze of an instructor. Full-scale data can be likely to encounter before he ever joins a ship. from either at-sea trials or in wind tunnel tests. In a simulation. Perhaps control strategy adopted by the pilot. The air wake is both visual and environmental. 8-3 the final operational envelopes to environmental simulator training and may. An air wake model of appropriate fidelity is. which are used to provide benchmarks for the modelling. There are several approaches to opens opportunities for the student to be exposed to acquiring air wake data.1 Background For the purposes of this paper. for changes in onset wind direction. The modelling of air wakes for high-fidelity simulation continues to present a substantial challenge. A typical SHOL appropriate values for the air wake must be generated at all required locations in real-time. will influence the and potentially larger operating envelopes. which. in extreme conditions similar to those encountered during the circumstances. therefore. The flow field is further complicated by changes in the speed and direction of the approaching natural wind. improved safety performance. The air wake has an trials. prove dangerous. there is a danger that the simulation and allows aircraft control limits to be pilot will lose confidence before he has even started. Recent advances in simulation a substantive challenge. Incorporation of correct temporal and most severe conditions until he is flying spatial correlations into real-time simulation remains operationally. modelling refers to the use of numerical techniques to extract quantitative air wake characteristics and to provide numerical data for input to a simulator. Interpolation 3. The use of modelling and simulation to important part to play in deck landing training as it support SHOL development offers substantial has a considerable impact on both workload and benefits in terms of reduced costs. as well as substantial levels of turbulence caused by separated shear layers and vortical flows. a fact which will flight deck. a student may a high fidelity air wake model that generates gain a false sense of security in the simulator. If the simulated the most important component of such a simulation is air wake is more benign than in reality. in turn. with collected. technologies. an essential element of a training simulation. yet there are often flow the same control strategy in the simulator as in flight.3 Simulation in support of deck landing training schemes are required to generate values for wind at Until recently. while. Red Green 4 AIR WAKE MODELLING AND VALIDATION 4. he is unlikely to experience the verified.

000 SFS. conditions were benign for some time and the air wake was thus generated by the relative ship motion. the IAR is In the UK. Simple Frigate Shape tunnel investigations. Future efforts will be focused on have been developed for incompressible industrial unstructured grids and the Renormalisation-Group k- flows which are possibly better suited to the problem £ turbulence model. primarily focused on developing solvers for although there were limitations due to a structured conventional streamlined shapes. there will exist having examined both it and the k-e approach separations and reattachments similar to those found (without wall functions). however problems can arise from an inability to scale accurately the smaller geometric details which exist on most ships. This involves sub-scale tests which are generally insensitive to Reynolds number. at least 600.2 CFD methods for modelling and wind tunnel results is reduced. Both Australia and Canada level of confidence in the data. efforts to date indicate that CFD produces a realistic scale conditions. The geometry of this model is shown in elsewhere [2]. Such wind tunnel employ Navier-Stokes solvers. 15°. The surface nature of full-scale atmospheric characteristics is oil flow revealed a surprisingly complex flow such that data can rarely be collected for long enough topology for yawed winds and served to reinforce the times to satisfy the stationary criterion. often in the grid. This is not currently possible in approach is included here. Even for focused on real ship geometries in addition to the relatively simple ship-like shapes. One notable difficulties of obtaining a CFD solution for a realistic exception was an Australian at-sea trial where wind ship configuration. The advantages of wind tunnel tests for air wake measurements are common to most wind Figure 2. However. Unfortunately. the agreement between the CFD 4. The Euler codes will tunnel test to determine time-averaged pressure be less effective in predicting re-attachments. community over the past several decades have been Investigations have been made of grid independence. The other source for validation data is the wind tunnel.8-4 stationary data from which to generate valid coefficients at approximately 200 locations on the representations of an air wake. resemblance to an actual ship. evaluate the strengths and weaknesses of different CFD is. The approach in Canada has been to Figure 2. The boundary layer to enhance the replication of full. the work to date [3] has primarily used conducting wind tunnel tests in support of qualitative Euler solvers with the justification that the separation and quantitative data for CFD validation. a promising technique. Canadian efforts have also been expended on a Part of the TTCP collaboration has been aimed at commercial Navier-Stokes package (not the one used validating CFD codes for a Simple Frigate Shape in Australia). a brief overview of each nation's vicinity of the ship. In Canada. grid points are required. Although the model bears little use the k-e turbulence model with wall functions. which . there are codes which the k-e approach. thus several million could be required for a more realistic ship model. These efforts are often real-time for incorporation into a simulator. The plans for the upcoming on typical frigates. 45°. since CFD strategies as applied to the air wake problem. the velocities can be determined at any point in the For completeness. Surface-oil points are well defined and the flows are often flow visualisation has been completed and a wind dominated by strong vortices. result for a 0° wind. of modelling a ship air wake. and finally 90°. solving for the air wake around the Canadian Patrol Several organisations in the four countries are Frigate. at first glance. while the UK and US tests should be carried out in a simulated atmospheric use both Euler and Navier-Stokes solvers. the SFS is scheduled for September of 1998. for example control of parameters such as wind direction and the ability to Several codes have been used to calculate the air average results over a suitable time to ensure a high wake over the SFS. and 90°. but as the yaw angle increases to 45°. These efforts are discussed in detail (SFS). Turbulence modelling to date has focused on transonic regime. actively computing the flow around the SFS for wind angles of 0°. Few studies have been undertaken to Flight deck assess the level of detail required for such wind tunnel tests. One advantage of A technique which can be used to generate an air the four country collaboration is the ability to wake for use in a simulator is to use CFD methods. It must also The Australian CFD effort has involved the use of a be recognised that CFD developments in the aviation commercially-available Navier-Stokes code [1]. The geometry was chosen for year include incorporation of a helicopter rotor in ease of defining a grid and to avoid curved surfaces.

The geometric properties of the ship from NASA. In one case a NASA. Also modelled is the attenuation of the vertical component of 4. In the latter case. The time simulation. number of calculation points where the flow pattern is expected to vary rapidly with distance. although smaller ones. This assumption is dimensions of the geometric feature to which they are attached. Before being added together the intensities will be independent of wind over deck relative magnitudes of the flows caused by each basic (WOD) velocity and therefore a new table is required for each wind azimuth case where all the results are flow element are calculated according to the normalised by the WOD velocity. longitudinal and lateral axes as the al [6]. A more recent in the model allowing it to be used to investigate version has been upgraded to also include a prediction of turbulence intensity (the magnitude of some of the operational problems associated with the deck landing task. the steady flow and the flow behind the edge of a bluff body (Edge field solution was developed using an Euler solver Turbulence). under a contract aimed at establishing the potential of In order to make use of the model results in a real- a simulator to support deck landing training. which can be multiplied by an . in support of their own studies into the simulation of The UK model was originally intended to provide an estimate of the steady flow field pattern only. the time varying wind component) as a function of the position in space. It is assumed that both can be predicted by adding together the effects of the pattern of steady winds and the turbulence basic flow elements located at various parts of the ship geometry. Much of this work is directed superstructure and the low velocity regions towards solving for the air wake around Royal Navy immediately downstream of the superstructure.1 UK Advanced Flight Simulator simulation which will not let any flow penetrate the deck and The UK has implemented an air wake model in the Advanced Flight Simulator (AFS) at DERA Bedford hence at this point the intensity must be zero. This requires a non- Configurations currently exist for the Type 23 frigate uniform calculation grid which packs a higher and the Landing Platform Helicopter (LPH). In another calculated from two new modules that relate to the effort [4]. a fourth-order time. Initially the turbulence is in small eddies and these rapidly amalgamate to form larger eddies of the order In the United States. values are typically uncorrelated. Future developments include incorporation of moving geometries and Time-dependent turbulence is generated in the strong inclusion of stack gas effects. can produce sufficient size to capture all the features of the model estimates of the air wake around a ship of any shape output without requiring a prohibitive amount of for which the model has been configured. The combination of these programmes has built a level of confidence function of the position in space. Significant CFD efforts steady air wake modules for the basic flow elements. The turbulence intensities are time-averaged data were not available. several stages of refinement since the creation of the first version of the model. The relationships between the magnitude expected to hold well for moderate and strong winds of a flow and a geometry of a particular type have but may become increasingly less appropriate for cases where the wind is weak and the flow pattern been derived from empirical data and have undergone may have significantly changed. CFD efforts applied to the SFS of that of the superstructure. a look-up table is generated of model. The UK is developing a appropriate random time history that represents the time-accurate inviscid code and is working towards profile of the turbulent winds.and space-accurate flow behind bluff bodies (Profile Drag Turbulence) code was used. 8-5 do occur on the flight deck.3 Empirical methods turbulence near to a horizontal surface (such as the deck of the ship) owing to the boundary condition 4.3. described by Woodfield [7]. towards air wake modelling on United States Navy Turbulence intensities are calculated independently ships have been reported by Tai [5] and Landsberg et for the vertical. ships to aid in ship design. as a the DI under the auspices of TTCP. dissipate further downstream. A large amount of data and pilot comment has been collected from using the UK air wake model Air wakes behind bluff bodies can be described by a throughout a range of research programmes. As the shear regions have used two approaches. as well as incorporation shear layers that exist between the flow past the of the helicopter rotor. The combination of steady wind components that vary model datasets have also been supplied to spatially but are independent of time and unsteady or government representatives in Australia and the USA turbulent components that vary in space and time. and less The model is built on the underlying assumption that the main features of the flow field in the lee of a ship closely-spaced points where the flow is expected to be approximately constant. computer memory for storage. fewer new turbulent developed time-accurate Navier-Stokes code was eddies are formed and the larger eddies decay into used which produced encouraging results. and the unsteady flow field followed that contribute to these effects are already held in the using a viscous solver. inclusion of viscous effects.

borrowed from NAWC Patuxent River. Simultaneous volume measurements were made of ship angular motion. data from actual deck landings and take- offs are used as input control strategies for a high- fidelity simulator which does not incorporate an air wake model.8-6 4. Since a laser anemometer is only able to determine the wind component along the direction of the laser beam.4. Furthermore.1 Full-scale measurements the lasers over the flight deck and along the aircraft The UK has conducted a number of full-scale trials flight path. in one case full-scale air wake Sensor (SALTS) is being developed by NAWC measurements made by two nations on a virtually Lakehurst. most data can only be collapses of the air wake data when velocities are recorded directly over the ship flight deck. and anemometer. an Australian trial was direction and analysis of the size of the lobes gives undertaken aboard HMAS Darwin. heading and standard wind-over-deck direction and speed. with UK Ministry of Defence (MOD) identical ship operated by those two nations produced funding support. Various scan patterns may be utilised but the simplest is the conical scan about a vertical axis. An unexpected . suggested by the US. The objective of the trial was to measure the unsteady fluctuations in the air flow. would be assumed to be caused by the unknown air wake. linear accelerations. and there are encouraging space of time. to determine the air wake is to use inverse simulation. speed. mast with ultrasonic anemometers in 3 orthogonal directions at 3 heights was located at a series of up to Measurement 14 locations on the flight deck. an FFG-7 frigate the wind speed. it is necessary to scan the beam in order to resolve the various components of the flow field. trolley-mounted laser anemometer scanning conically at 5 heights above the deck to gather data. a method which is discussed further in Section 5. Typical results from a conical scan The difficulties associated with quantitative air wake measurements for both full-scale and model-scale A significant limitation of current full-scale measurements have been discussed. A mast-mounted occurring through the launch/recovery area. and overcome these limitations.3.4 Experimental methods for validation Figure 3. A later trial on an Auxiliary Oiler Replenishment (AOR) vessel used a single. however. ship motion and air wake over the flight deck for a variety of relative wind speeds.4. mounted on or near the flight deck (Figure 4) a picture of the flow field can be generated by scanning 4. In this approach. the Shipboard Aircraft positions are normalised by ship hangar height. The UK is currently investigating the application of inverse simulation to the prediction of SHOLs. and/or helicopter position. Plotting the results of the scan in polar form Figure 4. hence can be used to improve future ship and aircraft was used in 1994 to record air flow measurements designs. All of the measurement methods is the inability to measure air collaborating TTCP nations have made such flow vectors over a wide range of points in a short measurements. Variations from the circle indicate [8]. Differences in force summation. between actual and simulated manoeuvres. to validate CFD codes and to model air wake over the flight deck of a Type 23 frigate. Analysis of SALTS data will increase aimed at collecting sufficient data to aid in the understanding of the air wake characteristics validation of air wake models. Principle of the SALTS system (Figure 3) generates two approximately circular lobes where the orientation of the lobes gives the wind In September 1989. This approach has not yet been closely looked at and may prove difficult because of solution sensitivities. By using three co-ordinated lasers a factor-of-three difference in the turbulence levels. To normalised by ship anemometer airspeed. A single systems for simulators. 4. LIDAR (Light Detection and Ranging) Turbulence However.2 Inverse simulation Another approach.

While this used throughout the TTCP community to validate permitted easy generation of 0° wind data. 5. deficiency in the recirculation region behind the hangar. it was a models and improve understanding of the flow greater challenge to generate data for yawed winds. structure over a ship flight deck. Even though an air wake was often generated by relative ship motion in the calm conditions. wind- operational constraints during the trial meant that the tunnel measurements were made on a 1/64-scale duration of recording at each mast position was often model of an FFG-7 during the early 1990s [9]. removed by first resolving the components into wind Landing Platform Dock (LPD) and Auxiliary Oiler axes. air therefore considered limited to validation of models wake measurements were made on the HMCS determined by other means. Halifax. it was found to make little difference to the 4. superstructure configuration changes between the Though a technique was successfully developed to wind tunnel test model and the actual ship. a Canadian Patrol Frigate. All of these tests contribute to a . This work is discussed in detail in [2]. Recent tests have included the UK axes purely through vector rotation. remove ship motion effects from the measurements [8].4. The measurements were undertaken to complement the parallel work conducted in the UK and Australia. only mean flow could be appropriate duration of 5 minutes). a serious heights) was inadequate (Figure 5). and were poor in this case due to a number of (c) at the measurement grid points for the turbulence. An extensive wind tunnel experiment was conducted in the UK on a Type 23 frigate in which the testpoints For a given nominal relative wind direction.2 Wind tunnel measurements results and so was not applied. Measurements were not corrected for blockage or interference.5 representation of the full-scale measurements in the metres and 9 metres. Limited model-scale tests are also some variation about the overall wind azimuth. it was concluded that the measurements Figure 5. After normalising using free stream velocity. For inadequate (typically 90 seconds rather than a more the wind tunnel tests. the purpose for obtaining both full-scale and View From View From wind tunnel air wake measurements by the Port Side Front Australians was to provide a model for use in simulation. in the summer of The revised strategy adopted was to provide a 1991. (b) within the prior to the commissioning of the ship. (AO) vessels. The usefulness of the measurements is In support of Canadian helicopter-ship integration. Also. 8-7 development of the trials was that wind conditions database of air wake measurements which can be were benign for much of the tests. There have been some attempts to collapse the air wake data from full-scale tests and wind tunnel tests. To complement the Australian full-scale trials. Australian full scale air wake trial . 4. the mean were chosen to coincide with the data recorded on the relative wind azimuth for each mast position showed full-scale ship. This type of approach is useful in providing a framework from which to build an air wake model.4. using the ship anemometer to normalise the velocities and the hangar height to normalise distances. which conducted on each new ship entering service in an resulted in variations in velocity components attempt to identify problems in the air wake before recorded by the mobile mast anemometers in ship the ship is built. obtained did not allow an adequate representation of mobile mast layout and deck positions the air wake covering the region surrounding the flight deck. This effect was Royal Navy's Landing Platform Helicopter (LPH). • Mast Positions Canadian wind tunnel tests were undertaken to provide wind tunnel air wake data for a configuration matching the as-built CPF.3 Australian full-scale data representation Initially. Encouraging results were found for two distinct frigates for several wind angles. Correlations measurement space domain for the mean flow. the number measured and the measurement probe was limited to of grid measurement locations (13 positions at three local flow deviations of less than about 40°. such as CFD techniques. These measurements were form of a model valid only (a) for the nominal made to verify earlier wind tunnel tests conducted relative wind directions recorded. Because of the deficiencies outlined above. Measurements were made at fifteen locations separate mean disturbance flow and turbulence on the flight deck at two heights above the deck.

the weight of the aircraft air wake will be modified primarily by the downwash and the power applied will also have an influence on from the main rotor. for a finite surface. of a helicopter operating over a ship flight deck. allowing the best' data set (in terms handling of the aircraft. established. The methodology adopted to represent turbulence was similar to that used at the Georgia Institute of Technology [10]. Initial attempts at real-time aircraft will also influence the flow structure. 4. and a similar inverse power model. model over the flight deck. To represent the spectral density more speed as the area of increased pressure below the faithfully. ground effect. 4. y helicopter increases with a consequent reduction in (lateral). when hovering near the ground. and z (height) directions. the power spectral flight deck will also influence ground effect. In simulation of these effects may be based on the . polynomials are used velocity is reduced and the pressure below the to curve fit the data. 135°. (noting some significant deviations) were treated as these factors will not affect the performance or duplicate cases. With the helicopter in the flow field. In the x (longitudinal). effect can be expected with recirculation. represent the turbulence only at each grid point as an the ground effect can be expected to be reduced.different wind speeds at the same nominal azimuth clear air. and two respectively for each wind direction of 0°. The area of the helicopter rotor in comparison to the size of the Using the coefficients obtained. such as a ship flight deck. a factor known as respectively three. the The angle of the rotor disc. 30°. efficient way to archive the data for future use in Ground effect also diminishes with increasing wind validation. Ship motion will constantly Shinozuka's algorithm [11]. and w of degree two. and 180°. five (but only three at aft of ship). but the tail rotor and body of the the flow structure. an additional coefficient was used in the helicopter is effectively blown away. the purpose of the modelling is to predict Modelling recirculation and ground effect presents a the behaviour of a helicopter operating in the air significant challenge. as shown in Figure 6. there were the power required to hover. This is done for the three velocity components at each location (i. Instead. In the Georgia Tech model. Hence. who attempted to obtain a model using the Australian test data.6 Modelling the helicopter-in-the-loop However. and possibly ship motion. 3x13x2 Ground effect and recirculation are both results of the coefficients) for each wind direction case at a rotor downwash influencing the freestream air flow. away from any surfaces or obstructions. particular free stream wind velocity. Recirculation effects variations with frequency. v. the rotor downwash For the mean disturbance flow. the local wind speed and different wind speeds at the same nominal azimuth direction. 60°. especially for real-time wake and so it is important that the effect on the air simulation. In the presence of a of measurement scope and standard deviation) to be horizontal or vertical surface. Ground effect is led to an abandonment of the prospect of a suitable normally considered in terms of an infinite ground model. the flow is chosen to represent measurements at that nominal constrained and may recirculate through the main azimuth. This partial 4th degree polynomial representation had 27 terms. Furthermore. four. alter the relative angle between the helicopter and the flight deck. as well as the effective area of the deck. The whole flow field will change for wake of the presence of the helicopter is also every position of the helicopter relative to the ship. The degree to which each effect is present at the DI The Australian approach differed in that difficulties depends on a number of factors including the size of in obtaining consistency between coefficients for the ship and helicopter. however. and three anemometer locations. power spectral density is represented by an inverse power model with two constant coefficients obtained by fitting the model to auto-spectral power Figure 6. To obtain a rather than the freestream influencing the helicopter.5 Effect of helicopter-in-the-loop and may influence both recirculation and ground The models described so far have attempted to effect. a spline fit was applied to and may be expected to contribute to the performance each coefficient.e. predict the influence of the ship on the air flow as it passes over the superstructure and flight deck. 90°. and the density can be calculated as a function of frequency dimensions of the hangar face will impact on the and a simulated time history can be created using strength of recirculation. it was considered appropriate to use polynomials for representing u. the approach adopted was to plane. rotor.

However. The air wake in this case was With the numerical rotor validated against the wind obtained from cross-hot-film anemometer tunnel rotor in uniform flow. were taken around a helicopter hovering that is traversed into the air wake. It is proposed that the effects increased the level of the unsteady rotor effect of the modified air wake on the aircraft could response. It technique has been used to represent the effect of the was reported that inclusion of spatial correlation helicopter on the air wake. Both scaling of the experiment is 1:50 and the rotor is sets of data were recorded in nil wind conditions. incorporated. will ensure the flow field clearest indication of the influence of the helicopter coupling is correct. a simple actuator disc space. but off-line cross-spectra or orthogonal component cross-spectra analysis is being conducted in the UK. and GenHel was then type of model would reside with the ship. terms of accurate air wake modelling. correct scaling of the same conditions and initial comparisons were the rotor aerodynamics. Measurements of rotor loads in the wind tunnel Simultaneous measurements from two anemometers experiment have been made for several onset wind permitted accurate measurements of the turbulence speeds and directions as well as with changes in ship correlation between two physical points in the air superstructure. while unparalleled in Canada through the use of CFD codes. the pilots who consistently remarked on the difficulties capability for cyclic inputs is not possible at this of holding position behind the hangar in comparison scale. The spatial the helicopter-ship system will require integration of correlation in the generated turbulence could the rotor code with a real-time simulated air wake. This flow field was developed. but can be to the hover in clear air. It has been determined that for a wake region. US and was attempted. 8-9 empirical approach of the UK air wake model. It rotor disc from a CFD calculation and feeding it to a was also found that the inclusion of the unsteady helicopter performance code. As a precursor to this research. with a seems likely that a higher fidelity representation of thrust-to-weight ratio of unity. but it "flown" through this numerical air wake. model of an air wake. the rotor module of the Sikorsky of the GenHel rotor to match the characteristics of the General Helicopter Flight Dynamics Simulation code wind tunnel rotor and then this numerical rotor will GenHel [14 & 15] was "flown" through a numerical be "flown" through the measured CPF air wake. This approach.6. were then analysed. The rotor has the capability for and solid boundaries on the flow field is given by the arbitrary collective settings (up to stall).2 Canadian research this experimental modelling. The geometric first behind a hangar and second in clear air. A modified loading vertical velocity into the air wake model induced over the actuator disc would then be applied as an greater unsteady rotor loads than inclusion of an input to the CFD calculation and an iterative process unsteady streamwise velocity.6. Thus the rotor is not trimmed. specifically advance ratio very encouraging [12]. used to produce a representation of the complete flow field. A "flying" hot-film probe was also fixed collective setting in a 35 knot wind. the IAR is helicopter influences the air wake over a flight deck. conducting sub-scale wind tunnel experiments Air flow measurements.9 metre wind tunnel.1 UK research wake. Measurements were made of the three orthogonal mean and unsteady flow components. for a rotor positioned 6 metres above the deck (full- scale) and over the centre of the flight deck as . Ground effect is also inherently included in 4. however. The calculated rotor loads at a given point in In the UK research [3]. the time- used in these tests to permit accurate measurements averaged rotor thrust coefficient decreases about 10% of velocities in the recirculation zone. Canadian work incorporating a numerical helicopter into an air wake was described by Syms and Zan These experiments will be followed by modification [13]. as output from GenHel. is not able to model the coupling of the rotor downwash and air 4. No modelling of the lateral velocity expected in the foreseeable future. powered by a small DC motor mounted on a six- The results were compared with a CFD analysis of component balance. In this case. changes in rotor loads measurements made behind a model of a frigate in determined from the experiment and the numerical the IAR 0. incorporate cross-spectra of either the streamwise or Real-time calculations of this type cannot be vertical velocity. the UK has In order to obtain quantitative information about the recently conducted a land-based trial in an attempt to degree of coupling between the rotor downwash and determine the degree to which the presence of the the air wake as it affects the rotor. perhaps the and thrust coefficient. wherein an model will indicate the importance of the downwash atmospheric boundary layer simulation was and air wake coupling. using a single laser involving a frigate model and a sting-mounted rotor anemometer. so that An algorithm to regenerate the unsteady a look-up table is used to approximate recirculation characteristics (spatial and temporal) of the measured and ground effect at any point around the ship. operated at any fixed collective position for a given traverse. This is to be expected since the turbulence be calculated by extracting the velocity field over the scales are of the same size as the rotor diameter. In this work.

it is attempt was then made to use this relationship to recognised that the tail rotor cannot be ignored. SHOL boundaries for beam winds are generally However. heavy seas.7 Effect of the tail rotor linear relationship between the torque ratio and While the discussion so far has focused on the relative wind speed was obtained for each region.1 Synchronised ship and helicopter measurements SHOLs. anemometers. though corrective actions that a pilot must take during the synchro inputs were used for data recorded from landing. and another was not hold a fixed collective during this manoeuvre. alongside the ship to the low-hover. the Australians obtained .6. however. Hence. the boundaries did not match well with determined by limitations in tail rotor authority. and the helicopter same control strategy were implemented for a was instrumented to record flight control positions conventional ground-based landing. Incorporated into the code was an actuator disc inspection of Australian FOCFT measurements model to represent a 5000 kg helicopter with a 16.2 Torque-limited SHOLs house developed unsteady-flow solver was used to In the absence of shielding of the air flow by the compute the unsteady air wake and the ship stack hangar. decrease collective normal ship instruments and some aircraft to execute the landing manoeuvre but these data instruments. and this echoes the plotted against relative wind speed for each 20 degree experimental (wind tunnel) work reported by Syms range of relative wind directions. which would represent the fully- coupled case. the information is incomplete in The primary focus of the work was the modelling of attempting to duplicate the trials in a simulation using exhaust gas concentrations in the vicinity of the flight recorded pilot controls to drive the simulation. or simply by the effort redundancy. However. An interactions of the main rotor and air wake. primarily because of Thus high-fidelity simulators used for training sensitivities of the boundaries to small changes in purposes or SHOL prediction must include suitably torque. it was assumed that the physical characteristics of the flow were reasonably constant.5 showed that torque remained relatively constant as metre diameter rotor. While a pilot will Australians was fitted to the ship. These effects and flight dynamic parameters. This approach will result in a relative wind speed increased for headings where the modified air wake due to the actuator disc. mostly analogue inputs. deck as opposed to investigations of rotor aerodynamics and helicopter flight dynamics. accurate modelling of the tail rotor aerodynamics. This the modified air wake is not fed back into the indicated that the flow was being effectively blocked. increased airspeed up to about 80 knots. The ratio between in the unsteady air wake are in a bandwidth likely to the measured and predicted torque was obtained and affect helicopter flight dynamics. which was utilised when instrument required to track the vertical and pitch deck motion in failures occurred. an S-70B-2 Seahawk helicopter operating from an These positions represent the transition from FFG-7 frigate during a FOCFT in March 1994 [16]. of course. andZan[13]. The work included modelling of the The blocking effect could be quantified in an off-line atmospheric boundary layer to enhance accuracy of simulation by using GenHel without any allowance the boundary conditions. more substantial areas limited As a first step in using analysis and simulation in by pilot workload can also be investigated. predict torque-limited boundaries of the SHOLs. observations from the FOCFTs. 5. into the helicopter. Additionally. The ship was instrumented to record suggest the rate of descent will be greater than if the ship motion and wind velocity. torque would be expected to decrease with exhaust gas trajectories for a DDG51 destroyer. though mean relative wind was unsteady components of the air wake remains to be measured using the mast-mounted ship and reference determined. ship hangar significantly affected the air flow.3 Piloted simulation 5 PREDICTION OF SHOLS Piloted simulation offers a more promising method for recreating torque and pedal limited areas of the 5. the fitted to the aircraft. Whether the wind tunnel rotor can be used to extract dynamic loads arising from the During these trials. the air flow distribution over the flight deck region and around the helicopter could not be 4. Within this range. The main determining SHOLs. a good 4. actuator disc model. as practised by A data acquisition system developed by the the British and Canadian navies. for relative wind headings where One US approach to modelling the rotor and air wake the ship hangar will significantly affect the air flow interaction has been reported by Landsberg et al [6]. In general. Duplication of some may be overshadowed by the unsteady rotor loads measurements from alternative sources allowed some arising from turbulence. The CFD air wake for ship air wake to predict torque for a trimmed prediction indicated that the predominant frequencies aircraft behind the ship hangar. An in.3 US research measured.-10 compared to a rotor at 10 metres above the flight synchronised ship and helicopter measurements for deck and displaced laterally by one rotor diameter. Each system allowed up to 56 results suggest the magnitude of quantitative channels to be used. The pilot will. 5.

8-11 obstacle to achieving a suitable model of the whole simulation of the air wake remains. The US has proposed landing operations. The authors gratefully acknowledge for all stages of the deck landing task. that in Australia and the USA are working collaboratively to this air wake model. potential benefits to maritime helicopter operational availability through improved ship design. The UK is currently FOCFT with simulation. at both model installed in the AFS. The second phase utilises prediction of operational limits using a variety of Wavelet Analysis to decompose the time histories of techniques. High fidelity . by the UK model. the pilot will be unable to in an air wake model currently installed in the AFS consistently achieve safe and accurate landings and and successfully used in trials aimed at supporting the SHOL boundary is reached. including piloted simulation and pilot pilot control activity into discrete events which are workload modelling. AO Auxiliary Oiler Simulation allows any wind condition and ship AOR Auxiliary Oiler Replenishment motion to be investigated and enables exploration of CFD Computational Fluid Dynamics areas of the flight envelope which may be dangerous DC Direct Current in flight. no account is taken of the the aim of improving simulation of the DI. however. though. data derived from piloted simulation. tool to estimate the control actions required to fly a Each nation has been collecting air wake validation helicopter model along a prescribed flight path that data through ship trials and wind tunnel experiments describes the deck landing task. When the workload is too high due more empirical approach is being pursued by the UK to the air wake effects. Each change in the air wake as a result of ship motion. several organisations in the UK. have demonstrated the feasibility of this method [17] and Australia will shortly attempt to reproduce The paper has described how. The project is divided into two inverse simulation as a means of deriving air wake phases. By modelling the air wake. The project the substantial contributions of their colleagues will culminate with the production of predicted without whom this work could not have progressed SHOLs which will be compared against equivalent as reported here.4 Pilot workload modelling directions and comparing results against each other A critical factor in the development of a SHOL is the and against wind tunnel data generated by Canada. It is anticipated that the then analysed to indicate the level of workload synergy of effort achieved under TTCP will make required. The simulation and distributing the results to other members of the incorporates a model of a human pilot with group. A piloted 7 ACKNOWLEDGEMENTS simulation trial is planned on the AFS at DERA The research described in this paper results from the Bedford to collect data for the assessment of the pilot culmination of effort of perhaps forty individuals in model and to identify suitable modelling parameters the four nations. forming a validation database. The first phase uses a constrained simulation data from aircraft position and pilot control activity. The Canadian wind tunnel work is a collaboration 6 CONCLUSIONS between the Institute for Aerospace Research and the This paper has shown how the modelling and Air Vehicle Research Section of the Department of simulation of ship air wakes offers substantial National Defence. parameters to allow the pilot strategy to be varied. Canada. The most significant challenge due to the complex and highly promising approach in the short term is to incorporate dynamic nature of the flow over the flight deck and the UK look-up tables produced using the air wake the difficulties of obtaining validation data. as with look-up tables produced facilitate the development of air wake models with by current CFD codes. the 8 GLOSSARY development of new technologies and qualification AFS Advanced Flight Simulator (DERA) techniques and more effective pilot training. considerable savings in the time and cost required to develop a high fidelity DI simulation and will The aim of the research is to enable the investigation significantly enhance the maritime operational of the mechanisms by which a helicopter performing capability of participating nations. It should be noted. a system is in representing the air wake. Initial trials at DERA full. A pilot workload. nation is using CFD techniques to calculate the air wake around a common SFS for various wind 5. The model will be developing a modelling capability to assess and evaluated independently by Australia and the US in quantify the effects of the air wake on helicopter deck similar trials in the near future. helicopter-ship DERA Defence Evaluation and Research Agency (UK) interface issues can be explored and SHOLs DI Dynamic Interface estimated before a ship is built. under the auspices of FOCFT test data using the look-up tables generated TTCP. This prevents the prediction of unfeasibly large or The air wake models have already been applied to the rapid control inputs.and model-scale. a deck landing is influenced by the presence of an unsteady air wake and by ship motion.

. 1995. 5 to 8 October 1998. Simulation'. TTCP The Technical Co-operation Programme 12 Manning. & Allen. No. L. 25. N.F. Spain. P. & Schräge. M. J. The paper is not subject to N. CM.N. September Paper 5.M.J.E.R. NASA Contractor Report 166309. Blackwell. Washington. RTO/AVT Symposium. 15 Howlett. Australia. G. Paper 6. 1. I. G. Aircraft Ship FOCFT First of Class Flying Trial Operations.8-12 DSTO Defence Science and Technology Modelling the Helicopter/Ship Dynamic Organisation (Australia) Interface'. Amsterdam. N. Sandberg. May 1995.Mathematical 2 Zan. pp. Moffett Field. Paper 8. W. & Padfield. S. 'Simulation and Analysis of LHD Ship April 1998. M. Britannic Majesty's Stationary Office. pp. the Challenges?' AGARD CP 577. D. Department of Defence. Elements of this paper are declared the work of the 8 Arney. of Patrol Frigate Air Wakes'.M. & Zan. Australia. Albone.P. 655-658. Analysis of the Non-linear Coupling Effects of a Helicopter Downwash © British Crown Copyright 1998/DERA. Vol. LPD Landing Platform Dock University of Tasmania. T. Paper 31. Air-Sea Interface'. MOD Ministry of Defence (UK) 10 Mavris. LPH Landing Platform Helicopter December 1992. 'UH-60A Black Hawk Engineering October 1997. J.F.. DSTO Technical Society Annual Forum. & Jan. Jr. 'A Review of Australian Activity on Copyright protection in the US.B.R. HMAS Her Majesty's Australian Ship 9 Matheson.Background Allen. & Young.. Amsterdam. Turner.. AGARD CP 552.P. NASA Contractor Report 166310. RTO/AVT Symposium. 'The Dynamics Problems of Vehicles Operating near Use of Modelling and Simulation in Support of or in the Air-Sea Interface'. 7 Woodfield. Reddy. 5 Tai. Organisation. Marseilles. K. 1998. J.R. Prasad. US Government. 'Ship Airwakes .M. 5 to 8 October 1998. C.T. Air Wake by Navier-Stokes Method'. August 1994.FFG-7 First of Class Interface Problem'. USA. Simulation Program: Volume I . Problems of Vehicles Operating near or in the Ames Research Center. Prediction of Ship Air Wakes over Report'. G. Soliman. & Simulation Program: Volume II . 'UH-60A Black Hawk Engineering 3 Tattersall. CM.. Rotorcraft Forum..C. 1998. in 'Flight Simulation . in Fluid Dynamics National Aeronautics and Space Administration.P. T. SHOL Ship Helicopter Operating Limits 1972. SALTS Shipboard Aircraft LID AR Turbulence 11 Shinozuka.. 25th Annual International NAWC Naval Air Warfare Center (USA) Symposium of the Society of Flight Test NRC National Research Council (Canada) Engineers. AIAA-95-0047. Analysis of Rotor Forces CFD Studies of Ship Airwake. A. 1 Jones. D. 54th American Helicopter Flight Trial Measurements'. 29 September to 2 14 Howlett. 1991. S.J.M. J. CTAC 97.. Unpublished DERA report. & Gilbert. in Fluid Dynamics National Aeronautics and Space Administration. J. 4 Liu.W. A. D. August Techniques and Applications'. Syms. A.P. 16 Arney. December 1981..J. with the permission of the Controller of Her 1995. 11th LID AR Light Detection and Ranging Australasian Fluid Mechanics Conference. SFS Simple Frigate Shape Journal of Sound and Vibration. & Renter. B. A. D.A. & Toffoletto. & Tomlinson. 5 to 8 October 1998. G. 'Digital Simulation Sensor for Random Processes and its Application'. & Cheney. Amsterdam.. 13 Syms. Flight Decks using CFD'. Analysis Model'.J. December 1981. 111-128. Problems of Vehicles Operating near or in the Ames Research Center. K. Higher Order Accurate Australian Airborne Trials of the Sikorsky S- Ship Airwake Predictions for the Helicopter/Ship 70B-2 Helicopter: Part III . Seville. R. California. 'A NASA National Aeronautics and Space Simulation Methodology for Modeling Ship- Administration (USA) Airwake Turbulence'. California.J..M. France. . Fieldhouse. 24th European Symposium.P. DC.. & Long. University of Adelaide. Boris. Hobart. Erm. May Report 0660. Air-Sea Interface'. Defence Science Technology 1998.R. Computational in a Ship Airwake'.V. USA. 'Comparison of WOD Wind Over Deck Air Flow Measurements Around the Flight Deck of an AOR and Type 23 Frigate with CFD and 9 REFERENCES Wind Tunnel Results'. Moffett Field. L. in Fluid 17 Fitzjohn. B. J.N. Washington DC.A New Generic Model for Piloted © 1998 National Research Council of Canada.Published with an Unsteady Airwake'.Where are © Commonwealth of Australia 1998/DSTO.D. AGARD CP 509. 6 Landsberg. A. RTO/AVT First of Class Flying Trials'. 'Measurements of the Airflow HMCS Her Majesty's Canadian Ship about a 1/64 Scale FFG-7 Frigate Model and IAR Institute for Aerospace Research Correlation with Full Scale Results'.

its decision was laid on Central main defense task was to restrain an aggressor from Aerohydrodynamlc Institute (TsAGI). the ship on role and the place of navy. bluff body with separated zones and vortex flow. models became basic method of investigation. on sharp bends of form. Moscow 107005 17. TsAGI's works for provision of aerody. The results did not give For a long time Soviet Navy task forces have not in open possibility to receive enough complete insight into the ocean the air cover. and principally . The works were directed on develop. of their functions. With account of specifics of this continental power. Radio Street. jects 1143. that prolonged experiment in wind tunnel. time the soviet leadership at forming of programs of ization of measures of reduction of these disturbances military shipbuilding gave absolute priority to ASBR. Since in this case the inertia- ets (ASBR) radically changed the views of soviet leader. This consideration allowed to go along the way of tests The boundaries of anti-submarine defense have moved of models of aircraft-carriers in wind tunnels. type effects prevail over the effects of viscosity. war actions on alien territory.3%. which was not supposed conduct of are described. the ship. Russia SUMMARY based naval missile-carrier aviation and submarines. ing of combat stability. ed not only increase of autonomity and seaworthiness Besides. domestic navy ships with aircraft armament. Therefore on many centuries its problem.A. positional areas of its deployment from anti-submarine tion and to estimate the effects of motion ship kinemat.2 m. The wind tunnel with open working part and nozzle each equipped by 14 anti-submarine helicopters. The tion to aircraft-carriers of enemy was laid on coast. INTRODUCTION Whereat one of principal tasks was found to be the pro- vision of aerodynamic compatibility of ship-based air- According to its geographical position Russia is the craft and carrier-ship. of pressure distribution on topside of HACC's hull for the project 1160 was made. For this aim construc- ic parameters relative to wind and of some features of tion of heavy aircraft-carrier cruisers (HACC) of the pro- ship architecture on the flow structure. flow picture around ship. These functions was laid on anti. 9-1 The experience of aerodynamic disturbances research behind an aircraft-carrier ship with elements for safe operation of ship-based aircraft L. The development of marine strategic forces demanded Numerous model tests of different ship architecture with time the decisions of the tasks of protection of the variants allowed to study the mechanism of flow forma. operation safety. Reynolds number. Its aim was design and real. This circumstance and ment of methodology of estimation of wind flow distur. that An opinion about some reasons of relatively late World War II showed leading role of aircraft-carriers appearance of the first aircraft carrier in the Soviet Navy among other classes of combat surface ships. giving for on the thousands kilometers from coast. ing. also economic difficulties of parallel creation of aircraft bances over flight deck and behind carrier by model carriers and ASBR have lead to the fact. armed with winged missiles. was deployed. equipped with VTOL air- craft Jak-38. put diameter 2. fixed tial adversary of atomic submarines with ballistic rock.Zharinov Gos NITs TsAGI. this gave the possibility of realization of of ships. and its gradients up to level required for takeoff landing capable to decide the strategic tasks. means of antagonistically side. five-tube air-pressure probes mounted on spatial coor- Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". The lack of oversee territo- ries and restricted emergence to Baltic and to Black METHODIC AND RESULTS OF INVESTIGATIONS seas set before Russian naval fleet generally the tactic tasks of support of maritime flanks of overland armies As the first step there was the choice of methodology of on insignificant small distance from its shores.O. In 1971 here the experiment on study Navy. having low level of turbulence. what demand. and published in RTO MP-15.Maslov N. The Netherlands. Anti-sub- marine cruisers of project 1123. .on resulting flow picture practically does not depend on its enlisted stuff of ships. was used.Valuev A. on scale effect. full-scale experiment the role of controlling instrument. ming aerodromes» had not found their place in USSR namic compatibility of ship-based aircraft and carrier defense doctrine. The particular features of Some recommendations were developed and created these ships applied architecture were a consequence on aircraft-carrier cruiser «Minsk» for deck flow level. 1143M and 11434. In spite of the fact. Such experiment on receiving of physical picture of flow over conception persisted up to 60th years of XX century. The range of relative wind angles was determined for safe landing on carrier «Admiral Kuznetsov». Freeboard and super structures of aircraft- carrier from point of view of aerodynamics are typical Appearance in the stuff of the naval forces of the poten. type of « Moskva». which Is very necessary for ensur. and counterac. west and south directions. 5-8 October 1998. and. The measurement of velocity fields around carrier-ship aircraft missile armament of the ships. An absence of experience of creation of such objects set before industry the quite a number of problems. equal to beginning of the application of deck aircraft in Soviet 0.V. held in Amsterdam. but also creation of principally new for the advanced investigations of perspective configurations. consequently. «swim- is expressed.

of ship is essential factor. has shown. obtains significant of flow were main works. Its heights are in the range 1 -2 m. lowing. 1 and 2 ture. mining boundaries of zones of separated flow above deck and behind island qualitatively. i. Opposite effect is observed at heel on leeward . ship are possible at presence of battle or operational damages. having less cost. They do not exert considerable influence on resulting cal flow washes en=f (y. Vy. around heeled or trimmed models. Hot-wire anemometers for measurement of fluctuating velocity Extensive dead fluid zone behind the counter flat stern components also were included in the equipment. and velocity field in wake behind Special series of methodic experiments in wind tunnel ship is formed under action of dead zone behind stern. pitch configuration with displaced on the starboard side and roll angles.As longstanding practice has shown. The vortex system undergoes some changes quantita- riers' models. forms of which correspond to classic tively in dependence on absolute value of angle (. The vortex sheet is curtained into large-scale vor. Vortex of analogous nature is frequency oscillations of scale model. was suffi. is shown on Figs. but it Is necessary to take into £v= f(x. on comparatively small angle 9=3°( is accompanied with ble with intensity of vortex. was eliminated in the case. formed on edge of deck. Flow is closed behind the symmet- vector of freestream velocity V<». process in wind tunnel with the aim of estimation of flow tex core. and with static angles of heel and trim. is eroded gradually or joins with more inten. behind its bow edges. angles on characteristics of stationary flow fields tex sheet. that tests of double mirror model for study diverting the resulting flow downwards (Fig. of disturbed air flow in the wake behind ship is more correct to use for wind tunnel instead of model on At ß>0. at regimes. structure above flight decks and in wake behind aircraft carrier of around aircraft-carrier and to determine quantitative the considered configuration. but qualitatively it remains practically unchanged.z) and Vyz=f(y. edges. Output data of information-measuring system of wind It is necessary to note that fixed flow separation zones tunnel were the components of local disturbed veloci. Described above picture is principal structure of flow sis of features of flow in wake behind ship. The considered flow picture takes place at ß<0. The oscillations are close to which angled deck is located. which are Model experiment showed that heel on windward side different from zero. This vortex is found vibration. behind aircraft carrier only one vortex is observed. which are regarded as basic for realization of Measurements of parameters of perturbed airflow were landing operations on carrier with angled deck sloped carried out in transversal cross-sections above deck on the port side. received on base of tests of series of aircraft-car. The measurements were com. angle of relative wind ß was counted direction of rotation are formed. from edges of deck and island. bow vortex conditions. Bow vortex. such account them at analysis of aircraft takeoff conditions. coming off from edge of sponson.7L (where L. what does not correspond to surveyed phe- downstream up to distances behind the stern equal to nomenon. and in aerodynamic wake behind model. Velocity field behind between longitudinal centerline plane of model and island is vortexless. Intensity of this vortex is compara. motionless screen and does not correspond to real However. Certain generalized picture of flowed as thick wing of extremely small aspect ratio. which propa. closing behind the stern. noticeable reduction of intensities of vortexes (approxi- Namely it exerts a crucial effect on parameters of air. as to estimation of influence of limiting static roll and pitch consequence of loss of its connection with feeding vor. and also on features of ship architec- island and angled flight deck. named further as «bow vortex». screen. 9-2 dinating device became main instruments. Third component of vortex system of aircraft carrier is the vortex.z). angles of vertical and horizontal washes. deck. Both hull and island are characteristics of flow.length of ship's hull). what leads to plemented with visualization of flow around models in curving downward of trajectories of vortexes. Vortex generations at ß=0 occurs only on edges of fore- Cartesian system of coordinates was used at the mea. rical structure. tecture. vector fields Vxz=f(x. set of information permits to carry out exhaustive analy. coming wind tunnel and towing tank. The reason of this is in high- gates over flight deck. Such positions of sive vortex. flow separation occurs from windward edge of fore.e. flow. These tests allowed deter. Due to aerodynamic interference in wake cient for flow measurement above deck. where two symmetrical vortexes with opposite surements. For example intensities of vortexes increase practically according to linear law at an increasing of absolute val- Vortex system of flow around aircraft-carrier is the fol. en=f (x. when wind flow incomes on the starboard side. as an example. Vz. are formed above surfaces of decks immediately ty Vx. In front of carrier the undisturbed horizontal wind flow acquires vertical upwash due to which the Aircraft carrier in real conditions is subjected to rolling. mounted both on even keel.y). and also distributions of angles of horizontal and verti. required accord- formed on windward side edge of ship's sponson. Influence of boundary layer.5-0.z) and sv=f (y. disturbances is impossible. Therefore the investigations were restricted 0. encountering island on its way. ues ( up to investigated range -25°<ß<25°).y) and velocity field in wake. generated by island at angles (. However correct modeling of this deck. coming from corresponding deck flow in zone of landing glidepath. Bow vortexes due to their small intensity are quickly destroyed. mately on 50-85%). which is formed on flow picture in bow part is similar to the picture at ß<0.z). determining flow in wake. on ing to kinematics criteria. pitching and heaving. Mounting model interacts with flow around this element of ship's archi- on horizontal flat screen. The However the investigations of quasi-stationary picture flow. 2b).

with rolling-off. architecture. However since a use of radical means of ones. Displacement of island to starboard from longitudinal The developed methodology allows easily to estimate plane and closer to the bow favorably effects on condi. Moreover vertical front wall of large spon. To decide this problem. Besides.5°. 3).7) from point of view of safety problem of takeoff/landing operations on the design stage.5° (trim by the stern) increases in 1. presented certain hazard for conditions to required ones for safety. There were made attempts of search of such shape. Kuznetsov» (Fig.5-2 accepted composition of LD included three flat and one times and correspondingly decreases at i|)=-1. zone. that intensive vortex tem of wake becomes difficult due to an increase of comes from windward side edge of risen foredeck. Flow in this zone is characterized with high deceleration of incoming flow Such range was found in process of model tests. on Figs. at which vortex formation behind it good convergence of the results with model experiment was minimal. covering up to 20% of runway. removed from glide path. in this case is impossible to create on equipped with developed LD complex. the determination of require- meters. The necessity of additional control- received in wind tunnel has shown.65V»). So.4). The vortexes at \p=1. Whole complex LD reduced lateral. ing it from flight deck. Therefore TsAGI rec- As a result the such variant. type of «Minsk» effects on flow above deck like trim by the stern. of which were confirmed subsequently by measure- ments on ship. intensity of edge and tactically acceptable from another side. just like for Jak-38. angles (. directed on air flow leveling. as it is shown in wake due to interaction. These phenomena. All ships of this type were deflected flaps. of VTOL aircraft Jak-38 from «Minsk». because of it Is observed more strong fall of The flat plates turned the vortex flow into contoured path of vortexes. For example the rise of foredeck bow 29K was as a point of departure. operational pilot's. 8). generated by island. mushing and in isolated off runway of given aircraft must be taken into account. In com- over its angled flight deck. although These data became basic criteria of conformity takeoff are not catastrophically. bow starting ramp and forward part of angled deck can lead to pumpage of compressor. Previous experience of satisfaction of demands to flow parameters along Jak-38 takeoff path needed in its Just for this cruiser the first TsAGI experience was evolution. recommendations are reduced to require. As test pilots noted. requirements from one side and was technologically Influence of trim is more essential. However the change acceptable level. ic conditions. the external aerodynamic condition on the first in USSR tions of landing. practically does not zone over deck and decreased deceleration of flow to depend on angles of heel and trim. panied with vibration of aircraft. propagating in aerodynamic wake passage between blast fence and curved plate. as example spoilers and accepted decisions. The flight tests confirmed correctness of the aerodynamic improvement. The fairing with symmetrical semi-elliptical profile practically removed separation Vortex. there is a noticeable effect of some features of ments to parameters of air flow on trajectories of take- ship's architecture on quantitative characteristics of off and landing for ship-based fighters Su-33 and MIG- flow around it. since in these cases vortex. was selected which ommended to set profiled semi-elliptical fairing on bow reduced parameters of flow to correspondence with the and side edges of ships ramp and forward part of spon- . and there are no practice. besides of kinematics para. in process of which turbed air flow with help of hot-wire anemometers dozens of so-called leveling devices (LD) variants were showed. How already it was noted. that it is impossible ling actions increased high psychophysical load on the to satisfy these criteria without modification of ships pilot (Fig. The scale of the phenomenon. part of HACC of the project 1143. inside of which flow parameters would be con- son leads to appearance of well-developed separated formed to the requirements. Therefore one of the way of vortex deflects by containers. This involve tightening of requirements to air flow para- meters too. located on its way just on increase of landing safety was search of range of flight deck. Here. generated aircraft-carrier of classic type named now «Admiral by it. any takeoff accidents in process of long-term flight ment of decrease of aspect ratio and relative width of operation of aircraft Jak-38 due to external aerodynam- island. as well as fairing on front wall of sponson dead fluid zone after stern grows exactly likewise at trim by the bow. deviat- behind ship (Fig. Study of data. That was in 1981 during the parison with ground-based aircraft the requirements to course of development of horizontal short-run takeoff accuracy of nominal glidepath keeping are increased. 9-3 side (Fig. simulation of land- At the analysis of safety of above-mentioned takeoff the ing outside the recommended of angle (range accom- special requirements to parameters of air flow on take. TsAGI jointly with industry have carried Measurements of non-stationary component of dis- out extensive research work. There are high velocity of approach and lim- obtained in purposeful action on flow characteristic ited deck section of touching by arrester hook. that frequencies and levels of fluctuations over tested. purposeful action on vortex sys- Aerodynamic tests have shown. 5 and 6. Full-scale measurements on HACC «Minsk» showed formations of island. of characteristics of another elements of vortex system vertical and longitudinal disturbed velocities and its gra- leads to some change of resulting flow intensity picture dients over flight deck to required values. cases . The curved plates. results (up to 0.

carrier became important result of whole complex of the works. in particular. itary and civil ships and marine drilling platform.■ / / / //*"" c) / / / ///-"' I / I I //*" \ \ V \ I I i ///•" / Fig. This permitted to eliminate formation of extensive Almost decennial experience of trouble-free drill opera- separated zones. Practical experience was refinement of external conditions of basing of marine obtained in development of measures for ensuring of aircraft was used also at decision of the same ques- flow parameters required for safe operation aircraft tions. - . connected with operation of helicopters from mil- from concrete ships of given class.9-4 son. spective aircraft carriers with group and single basing turbed air flow over deck and in wake behind aircraft of aircraft. CONCLUSION Generalization of accumulated materials formed base for development of requirements to architecture of per- The developed methodology of investigations of dis.2 Velocity fields in the wake behind carrier \ \ I I t i / / ' \ \ model at different angles of relative wind \ \ \ \ a)ß=-7° b)ß=0 —<V\V c)ß=7° £•■£-'. model experiments. developed on base of degree of non-uniformity of flow. software and model Developed in TsAGI methodology of estimation and technology were created. correctness of operation of its powerplant in consequence of high recommendations of TsAGI. unfavourable influencing on condi. island vortex bow vortex sponson's edge vortex Fig. tion of deck fighters Su-33 on aircraft-carrier «Admiral tions of flow around wing of aircraft at takeoff and on Kuznetsov» confirmed.1 Picture of flow above an aircraft carrier dead fluid zone t t / / / / / / *. Some test facilities.

5 Scheme of fore part of flight deck neibourhood of HACC «Minsk» a) original version.1 L and ß=-7° a) at trim ip= -1.1 L and ß= -7° a) at heel on the starboard b) at heel on the port ..5° by the stern modified version required level containers LD complex Fig. \V $ \ \ \ \ J:: \ \. f-*"f N \ \ \ \ . ' / ' Fig.6 Flow field over flight deck of HACC «Minsk» model ß= -4° a) original version b) modified version . ^ * \ \ \ \ \ v VX -})!/.5° by the bow b) at trim ip= 1.b) _ (\v ~* s x\\x 'H' '' '" |s WO» ^EB^r -.V*• X V V .. ™= Fig.4 Velocity fields in cross-section at X=-0.3 Velocity fields in cross-section at X= -0. b) modified version. 9-5 . c) comparision of flow parameters over deck original and modified versions b) \ \ \ \ \ \ \ t . Fig.a) -.

9-6 SIfitf Näss?» ■äHEp 1111Hife P Fig.7 Heavy aircraft carrier cruiser «Admiral Kuznetsov» 8 h> deg.8 Distribution of horizontal flow washes Eh= f(X/L) along glide path measured -10 behind model of HACC «Admiral Kuznetsov» at different angles of relative wind ß . 4 Fig.

and published in RTO MP-15. The Netherlands.es de l'Ecole enveloppes operationnelles sont presentes ainsique Polytechnique de Montreal.navire est basee sur les caracteristiques simulation such as aircraft launch and recovery. SMS defines a seaway. Sur le pont. envelopes are discussed. . en general. Sample pont ou Vindication d'un decollage non controle. such as visual aids. soi les differents obstacles ou l'etat du pont. An analytic approach to helicopter/ship ecoulement extreme de Vair provoque par des dynamic interface testing is presented. and flight readiness or availability. maximize operational flexibility. Factors affecting an air au moment oil un incident est identifie. such as. ship airwake turbulence. hangar de navire et son application dans la simulation de wall generated turbulence. gradients de vitesse violents (rafales) et par la turbulence atmospherique. Bernard LANGLOIS Senior Test Engineer. It is performed to reduce risks and principale du mouvement irregulier de la coque. hangar et l'exterieur. un glissement du train d'atterrissage sur le deck conditions (wet. The operationnelles et des facteurs de base des etudes de Ship Motion Simulation (SMS) Model is derived from l'interface dynamique aeronef-navire. floating object's response to the dynamic loads on its l'effet de sillage provoque par l'ecoulement autour de la structure. Spin-off projects into other du decollage ou du transfert de l'helicoptere entre le fields of growth. held in Amsterdam. dry. Dept. L'interface the relationship between the wave and ship motion dynamique est definie comme etant les relations entre spectrum. The lors du passage du navire. Virginia U.A. Quebec Canada Helicopters operating from small ships are limited in Sommaire the maritime environment by high winds and rough Le bilan de revolution d'une simulation du mouvement seas.S. Countries with a large mouvement ä six degres-de-liberte don't plusieurs number of platforms conduct DI testing as a matter of peuvent etre couples. and del'appareil. Dynamic Interface (DI) is defined as que les porte-avions sont limitees. etc). computes the de minimiser les risques operationnels en maximisant hydrodynamic and hydrostatic forces imposed on a la flexibilite tactique d'un Systeme aeronef-navire. wind-over-deck. 10-1 Simulation Tools in the Calculation of Aircraft-Ship Interface Operational Limits Application des outils de simulation pour le calcul des limites operationnelles de l'interface dynamique aeronef- navire Dr. The simulation is an extensive treatment of a sont le mouvement du navire. oily obstructed. Dynamic Interface Program Bombardier Services Arlington. präsente. deck spectrales du navire et de la mer. superstructure et les conditions physiques du navire. Ces deux facteurs sont la cause a moving platform. The est le produit des fonctions de transfert et du spectre de Aircraft/Ship Interface Simulation (DI) is a la mer definie dans le cadre du spectre rencontre. Le spectre de reponse handling. Bernard de FERRIER Engineering Manager. L'incident est vehicle on a moving platform are primarily ship defini par un renversement en roulis ou en tagage motion. Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". La reponse The application of ship motion simulation as a du navire est definie par la resistance cree par la houle developmental operational tool is introduced. ship (defined as the product of its transfer function and the seaway) and calculates a resulting ship time Les facteurs pouvant limiter la disponibilite d'aeronef history. are developed. En mathematical description of conditions limiting the simulation d'interface dynamique. d'autres applications telque les aides visuelles. ingestion. applies various definitions of seaway spectral Les etudes de l'interface dynamique sont realisees afin formulation. inappropriate deck lighting and markings Les operations d'helicopteres ä bord de navires autres limit helicopters. Ces conditions climatiques A brief synopsis of the theory and calculation of the maritimes sont ä l'origine des contraintes ship motion simulation program is presented. une limite est definie availability of an air vehicle. It incorporates seakeeping philosophy and un vehicule aerien et une plate-forme en mouvement. 870 (UAV) Bombardier Services Abstract Mirabel. Un helicopter/ship interface operational limits or incident peut etre identifie au moment de l'atterrissage. 5-8 October 1998. La simulation du primary application of the SMS is in operational mouvement du . ship stack hot gas motor l'interface dynamique aeronef-navire est id. par des the study of the relationship between an air vehicle and vents et de la houle. man-made obstacles. la turbulence sur le pont. il en resulte un necessity. Des exemples du calcul des Ferrier et Langlois sont diplom. In addition.

simulating any kind of air vehicle over and in France. Neither method alone can produce a w comprehensive and timely solution of the DI problem. simulating any kind of retention or handling paper is to present highlights of the analytic system. The traditional approach is experimental DI. such as. and artificial (degraded modes). have on the deck. Launch limited in the maritime environment by high and Recovery tests are rated by the pilot on an winds and rough seas In addition.10-2 experimental testing has numerous objectives. Experimentation investigates operational launch and recovery of vehicles. and analytical which is concerned with mathematical analysis and solution [3] . helicopters accepted scale. The purpose of this 3. Dynamic Interface environment [4]. approach to dynamic interface testing and 4. Simulation a matter of necessity. Recent and near future capital motion condition. USA). It is performed to reduce risks and maximize operational flexibility [1]. evaluation and deck handling. such as. simulating any kind of environment natural application. effectiveness. DCN Toulon (France). Other experimental analysis (DI) is defined as the study of the relationship are (but not limited to): aviation facility between an air vehicle and a moving platform. The pilot assess workload resulting hangar wall generated turbulence. While . Testing methodologies and procedures have been standardized by laboratories. increased interest in DI. such as. the Pilot Rating Scale are limited by man-made obstacles. simulating any kind of ship motion and ship capable ships [2]. Introduction the primary activity is on launch and recovery Helicopters operating from small ships are envelope development and expansion. Dynamic Interface Studies Brief Overview Dynamic Interface is divided into two broad categories: experimental or at-sea measurement and analysis. and DRA Bedford (UK). (PRS).The methods are not mutually exclusive. Countries with a DI analytics use mathematical modeling and large number of platforms conduct DI testing as simulation to support flight testing. ship stack from aircraft control margins. Britain and Germany. Naval Air Warfare Center (Patuxent River. such as. "Shipboard suitability testing" assesses the adequacy. The American Navy can be used to help define operational limits of matrix alone accounts for over a dozen any air vehicle/ship combination by: VTOL/VSTOL manned and unmanned vehicles and more than 20 classes of aviation 1. and performance in the shipboard lighting and markings. RAST and SAMAHE. such as 2. aircraft flying hot gas motor ingestion. vertical replenishment and helicopter in-flight refueling envelopes. acquisitions by medium sized navies. inappropriate deck qualities. and safety of shipboard aviation. engage and disengage of rotors.

Bretschneider [6]. Ship Motion Simulation The ship response spectrum is created as the product of An important DI analytical tool is the Ship Motion the RAO and die driving sea spectrum (figure 1) over the Simulation (SMS) which was initially developed by entire range of frequencies. computes the hydrodynamic and hydrostatic where forces imposed on a ship (defined as the product of its transfer function and the seaway) and calculates a Ar DOF amplitude resulting ship time history. The aircraft model is considered an extension of the ship. spectral analysis functions (k=48). In DI the limitations can be defined as the point at which an /-1994. history trace. oily. Factors affecting an air vehicle on a moving platform are primarily ship motion. Synthetic time histories are created probabilities in order to produce deterministic synthetic stochastically by summing the harmonic components time histories. The simulation is an w: a circular frequency extensive treatment of a floating object's response to the e: phase angle dynamic loads on it's structure. obstructed). and pitch back limitations. It incorporates seakeeping philosophy and applies various definitions of seaway spectral n=l formulation. over a given time period. Deck handling studies determine turnover limits. One Ship Airwake Turbulence. O'Reilly between 1973 to 1984 for the United reduced to sets of harmonic components for each degree- States Navy. lateral drag area and center of pressure. w: wave frequency (rad/sec) DI Theoretical Synopsis Sw(w) :seaway spectrum (m^-sec) Motion of an aircraft on the flight deck is calculated in Hs: significant wave height (m) terms of ship motion as a function of the aircraft model. which is given by: An example of DI analysis involves deck handling. Envelope dynamic ship responses for a specified load/operating studies will always require physical verification. such as. SMS by experimental or computational developed center of gravity. SMS defines a seaway. SMS can apply various definitions for the seaway. The model is defined by its landing gear footprint. sliding freedom. Wind Over Deck. of the most common is the definition called the dry. A typical time history equation is given by: SMS Theoretical Synopsis The Ship Motion Simulation (SMS) Model is derived from the relationship between the wave and ship motion spectrum [5]. deck The spectral characteristic of a vessel is defined in the location and orientation. In summary. Sr = S w(w) • RAO • / (V. it cannot replace experimentation.5\ aircraft/ship incident occurs. tiedown forces. where: TJJ: period (sec) traversing factors. transfer functions termed Response Amplitude Operators The aircraft experiences ship transferred forces and . 10-3 While analytics may seem less taxing to the DI study (RAO). and deck conditions (eg: wet. aircraft weight and intertias. and flight readiness or availability. Figure 2 displays a typical time and the calculation motion histories in the time domain. the Ship Motion Simulation The SMS fundamental relationship is: creates deterministic measures of ship motion from a probabilistic spectrum. condition [7].m): Frequency mapping Aircraft/Ship Interface Simulation (DI) is a mathematical V: Velocity description of conditions limiting the availability of an m: Relative wave angle air vehicle. pitchback or on-deck slide Sw(w)=iP4H. Incident means an occurrence of aircraft turnover.2 4 (2) at any point from touch-down to hangar stowage and W5T0 back to launch. The response spectrum is Peter JJr. The response amplitude operators define the process. The f(V. deck RAO: Ship transfer functions handling. Bretschneider. Time histories are produced by the sum of 48 synthetic SMS is divided into two basic themes. m) (1) Aircraft/Ship Interface Simulation where: Sn Ship response spectrum The primary application of the SMS is in operational Sw(w): Seaway spectrum simulation such as aircraft launch and recovery. The program methodology uses spectral of freedom.

an unidirectional Fix> continuous wind model. these accelerations on the air vehicle.o^ev.\|/) (transformation matrix from induced by ship motion is given by.8.. as follows: VAzcgy Spectrum for given SEA SPECTRUM (Ship's Speed and wave heading) i£ <j=J significant wave height and given encountered conditions Sw to each DOF for given <X3 encountered conditions and for each of "n" freq DAn. ship's axis system to horizontal level/vertical axis system).. EVn Aen.*••■. The lateral component is used to ^long^^lat Aycg = accelerations compute the lateral force.<02.^. to which it Y = T21T22T23 Fiy (5) is exposed ZJ Vr3iT32T33AFiz+wJ The inertia! loads at the helicopter center of gravity where: Tij = T (<{>.W2. •■ ■ COMPONENTS: *• '• "I. ». for various ship motions [8]. A„. *«. «">„. eYn Figure 1 . d>».eY^ """components HARMONICS?-'«. AB.ee2 A02 vz . In essence.ü) 2 'e 0 vi2 AY2. To compute the lateral force applied at the W = aircraft weight aircraft Center of Pressure due to the wind. the X\ /TllTi2Ti3\/ Fix \ aircraft is displaced as the sum of all forces.HEAVE I ROLL I PITCH I YAW' I SWAY I SU SURGE I C3SR = SwxRAO 1 At each of "n" freq ! h RESPONSE Ä"" „ I -4 *""' o I A"**" c I ""c .U2. is applied. a wind force is added to the ship motion induced forces In the Ship Motion Simulation.«.10-4 moments which create rectilinear and angular In the longitudinal. The wind vector is defined by its magnitude (Vwod) and its direction (\|/wod). roll 9= pitch W*AZcg yaw where. Next..^.e.Shin Motion Simulation Flow Diagram . Een A0„. whos vector is in the same fFjy ! = inertial forces due to ship motion direction as the seaway. Av2. I . »„. . lateral and vertical directions. W*AXcg are! WAYcg (4) 4.ev2 Ae2. The accelerations can be inertial loads become: numerically integrated to determine the position and attitude of the helicopter relative to the ship as function of time. the Vwod is resolved along the normal to the aircraft center line /AXCg\ ). """ c » " " (P «e lag at each SPECTRA ule u)e We C0e (Oe (oe I I I I I I O Harmonic Analysis imo N.eY2'•'•': .Ul. <V £0n Yn. AYl.---"... A^. A AVn.

10-5 20 40 M M 100 120 144 1*0 1*0 200 3» 244 1*4 2*0 TIIC IN U09NM •0 «0 M 00 100 1M 144 IM IM »0 «0 1*0 IM WO 10n IM 140 tM IM MO t» *4A »M Figure 2 .Turnover OttUnc« Rotldual TODR -on RTA. InortltJ Fore« Lift TODR.Residua! Tumovor Angle Rotor Static Dynamic Turnover .Typical Time History Trar«.

-»*^) (6) A( Klnertie) = tan'-1 (11) (K)(LBL) where: Ay =Aircraft projected area normal The simulation model assumes a constant wind. The model is lined up with the ship centerline and is rotated on the deck to find the least stable. The aircraft operational limit is produced LßL = Left wheel butteline owing to the break-down of the aircraft handling system. The azimuth of these two lines are then determined with respect to the ship's FRMGdnertie) = Y [ LßL +"RßL J (9) longitudinal axis. They define the distance that the CG should move F RM(3nnertje\ =Right main gear axial force due to the for a turnover to occur (right or left). therefore. P ) water or oil. Fwy =Lateral force applied at the aircraft However. an aircraft incident is WLQ = Ground wateriine registered. the 'worst case' hinge line on the flight deck about which the aircraft will turnover are Axial forces on the main landing gear due to aircraft defined by Rto and Lto (right turnover and left turnover). A maximum value of the encountered force load or Fwy = Wind lateral force component geometric ship position is preprogrammed. nose gears are modelled unlocked and castored for turnover.WLG.10-6 FRMGÜDSrüel \ F*. as: Assuming perfect rocking. The distance from the aircraft center of gravity (CG) to where: each line is computed as TODR and TODL (right and WLQ =Center of gravity wateriine left). They are expressed in equation (5).(^y > 7£L CJ ro affects aircraft stability by changing the coefficient of friction between the aircraft landing gear and the deck. F ^G(Inertic) = -F^G(Inertie) (10) They describe the angle between a vector from the CG normal to the Rto and the Lto and the vertical. The total incremental change in the aircraft roll center of pressure due to wind with respect to the slup is given by: The axial forces on the main landing gear due to the wind force Fwv is given by: Afytotal) = A<|)(vent)+A<t»(inertie) (12) Deck conditions. but realistic. is given by: Each line is computed from its main gear position to the nose gear swivelled for turnover. A<t>(inertie) is continuously changing with ship motion. where: Aircraft handling systems are handled much in the same FRMG WIND = Main Gear ri nt ^ 8 ) "^ force way. The hinge line about which the aircraft is The incremental aircraft roll due to inertial loads is likely to pitchback is the line which joins the right to left determined by: main gear. inertJal forces in the plane of the main gear. The distance from the CG to the hinge line is defined as FBD (pitchback distance) and expressed as: . RßL = Right wheel butteline Scenarios are programmed for the "worst case" condition. where: FLMG/WfljgN = Left main gear axial force Similar boundaries are computed for the pitchback condition. AZrto and AZlto. When either force loading or ship angular position is greater than the WLcp = Center of pressure wateriine manufacturer's design limits. This parameter FRMGwpM . where: K = spring constant Referring to figure 3. The incremental aircraft roll due to the wind is given by: For the greatest landing gear deflection. These lines describe lateral inertia! force Y defined an angle TOR (right) or TOL (left). to the VWj« component A$(wind) is constant throughout the simulation run. is a variable in the program. such as. the axial force on the left TODR 4 TOR (13) main gear is vectorially opposite to the force acting on ^VWLW-WLG the right main gear *TOL=tan-l( TODL (14) <WLW . eg: dry or with substances. orientation (6gure3).

The intention of analytical DI is not to replace experimental DI but to .Aircraft Model n»finirtnin by ship motion. In the sum of PBD = (CGx-MGx) (IS) forces. when the vertical where. PBD= pitchback distance and PBD essentially reflect system stability. TODL. the WLw = Wateriine to the aircraft CG deck friction goes to zero and an unintentional liftoff is WLg = Watcriine to the ship deck indicated. Dynamic turnover caused by the rotor disk (uneven loading of the rotor) or general flight readiness or availability. deck handling and causing the aircraft to turnover. The sum of these incidents trace aircraft-ship operational envelopes. Turnover incidents are static or dynamic in character. The resolved Limits weight vector migrates beyond either the friction forces The objective of analytic DI is to identify operational causing the aircraft to displace or the reaction forces envelopes for launch and recovery. 10-7 Forward WATULINIOnO STAW Starboard Y BLW Figure 3 . when The associated pitchback angle or PBA is given by: the landing gear friction values are exceeded by the combination of aircraft apparent weight and induced *PBA= „„. but usually undesirable geometric solution.l^MGx-CGx^ inertial forces. In similar fashion. inertial force equals and opposes the aircraft weight. The aircraft center of gravity is in motion. the system is MGx= aircraft main gear station unstable and will seek to find a more stable. or ship motion or both. the same phenomena occurs. The distances TODR. Calculation of Aircraft/Ship Operational Static turnover is the same as on shore. Aircraft slide will tan" \w^mt) <16> continue until the aircraft frictional forces are greater than the disturbing inertial forces. At the CGx= aircraft CG station point where a distance becomes negative. slippage will occur. Finally. the weight vector is continually modified in response to inertial forces applied by either the rotor disk where.

the entire data point is declared out-of-limit. 3. Here Office for the Direction des Constructions Navales (DGA the models A and B show better limits than helicopter C France). The aircraft are set on the ship's relative wave angle. In the following seas conditions. The helicopters are vehicle exposed force loads as a function of a given ship's modelled centered at the bullseye. the deflection and forward gears are modelled unlocked and rondelle shows increased encountered loads at 75 degrees castored for turnover. 6. every 15 degrees Sig. Areas within the shaded areas are inside operational limits. and significant wave height For example. of gravities and corresponding minimum mission weights. Here the shaded zone indicates air and fuselages unfolded and locked. 9 metres METRES (Equivalent Sea State «. as centeriine and rotated to -20 degrees to provide the least one would expect. 5. stable orientation. The bow of the ship is along the principal axis to the right out to 0 degrees relative wave angle. The air vehicles were modelled both secured and Rondelles maybe used to indicate encountered loads as unsecured on the deck with rotors spread and free to rotate presented in figure (6). Ship speed. 3. significant wave height and modal period are the primary ship motion markers.Encountered force loads . ondeck slide or uncontrolled liftoff.180* . A wavt angle typical test matrix is furnished in table (1). Several aircraft were modelled with high center for the same ship conditions. 9.Tvnical Test Matrjx Figure 4 . encountered loads are minimal. Once operational envelopes are DtckSafMrlSkMUWatl»«* iwmauttnmt) calculated. Inc A comparative operational limit sample between various Canadair Defense Systems Division's Dynamic Interface aircraft on the same ship is presented in figure (5).8 0. relative wave heading. Interface testing is performed according to the test matrix ii» indicated in table (1).Samnle Deck Envelope Several studies have been achieved by the Bombardier. Ship Velocities: 05. Deck safety rondelles are created as mttres a function of ship velocity and deck condition deck condition (figure 4).5 Wind-Over-Deck: 0 • 50 knots Table 1 . At any point during a simulation analysis an incident is identified. 11. pitchback. 6. 15 seconds additional matrix attributes Deck Condition: DRY and WET (water) (Coefficients) 0. DI test engineers would randomly verify selected data points in and out of the envelope. 180 degrees in bearing 270 (by symmetry 360 degrees) and a maximum of 50 knots wmd-over-deck Figure 6 . Incident means occurrence of aircraft turnover. In the examples which follow. 10. envelopes are based on limitations defined by the point at which an aircraft/ship incident occurs. Each concentric ring relates a relative wave height and significant wave height All cases are tested in seas ranging from 1 to 9 meters. IS. 7) Modal Period: 5. 20 knots Wave Angles: 0 .10-1 compliment the activity. The landing gear velocity.Wave Height: 1.

the index contains quiescent ship motion periods. The angular displacements and velocities weighted by system furnishes this information about any kind of dynamic coefficients. it capitalizes on the individual degree-of-freedom. aircraft lateral. and vertical. These limits are expressed by a LPD code calculates the rate at which a vessel can scalar empirical formulation. This does not suggest that the accelerations. sum of the squares of the various parameters and terms LPD is designed to reduce pilot workload in completing representing real-time ship/aircraft interface motion. The index is modular in design with Application of DI tools to the operational environment the capacity of incorporating other parameters (e.g. the energy index. In short. ship interface activities by improving operational security in the reduction of helicopter hover time. their rates. One wind-over-deck module) to improve energy index such improvement is the CL352 Landing Period significance and applicability. the index air vehicle. The degrees of freedom. are considered the most important for structural dynamic limits. Real-Time DI Applications. This ability is based on acceleration. All of the parameters are weighted by index is predictive (using historical information to dynamic coefficients which are weighted according to the extrapolate into the future). vertical supplies real-time information about the motion of any velocities and accelerations as well as roll and pitch vessel as a function of helicopter operational limits. The Energy Index Designator (LPD) helicopter landing aid. lateral and vertical velocities and in the near-term future. The limits of the helicopter. structural and dynamic operating limits of the matching velocity and acceleration terms. (yaw and surge) are monitored for motion within certain limits and may be incorporated more actively later if The energy index is an empirical formulation warranted. . the coupled degrees of rate at which a vessel can displace due to natural freedom and normalized according to aircraft hydrodynamic forces as a function of the structural and characteristics. Rather. termed.: has produced numerous real-time improvements. The energy index uses eight parameters roll furnishes information of the motion a ship must travel and pitch. pitch. EI= a1y2+a2y2+a3z2+a4z2+a5r+a6r-ra7q +agq (17) (where a„ z^. The equation in the Mk m is the aircraft in any sort of sea condition on any sea vessel. velocity and displacement terms which ship motion as a function of the mechanical and dynamic determine the motion of the ship in the near future.. and user experience into a motion sensitive tasks (in particular launch and recovery of air vehicles). viz: roll. T*>»w '• i»^*—r~ —r-j-v« Hl/ VM 15t/ V» 15 \L V» /»• /»• *lt\ $m *«•'*■—L_ l*-*l« / "• ii A. Visual Aids meaningful value. displace due to natural hydrodynamic forces against the The index identifies ship quiescence using displacement.. «•^"^-J— —\-~*$n H«\ «•*—1 _J—-^JM AS532-FrigateSTAB AS5365-FrigateSTAB NH90-FrigateSTAB Figure 5 • Comparative fWrtttinn»! limit. designed to convert ship motion characteristics. The LPD equation of LPD Mk III measures lateral. are weighted dynamic coefficients) The LPD may show improved recovery opportunity from its ability to identify the onset of As indicated in equation 17. The remaining two degrees of freedom dynamic characteristics of the approaching air vehicle. 10-9 AS532-HandlingSys-Frigate AS565-Handling Sys-Frigate NH90-HandlingSys-Frigate Handling Sys-FrigateSTAB ui^^-t —T^N^** '«^""T".

pitch angle and pitch rate. A23. The energy index for aircraft activity. I LAND NOW I The phase lag between coupled degrees-of-freedom contribute to the stability of the energy index.Lunrfin«. Exceptions to this object of much investigation. and pilot-in-loop factors (see figure 8).il*« The third step compares the aircraft limitations scale completing the calculation of the appropriate When the index is low the ship is stable and the ship weights of each degree-of-freedom. is directly based on the ship normally refined during initial LPD sea trials. during Calculation normal environmental conditions. (see eq. the significant wave height and the modal period. For a given are communicated to the user. The ship can only displace from a is then calculated and compared to the established stable to a high risk condition by the introduction of threshold (green. quantity must reflect a physical state of the aircraft/ship high energy intensity wave fronts such as those created combination in a given sea condition. In the first step. this minimum The meaning of the index value has been the measure is about 4. t. A6 A61 A62 A63 A7 A71 A72 A73 A8 A81 A82 A83 "Al""l f All • A12 • A13>] CONVERGENCE TEST A2 A21 <■ A22 -• A23 A1 TO A8 RESET FLAG = 0/1 A3 A31 <. COEFFICIENT MATRIX CALCULATION lateral velocity and lateral acceleration. and vertical MATRIX A = A1 A11 A12 A13 A2 A21 A22 A23 velocity and vertical acceleration. These relationships are A3 A31 A32 A33 direcüy related to the ship's velocity. For the 7.5 to 5. A32 <' A33 A= A4 -< A41 -• A42 ■■ A43 *► ENERGY INDEX CALCULATIONS E1=A1 PH12 = A2 PH1 VEL2+. For a FFG-7 or Type 23 class ship.0 seconds. BEFORE the coupled peak in vertical displacement Kgure 7 . red) scale the results of which certain quantity of energy from the sea. by an earthquake or weapon explosion.10-10 Methodology for Coefficient Calculation INSTRUMENT PACKAGE The calculation of dynamic coefficients is SHIP MOTION PROCESSOR performed in three distinct steps executed simultaneously. A82 <. as defined by the deck (PRS) [9]. the relative wave A4 A5 A41 A51 A42 A52 A43 A53 angle. mass of a Destroyer class of ship. of helicopter and ship. A5 A51 ■■ A52 -■ A53 (18) = (A)*(D0F) A6 A61 • A62 <' A63 "Hü THRESHOLD CRITERIA PERFORMANCE EVALUATION A7 A71 < A72 -• A73 E1 « EISAFE (FOR PROTOTYPE ONLY) L J UNDING SIGNAL = 0/1 A8 ^A81 .. The product of the motion is small.18) produces the danger threshold the landing deck motion is acceptable energy index coefficients in real-time. time necessary to raise the deck from a stable to an unavailable condition can be derived experimentally The flow-chart of the energy index is presented on figure from the calculation of the maximum Elmax. A relationship is derived for roll angle and roll rate. Pitch and vertical ■ w motion usually occur together though rarely in phase. The definition of each deck security zone is security zone criteria. this measure is about Methodology for Energy Index 'Motion Zone' 5 seconds. Deck motion The energy index value is analogous to the level security limits must be established for each combination of kinetic and potential energy contained in the ship. aircraft limitations (defined). Coupled means that the degrees-of. These limits may be measured . To be applicable. characteristics (measured). P»rjnH IWp. the scale is initially divided into four 'deck security' or 'availability' zones similar to the Tilot Rating Scale' The deck availability. = DECK ACTIVITY freedom are directly related and can only occur independendy in very special cases. A83^ PILOT DISPLAY The degrees-of-freedom that are considered 3D highly coupled are roll and lateral motion and pitch and MOTION DATA VS AIRCRAFT LIMIT vertical motion. For expedience.«^ Cnmnnni. When the index value is below the element coefficients Ai i. yellow. A maximum in pitch will often occur some time. condition. the rule occur when encountering longitudinal propagating. relative coefficients are SHIP MOTION DATA AIRCRAFT LIMITATIONS established between each of the following four degrees of PHI PHI THE THE VT VT LT LT PHIMX THEMX LTMX L1MX VTMX VTMX freedom and their derivatives.

average about 50 seconds longer than day landings (other [7] Baitis. C. Washington. inherent operational limitations. Analytic study is used to rapidly delineate conservative measure. In a determine the maximum safe air vehicle/ship platform red light helicopter recovery. a limit condition is identified. Applebee. J. YAW. the LPD. W. Monterey. W (1953). New York. DI-NAWC. Night landings without the LPD DTNSRDC Report. Proceedings of the First DI Simulator and at-sea testing have been conducted by the Working Group Meeting. in order to provide experimental DI with the necessary data to more capitalize on ship physical motion constraints. Vol Differences were detected between LPD day and night. Simulating the Helicopter- ship's structure is not sufficient to displace the ship into Ship Interface As An Alternative to Current Methods.Time to Land from Common Way- that is drawn on the energy index scale for a given ship. FFG7 and the FF1040 Ship Classes". Naval Postgraduate School. the effectively set testing strategy to probe the limiting operator must await a flashing green signal. Therefore. B & Semenza. VERTICAL) Figure 8 . The primary analysis [3] Ferrier. British and German navies.3 for Beach Erosion Board. Given an air/ship system and exceeded acceptable aircraft limits.S. However. R. US. A (1988). Figure 9 displays this result for both day and [4] Ferrier & Semenza (1990). AIRCRAFT) AND FORCE DIRECTION DEPENDENCE (i. The energy conditions. again between no LPD day and night calculated from a [6] Bretschneider. night. T. From the data. A limit is 0 100 200 defined by the impact that a certain ship motion condition may have on the structural integrity or TIME TO LAND (SECONDS) dynamic response of a given helicopter. R (1976). "A parameters held constant). and 61. D (1988). Proceedings of the AHS. one or more DOFs have operational limitations. On the Motions of Ships in Confused Seas. NATC Manned after concluding that the LPD performed as per Flight Simulator VTOL Ship Motion Simulation and specification was to compare recoveries with and without Application. a red line condition in under some specified period of NPS67-86-003. took more than 25 seconds longer to complete than the [8] Blackwell. Transactions of SNAME. Model of the On-Deck Helicopter/Ship Dynamic Interface (U)".t. During the day without the Non-Aviation Ship Motion Data base for the DD963. time. Aeronautical Research Laboratory. Melbourne. . "Wave Variability and common way-point to the ship deck. The sum of these limits produces a red line Figure 9 . 10-11 CHARACTERISTICS ACnVTTY CHARACTERISTICS (l. night recoveries take on Technical Memo No: 118. Patuxent River. L (1959). recoveries with LPD. [5] St Denis & Pierson. J (1990). The deck is available for aircraft system limitations. Army Corps of Engineers. LPD. [9] Ferrier & Semenza (1990). deck and energy index traces were used (see Figure 4. DI strives to increase deliberately assigning the red line several scalar points tactical flexibility for any set of environmental under the calculated absolute limit is a prudent if not conditions.Threshold Criteria experimentally or calculated analytically. If the condition exceeds an operational specification. U. an example). flights lasted on average almost as long as night CG26. The red line is absolute. Height over the Wave Spectra for Wind-Genertated Gravity Waves". J and Feik. point CONCLUSION All energy index values under the red line infer The overall objective of dynamic interface study is to acceptable deck motions. Aerodynamics Technical Memorandum 405. defined for a flashing green condition infers that the REFERENCES potential energy being transferred from the sea into the [1] Healey. Meyers. The calculated system limitations activity under the red line. A E. Washington.e. Washington. ROLL. FF1052. SPD-738-01. [2] Carico. with and without the LPD. "A Mathematical same mission with the LPD. G. Val (1986).

The UAV automatic pi- discussed. These results show that this ap. It is easy to understand why bullseye centred an assessment of the proposed concept obtained by sim. data has demonstrated that state vector (which consists of a velocity vector and an Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". within the framework of a Franco. The operator can only inter. The key ideas of the proposed approach are sessed only at the instant of recovery. A shipboard manual operator does not ginates from the ship's inertial navigation system. 2 Design of the system The automatic recovery system concept proposed com- 1 Introduction bines certain essential hardware and software elements.that is used in describing that data that is subsequently acy. guidance and control.recoveries are difficult in agitated seas.difficult. The guidance/piloting system's objective is to advantage of this guidance and control approach is recover the air vehicle to the deck bullseye based on the complete decoupling of the UAV from ship mo. However. held in Amsterdam. recovery becomes increasingly proach is more robust than conventional motion sens. The various methods used in each module of system advantage is its basic simplicity. Strategy chosen in the development of the lot must be at least as rapid as sea generated perturb- whole automated recovery system is described. In addition. and landing deck position predic. the quasi inertial reference system. gin of safety.to the ship deck. When resolved. These systems the estimation of the velocity/position with respect are driven by hardware solutions limited by the guid- to the ship in a quasi-inertial coordinate frame and ance system's capacity to judge accurately the relative the guidance/control of the air vehicle to a virtual position and velocity between the air vehicle and the target (predicted position of the landing grid). including: performance is directly related to the sea condition.trajectory of mouvement. its the whole recovery system are presented. Experience has shown that manual recoveries are accumulated. The system is based on two essential ideas (see figure This article summarizes a study defining an automatic shipboard recovery system for a VTOL Unmanned Air In the first instance. The reference system is applied to UAV made with considerable difficulties in seas greater than tracking. filtering. for example. Reboulet(*) P. 2 Av. Canadair Defense Systems Division 10000 rue Cargo A4 Mirabel (Quebec) Canada J7N 1H3 Abstract automatic recoveries are required in seas greater than sea state 4. Virginia 22202-4306 (USA) (***) Bombardier. 11-1 Piloting of a VTOL-UAV to Shipboard Recovery C. . The ship. As the seas ulation is provided. The Netherlands. Mouyon(*) de Ferner. This requires that the sea state 3. (***) (*)ONERA/DCSD.Canadian collabora- tion project between ONERA-DCSD and Bombardier. This is the reference system pret shipborne movements with some degree of accur. pratically itive techniques. B(**) Langlois. Inc (Canadair). In tracking. ations. This have all the information necessary to ensure secure furnishes the information to characterize the vessel in vehicle recoveries. The operator can . 31055 TOULOUSE cedex(France) (**)Bombardier Services Corportation. The system presented below attempts This activity is sponsored by the Delegation Generale to avoid the problems described above by an approach de PArmement (DGA France) and was accomplished considerably different than that of traditional systems. Modul performances based on test results are that experienced by the ship. 5-8 October 1998. Finally.data received from the tracking system. 1111 Jefferson Davis Hwy # 510 Arlington. The ship deck position being as- scribed. Ed. Dr. In seas greater than sea state 5. The recovery tions. identify the effects of wind gusts only after the gust it formulates a coordinate system describing the ships has displaced the vehicle. Current automatic recovery systems gen- A system of guidance and control ensuring the safe erally use a tracking system to recovery an air vehicle recovery of a VTOL type unmanned air vehicle is de. The system promises to dramatically impossible to have recoveries with a reasonable mar- increase UAV operational limits and recovery efficiency. UAV accelerations are equal to or greater than tion.essence. and published in RTO MP-15. Belin. B.become more agitated. the data being handled ori- Vehicle (UAV).

however. important impact on the architecture of the system. A method based on pseudo. An important advantage to this approach is that the guidance and the flight control modules are de- coupled from the perturbations caused by ship motion. such as. This alone is the state of the vehicle and ship (state variable model) insufficient. The equations de- navigation system is combined with that from the UAV fining the position of the UAV are those of a classic . It is for this reason culated measurements are compared with actual meas- that the information originating in the ship's inertial urements from the tracking system. The definition must be converted to the and a model describing the tracking system. The air vehicle position be tracked using angular sensors. it is possible to determine the po. The mathematical model which de- that radar is used. This methodology is described below. Y(UAV) erence system. based on these data. When the ship averaged accelerations are con- ure 2.scribes the system is composed of: model describing sition of the UAV relative to the ship. the helicopter-ship system responds. an optical or radar Using only angular information in a similar way tracking system. of the AM vector and its derivat- The primary structural modification is the creation of ive located in the inertial reference system (see figure various data streams coming from several data sources 1). ery systems. Ship INS Figure 2: Sensor architecture ^hip-y/VV^syrsteni ship accBlaratl \' noise Figure 1: Proposed approach error (measurement/ model) The second essential element is a spatial defini- ^&*n tion of the bullseye which is used to guide the UAV to its centre. into the inertial reference system. The primary advantage or this system UAV INS is to pilot the vehicle based on control laws unaffected by ship's random mouvement. inertial navigation system. The vehicle may vehicle as function of the ship.output is the definition of the relative position of the ceivers provide angular positioning. defining the deck attitude. The illustration is of an autoland architectural sidered as input (originating from the inertial naviga- system using a proposed sensor arrangement necessary tion system). This is contrary to the instanteous defini- tion of the bullseys used by traditional systems. The proposed system calculates the virtual position of the bullseys which is independent of sea created ship pur. R (V^) r--. The data handling methodology used is classic. The technique of data fusion used is illustrated in fig. will have an The problem is to determine the components.11-2 UAV positioning vector) be reassembled in the new ref. be 3 Tracking UAV used. This approach.measurement techniques was presented in [1]. Figure 3: Data fusion method turbations. other techniques may. These cal- quasi intertial reference system. INS data used are essen- It is reasonable to assume that this approach can be tially accelerometric and vehicle orientation converted considered more robust than that of the classic recov. however. is measured by a device. The for continuous vehicle tracking. Two optical sensor re. Several methods may be used to track the vehicle. This is illustrated in figure 3.

ditions.1 Choosing the virtual bull's-eye X(k+1) = A . It is this order of value difficult to predict with precision the exact moment of what will be used in the vehicle piloting cycles. Excellent results are produced when combining these two types of data collection sys- tems. On the other hand. Average motion calculations could in agitated ing made with 2 optical antennae providing a RMS of seas produce amplitudes greater than the dimensions 0. The simplist method displayed in figure 5. to predict the position of the virtual landing deck Various strategies for vehicle shipboard recovery may at the exact moment of recovery. This lß=g(X) +W2 definition is used by traditional recovery systems with limited performance results (particularly in higher sea states). [7(UAV)-Y(shlp) ]+v As discussed earlier. This last comment is important to the inertial frame (which may be accomplished as a because this differentiates between the classical fixed first approximation) the position of the vehicle can be wing UAV recoveries from that of a VTOL UAV. It is difficult to im- be considered.5m and a sampling rate of 120ms. It is feasible to place the descending technique is used which provides the ship position as vehicle on station hold at various points during its re- a function of the average Rj which is related to the covery phase. ated to be 0. X(k)+ B. It is also possible. using a control law. The UAV is unable to follow the ship's mo- tions in agitated seas.002rd. ship / Instantaneous / ship location / average trajectory : Y(UAV). recovery for a fixed wing UAV. lected since its precision is limited in higher sea con- The second part of the curves represent vehicle track. At every step the state vector is cor- rected by true measures which are a function of the variation between calculated and measured values of ^JÜ\J"""""i i _ the system. the information provided by accelorometers are more accurate in the short term while drift is an important factor in the long term. .1m/s2. It is known that the data provided by the track- ing system (optical. This method was not se- with a RMS of 0. it is possible to predetermine the length of time the recov- 4 Various strategies ery phase will be. Refering to figure 1. From the transformation of the Ri frame phase in advance. of the flight deck. 11-3 Kaiman filter. A classic data filtering over the deck. The results are culation of the virtual bullseye. By using this technique. Figure 4 is a flow diagram for the mod. The noise from accelerameter signals is estim. This decouple the calculation from random ship deck excursions. It is possible to program the descent trajectory. for example) are excellent in the long term but are relatively noisy. There are many plement but creates a much more preferment system. However.Y(sWp) 4. It is deduced as a function of Ri. the system does not use an in- |a=f(X) +W1 stantaneous definition of the recovery bullseye. For a VTOL UAV. terminal point of the guidance recovery trajectory. The first part of the curves are is to take the average of the position of the bullseye produced by vehicle tracking made by a relative GPS over some period of time. The purpose of this module is Figure 4: State representation to predict the location and attitude of the bullseye at the moment of UAV touch-down. An Extended Kaiman Filter is used to calculate Several methods may be considered in the cal- the state vector used by this model. the latter is prefered since it allows decoup- ling of the vehicle command and control system from ship motions. the reconstruction of the state vector is just as easily calculated in the ship's Figure 5: Kaiman filter results reference frame as in the inertial coordinate system. The strategy selected uses the unique character- All that remains is the calculation of the ship istics of a VTOL particularly in its capability to hover position in the inertial frame.choices to make: the recovery point is defined as the elling of this system. The optimal beginning of the final recovery phase must like- w (UAV location) wise be defined.

For more details on the LPD. The autoregressive ship model (see figure 6) is derived from classic least squares techniques... y. ship sea motion Chin disturbances e — S# & »-*■ Figure 7: Virtual bull's-eye trajectory b(t) y A white noise (0.y(k — 1). . This given by the expression: The second point consists of selecting the optimal in- stant that guarantees a safe recovery. heave). At the inverse Figure 6: Identification scheme the outputs y(k). The deck posi... The energy X(<7-1) and Y{q~x) correspond to the minimal index is constructed such that 4 energy states are de- solution scribed.11-4 4. The calculation may be made off-line based on ship motion characteristics. virtual bullseyes trajectory during the landing phase.1) / least square -«£= algorithmes The first term of the right side of equation 1 can- not be calculated since at k instance.2 Virtual bull's eye position predic. it is necessary to pre-calculate all the equation: the polynomial Y(q~1).b(k + m) is unknown. Thus. the This model is used to predict the location of the best prediction for the output y(k + m) is given by: bullseye at the instant of recovery.. This calculation must be completed 5 secondes prior to the beginning of the vehicle descent. sway. considering the situation at time k. The left portion represents motion in the ver- b sequence of independent random variables tical plane while the right part indicates motion in the k current time horizontal plane.. It is important to remark that each values of Y(q_1) are only valid for a considered *(*) = Ä^W time m. where Figure 7 displays the performance of this ap- y output sequence (ship motion) proach. The position of the deck is given by k + m where 4. z directions (surge.3 Selecting the recovery moment m is the position prior to the instant of recovery. y(k + m/k)=Y(q-1)y(k) tion is calculated with a model defining each of the It is the term used to predict the location of the three degrees-of-freedom mentioned earlier. deck motion while the heavy lined graph represents the troduced by Astrom. A calculation is made using an energy index which determines the y(k + m)= _n&(fc + m) period where ship motion is below a level which en- A{q~i) sures a safe deck. It is noted that the model A(q~1) is valid at any given instance. Vertical plane Horizontal plane tion The calculation methodology begins with a five minute ship motion time history in the x. In brief the Landing Period Desig- rewriting the equation nator (LPD) is a helicopter recovery aid which calcu- lates ship motion as a function of helicopter operational limits. the random se- quence b(k +1). a discrete-time single-input-output plant governed by In summary. The energy index collapses ship and helicopter y(k + m)= Xfa"1 W* + m)+ Y^^k) (1) limits into a scalar value while retaining the physical significance of the helicopter ship system. Consider deck. The lowest energy state (very safe) guarantees with respect to X of the following polynomial the aircraft operator a quiescent flat deck for a delayed equation: period of time consistent with basic principles of New- tonian mechanics. . please A(q-1)X(q-1)+q-mY(q-1) = l (2) refer to [2]. have been already ob- served. The dotted graph displays the flight q forward shift operator qky(m) = y(k + m) in.

radar. This is obtained by the use of an automatic recovery as defined above. winds and sea conditions. This is based on state 3. trajectory. 400meiers above the flight deck. ship classes are included. dis- the landing phase (note Ri this reference coordinate tance from the centre of the recovery grid. This approximate state condition is shipmotion. used by the control laws module to map the vehicle in The automatic recovery algorithms were pro- cartesian space. It serves to pilot the UAV to the predicted ■hip Bccal »rations position of the deck. There remains the of the results provided an estimate of the feasibility of calculation of its vertex. etc). even in the heaviest seas. Figure 9: automatic deck-landing results 6 Autoland test results Simulation studies of the entire system were performed 5 System general architecture using an AD-100 combined with a VAX3500 worksta- tion.5seconds the LPD signals that the deck is very safe Figure 8: System general architecture and a quiescent period has begun. A test matrix was programmed showing navigation data filter module has two functions: seas from sea state 3 to 8. relative ho- system). It is noted that as the prediction time is reduced. The simulation pro- tial navigation system and a tracking system (such as gram contains an environmental model which defines GPS. Between sea state 4 to 6. The inertial agitated seas. From the studies it can be concluded that safe re- The second function is to calculate the virtual covery can be achieved in seas upto and including sea bullseye in the Ri reference frame. The program includes a six degree-of-freedom The system architecture is shown by figure 8. the virtual bull's-eye's position is cor- rected. After hundreds of simulation cases. At each calculation update (several times/second). The simulation model of the dynamic and mechanical defin- tracking module reconstructs the state vector from in. as well. The guidance law fixes a descent trajectory which permits recovery within bseconds. itions of the UAV. Transfer function models of various formation furnished by the ship and air vehicle inter. The statistical analysis the ship inertial navigation system. At the onset of recovery the UAV is normally located in hover at 5m above the deck. scribed earlier. This included various mech- . The flight commands to the UAV are of paramount importance in the first instants. The coordinate orientation is known from rizontal and vertical velocities. This identifies shipmotion quiescent periods for vehicle re- module brings the air vehicle to the hover station at covery. the prediction is improved. The descent slope must be likewise corrected in order to respect the bseconds recovery time. rarely surpassed 50cm. 11-5 the prediction laws described above. and reference coordinate system related to the average ship so forth. It is a reference coordinate system that can The assessment of system was based on a set of be regarded as quasi-inertial related to the duration of crieteria which examined UAV tilt on recovery. test results show that . the flight guidance module descends the air formance of the entire system in various conditions of vehicle to the virtual bullseys position. At that moment the motion predictor calculates the location of the virtual bull's-eye at the moment of touch-down. The coordinate system defined by Rj is paramount in the autorecovery py fl system. At 476. the deck prediction error. Figure 9 represents the strategy de- Cm' «"""ESU.The first operational function is to detail the anical conditions including loss of tracking system. a high pass filter. grammed along with a model representing the Landing The path planning module generates the flight Period Designator (see reference 3 for details) which path trajectory in the final approach phase. poilHor Each of the modules described above were tested •»tlmaöon by simulation individually and assembled into a com- tracking m*a«ur«m*nt pleted system. From the hover The simulation objective was to assess the per- station. It is for this reason that the prediction estimates need to be good early in the process.

Simulation. sidered. An architec- ture using a redondant optical tracking system is an acceptable tracking technique. LANGLOIS. tem" Proceedings of the American helicopter Society. must be equally considered. favoured the optical tracking system. unlike radar systems. "Conception et developpement d'un Systeme d'appontage mari- time automatique d'un vehicule generique de type . V.and J. Simulations res- ults. BARRAL. de FERRIER. it is highly unlikely that both tracking systems would fail at the same time. BERGERON. optical tracking sys- tems are not confronted with the problems associated by mulitpath. The costs involved. The results were comparably acceptable. data fusions techniques used in this study were achieved using Kai- man filter applications. "Development.and C. REBOULET. de FERRIER. pp 25/1-25/8. VTOL" AGARD CONFERENCE PROCEEDINGS meters (none of the success criteria failed). May 1994. Us- ing redondant systems. FUERTES. The task at hand validate the system is by assembling and testing the system at-sea. The purpose of this article was to describe the design C. Data handling fusion tech- niques from several optical receivers may adequately compensate for the deficiencies noted.11-6 the prediction technique well within acceptable para. sea states. in- spite of the known problems associated with its use in the maritime environment. May 1998. What ever the technology chosen. however. It is important to recall that the primary advantage of the autoland system proposed its robust character and its capabil- ity to operate even in agitated seas beyond sea state 6. SEMENZA. From simulation results serveral an automated UAV-VTOL shipboard recovery sys- tracking systems are possible. REBOULET. J-C. In higher 591. References [1] B. Simulation results supported deck recoveries un- der 50cm from the centre of bullseye. F. ignator (LPD) Helicopter Recovery Aid" Proceedings of the American helicopter Society. Two systems were con. Preliminary analysis has shown the system to be feasible. Wash- ington 7 Conclusion [3] B. One tracking system used two optical receiv. The module calculates the virtual location of the recovery deck at the precise moment of touch- down. The deck prediction module is considered indis- pensable to ensure safe recoveries in agitated sea con- ditions. and Testing of the Landing Period Des- rate predictor should be included in the final system. de FERRIER. In addition. (1996). Prom simulation res- ults optical tracking system'failures may occur during the most critical phase of the vehicle recovery. however. Careful trade-off analysis is necessary when deciding on the components of the autoland sys- tem. "The design and test using simulation techniques of manned Air Vehicle.Washington ers while the other used a radar tracking system. This infers that a vertical [2] B. The Kaiman filter application reduces the tracking error to under 20cm permiting the system to be a feasible solution. the only failed recoveries occurred owing to exceeded vertical velocities. of an automatic recovery system for a VTOL type Un. B.

of SESs is usually of the finger type. The new results for added resistance are frequent motions of SES. oped a non-linear six degree of freedom motion program. Kaplan and Davis[l] presented one of the magnitude and origin of added resistance of SESs.[2] devel- new experiments are still in progress at the time this pa. The seakeeping of an SES is quite different to the air cushion. higher pressure than the air cushion. The cobblestone effect causes SESs to have a poor ride quality. An air cushion is enclosed by the side-hulls. One The computed motions and cushion excess pressures or two internal webs restrain the aft side of the bag. Only very little literature can be found on the topic of Up to now Surface Effect Ships (SESs) or air cushion added resistance due to waves of SESs. held in Amsterdam. and published in RTO MP-15. the deck. diction method for motions and added resistance due to Kapsenberg[10] measured a large added resistance due waves. per had to be delivered. Some air will leak from the cushion through gaps component was believed to be the largest. The bag is open to ratory of Delft University of Technology. Most SESs have a bag-type stern seal. the sides where the seal plenum is closed by hulls. The first papers on motions of SES. Ehrenberg[9] states however that an SES has much design of large SESs requires therefore an accurate pre. the so called cobblestone ef- reasonable agreement with the computational results. In these days however both naval and civil oper. Steen[5] periments show that the added resistance of the air cush. a row of loops of tance that was measured by MARIN.[8] supported catamarans were mainly operating in sheltered presented a comparative study on the speed loss and waters. S0rensen[4]. system of fans. and Ulstein[6] published extensive studies of the high ion is not large. SESs appear SES has a slightly larger speed loss than the catama- to have a large speed loss when sailing in waves. This leakage flow is compensated by a the computational method are compared to experimen. Several authors presented studies on motions of Surface The aim of the new experiments is to get insight into the Effect Ships. Faltinsen et al. ators show an increasing interest in large SESs sailing in They did not however include the added resistance due open seas. 12-1 Motions and Added Resistance due to Waves of Surface Effect Ships Joost C. He also showed that the usual quadratic rela- Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". This discrepancy flexible material which are open to the cushion side. The remainder is carried by the buoyancy of the added resistance of the air cushion. Figure 1 presents a longitudinal cross section and added resistance due to waves of Surface Effect of an SES. The Netherlands. The results of under the seals. a tal results of MARIN and to results of new experiments bag of flexible material which is pressurized at a slightly which are carried out at the Ship Hydrodynamics Labo. the water surface and flexible seals at the bow This paper presents a computational method for motions and stern. Moulijn Delft University of Technology Ship Hydromechanics Laboratory Mekelweg 2 2628CD Delft THE NETHERLANDS Summary A Surface Effect Ship is a hybrid of a catamaran and a hovercraft. The new ex. to waves. 5-8 October 1998. The present paper does not deal with 1 Introduction the cobblestone effect. was the major reason for the new experiments. operability of a catamaran and an SES in a seaway. fect. The bow seal resistance is however much smaller than the added resis. The ran. dynamics of hovercraft and SES. thus agree well with the MARIN results. less speed loss in waves than a comparable catamaran. The computed added resistance only includes the cushion. Kaplan et al. Nakos et al. This added resistance hulls. The computed added dividing the bag into two or three lobes. . The result of the study is that the from the seakeeping of a conventional ship. This paper presents therefore Doctors[3] presented an extensive overview of the hydro- only some first results of these experiments.[7] showed that these high frequent mo- tions are caused by acoustic resonance of the air inside the cushion. Most of an SES's weight is carried by the air Ships.

F6 are the moments which act on experimental results. and . pc.772. 2 Computational method a force due to the air cushion and a force due to the seals. F3 are the forces Section 4 presents and compares the computational re. Each force is split up into the following components: a gravi- tational force. The forces due to seals and following waves as only heave and pitch motions are follow from the appropriate seal models.1 The equations of motion waves that was measured at MARIN does not however follow from the computations. tor if = (771.2 The equations for pc and ps Consider an SES sailing in waves at a forward speed U. f?3)T and a small rotation vector Ü — T (T/4. E : dt 2 =F i = i. 2.e. The excess These results are compared to the results of experiments pressures are assumed to be constant in space.776 The main goal of the present PhD research project is (i. which acts on the vessel in ^-direction. The large added resistance due to 2.?75..d2i}k The next section presents a description of the new compu. The hydromechanic force is computed by means This section presents the computational method for mo- of a three-dimensional panel method. Mjk. y-axis.. so acoustic that were carried out by Kapsenberg[l 1] at MARIN.7?6) . The force due to tions and added resistance due to waves of Surface Effect the air cushion follows from integration of the cushion Ships. roll. and sults. and the for motions and added resistance due to waves of SESs. equations: tance due to waves of SES... tions and added resistance due to waves of SESs. ps.. a propulsive force.F4. heave. y-direction.(in) 1 H fp\ i( T Q<fut) Q cushion X^^S„ ^ = ^ plenum Figure 1: Longitudinal section of a Surface Effect Ship tion of added resistance with wave height does not hold small motions are defined by a small translation vec- for an SES. of the generalized mass matrix. a hydromechanic force.12-2 ■. computational results are in good agreement with the MARIN experiments as far as motions and cushion pres- sure are concerned.. surge. The six small displacements. and z-axis. These for the additional unknowns pc and ps have to be formu- . These experiments can be linearized to the following system of differential give new insights into to the origin of the added resis. The vessel carries out small oscillatory motions which are Next to the equations of motion two additional equations superimposed to the translatory forward motion. The effects of the air inside a plenum cannot be resolved. Up to now the method is only suitable for head excess pressure over the deck. the vessel around the z-axis. F2. This Next to these unknowns two additional unknowns occur: paper presents results of a new computational method the excess pressure in the cushion plenum. 771. Nev- ertheless most of the theory is presented for six degrees of freedom. The effect of surge motion is neglected. the experimental results of MARTN and some new z-direction. excess pressure in the stern seal plenum.. pitch and yaw displacement) to investigate how one can accurately compute the mo. The propulsive force is balanced by the resistance of the vessel.. are the actual unknown motions that have to be solved. sway. Fi. Section 3 describes the new experiments where Mjk is the is the kth component of the yh row that were carried out at Delft University of Technology.( (1) tational method for motions and added resistance due to k=l waves of SESs. considered. Then been carried out at the Ship Hydromechanics Laboratory the equations of Euler for the motions of a rigid body of Delft University of Technology. New experiments have The motions of the SES are assumed to be small. F5.

Furthermore This process cannot be captured by a linear formulation.4. The be found in reference [18]. the seal will touch the water surface so the leakage area equals zero. The seal canvas is assumed to have no bending stiffness. These equations follow from conservation of mass one-dimensional analysis leads to good results: for the air in a plenum and the equation of state of the air in a plenum.4 Air leakage determine the bag geometry and the tension in the seal canvas: heave displacement. Q(m) most suitable for head and following waves. 2. it only transmits tension.3 The fan system This implies that the wave height is assumed to be con- stant in transverse direction. The following assumptions underlie the stern seal model. 12-3 lated.+Pa) dt Ws S ^ ' dt wave height at the seal is large. pitch displacement. Nevertheless duce the air leakage gap. Q(in) = Q(in) + Q(m) whjch ^^ sum of ^g fl()w trough the cushion fan and the leakage flow from the stern seal. Durkin and Luehr[12].5 The stern seal age flow under the bow seal and the leakage flow under the stern seal. When the bag age to be important and highly non-linear. Masset et al. This Steen[5] and Masset et al. The cushion volume Vc depends on the This subsection presents a brief descriptions of the heave and pitch displacement of the vessel.[14] a simple stationary is due to the absence of information on dynamic charac. In the present model and in the models of ing the present method. This air flow will result in a dy- box.[17]. He found that inertial effects in a dynamic way to oscillating back pressure. seal excess pressure. Lee[19] neglected this pressure all computational methods for seakeeping of SES. Ulstein[6] carried out extensive air leakage com.[20]. Nevertheless. and Qiout) = Q^6ut) + Q&"0 which is the sum of the leak. at the seal and mean wave slope at the seal.[20]. A more complete description can of the seals and the wave height in the air cushion. Ulstein[6] developed a seal model which in- Masset et al.[14] and Witt[15] show that fans respond cludes the inertial forces. leakage gap. Therefore the seal model is The air volume flux through a fan into a plenum. Lee[19]. K = cp/cv « 1. The equations forpc and ps read: where c. specific heats of air. air leakage is often linearized or even neglected. 2. The leakage area under a seal is highly non- Vs dps _ (n(in) n(0Ut)\ dVs linear in the local relative wave height. neglected. Masset et al. use static fan characteristics. one-dimensional modeling of the leakage flow is used. Vs is the stern seal relative wave height at the seal becomes smaller a leakage volume. The leakage areas under the seals follow from the (2) seal models.[13] show that the dynamic behavior of the fan does not touch the water surface air will leak from the has a damping effect on the heave motions of a hovering cushion under the seal. distribution. namic pressure distribution under the seal which will re- tant effect on the overall motions of an SES. . tational and inertial forces that act on the seal canvas are teristic. geometry and the tension in the seal canvas are known. which results in the concave the following formula which follows from a stationary cushion facing part of the seal. teristics of lift fans. When the seal van et al. the geometry stern seal model. For a given bag configuration the following parameters 2.[13]. Moulijn [16] also showed that the fan has an impor. pa is the ambient pressure and K is the ratio of gap will occur suddenly. Ai is the leakage area. Sulli. Steen[5] the seal volume and seal force can be computed easily. When the relative ■K(p. is the leakage coefficient which depends on the local geometry of the orifice. Figure 1 presents the system of plena Q{ ciAi (3) P and the air volume fluxes of an SES. cushion Many authors (see for instance Nakos et al. When the where Vc is the air cushion volume. are important for high frequent motions. includ. showed the importance of air leakage for the cobblestone The seal may either touch the water surface or leave a effect. He found that larger than the seal pressure. In the bottom case the cushion pressure is putations using a non-linear panel method. Ap Vc dpc dVc is the pressure jump across the orifice and p is the density K(Pc+Pa) dt dt of air.[14]) consider air leak. mean wave height McHenry et al. The gravi- is approximated by the linearized steady fan charac. excess pressure. Steen[5] and stern seal volume Vs depends on stern seal geometry.[14] presented similar seal models.[7]. The equation of state is taken to be the out) 2Ap isentropic gas law. and afterwards grow linearly. Sullivan et al. The model is two-dimensional in a longitudinal plane.

The part of the seal that is bent backwards at the water surface is neglected. A more elaborate description of the and Beukelman[25]. When an SES three-dimensional Rankine panel method. plenum. thus avoiding a complicated time step- ping algorithm and saving much computational time. Ä(Be)=Pc-(C6-C. tra terms which account for the momentum of leaking air should also be included. £s is the wave height at the stern on the hulls follow from integration of the pressure over seal and Bc is the width of the air cushion. Therefore the where the over lining denotes that the time averaged value theory of Cummins[23] and Ogilvie[24] is used to trans. and no air will es.12-4 2. the added resistance due to sinkage. so a larger part of the SES's around the undisturbed flow. should be used. The bow seal model is also two-dimensional (in called added resistance. The tangential derivatives of the unknown poten- tial which occur in the free surface boundary condition The resistance of the air cushion follows essentially from follow from analytical differentiation of a bi-quadratic the following equation: spline approximation of the potential. the seal will leave a gap above minus the resistance of the ship when sailing in calm wa- the water surface. the pressure and the wave height where pc is the cushion excess pressure. £(> is the wave can be calculated. Some ex- the wetted part of the hulls. The boundary condi- the seals. The seal is represented contribute to the added resistance are distinguished: by a flat boundary which roughly coincides with the fore- most part of the fingers which actually closes the cushion 1. the amount of air leakage increases be- is assumed to be incompressible and non-rotational. The hydrodynamic forces which act height at the bow seal. The added resistance due to sinkage is basically an in- The hydromechanical problem is solved by means of a crease of the steady resistance of the hulls. 2. The rest of it is els. Most authors consider the added a longitudinal plane).6 The bow seal form the frequency domain results of the panel method to the time domain. This leads to a Kelvin free weight has to be carried by the buoyancy of the hulls. The frictional forces that act 2. and air will escape from the cushion ters at the same speed. When the boundary value problem is solved. This results in a decrease of the mean excess tions on the free surface and on the hulls are linearized pressure in the air cushion. and a Neu- a greater resistance of the hulls. It tion method for the hydromechanical problem of an SES can be computed for instance by the method of Gerritsma sailing in waves. method can be found in reference [21]. Nakos[22] also uses this spline scheme. tion.)-5e (4) gularity distribution on each panel. which leads to terms due to the presence of the air cushion. The added resistance due to waves of a ship is the time cape under the seal.7 Hydromechanics The usual added resistance of the hulls is caused by the diffraction of the incident waves and the radiation of the waves due to the motion of the vessel. The hulls and a part of is partly caused by an increase of the frictional resistance the free surface are paneled with flat quadrilateral pan- due to the larger wetted area of the hulls. Again the mean wave height and resistance of SESs to be large. The modeling of the finger-type bow seal is much more simple than the modeling of the bag-type stern seal. the usual added resistance of the hulls. 3. For convenience it is henceforth plenum. the lower part of the fingers is simply bent backwards at the water surface. The surface boundary condition which includes some extra draft of the vessel will therefore increase. Each panel has a constant source an dipole distribu- caused by an increase of the wave making resistance. When the local deck height at the bow is smaller than the 2. but he also uses a quadratic sin. The water flow sails in a seaway. the added resistance of the air cushion. This added resis- This subsection presents a brief description of the solu- tance component is also experienced by normal ships. . This greater resistance mann hull boundary condition. on the bow seal are neglected. Then the added resistance due The unsteady hydromechanic problem is solved in the to the air cushion follows from: frequency domain. Three components which wave slope at the seal are used. When the local deck height is larger averaged resistance of the ship when sailing in a seaway than the height of the seal.8 Added resistance due to waves height of the seal. so cause the ambient waves cause large leakage gaps under potential flow theory can be used. The R("$ = R^'Hin waves) — i?(00>(calm water) (5) motions have to be solved in the time domain because of the non-linear cushion and seal dynamics.

as the experiments were a Computer. in fact they are only 12mm tance due to waves of Surface Effect Ships. The RPM of the fans is controlled by are not included in this paper. The hulls of the insight into the magnitude and origin of the added resis. The second version HYDROSES project. . HYDROSES project. added resistance due to sinkage and added attributed almost entirely to the air cushion. Otherwise the model would capsize. This version of the model will be referred to as both pressurized by axial fans. The other components are not likely to be large either. the diaphragm. a large collective research project The second version of the model is equipped with more on seakeeping of SES. 3. the model with hulls. and air from out- periments. The new experiments are still in progress at side the cushion will flow into the cushion. Loa. These hulls are usually very slen. still in progress at the time this paper had to be delivered. Therefore the forces that were distinguished by Section 2.1 The model is connected rigidly to the carriage. To 0. The usual added resistance of the hulls should there. Kapsenberg[26] presented a paper on computational and experimental results.745 m fore be small.050 m the larger air leakage flow. In the present experiments the main purpose of the diaphragm is to prevent very large pres- 3 New model experiments sure amplitudes which might lead to a negative cushion excess pressure. Therefore it is tested in a captive setup. An attempt is thick plates. The fans are kindly on loan from MARTN.000 m buoyancy of the hulls. This resulted in a reduction Pc 300 Pa of the added resistance by only 6 percent.8: added resistance that are measured during the testing of this version can be of the hulls. obtained from free sailing calm water test with a reduced The air cushion plenum and the stern seal plenum are fan RPM. Table 1 presents the basic dimen- of the model is also tested in a free sailing setup. Table 1: Main dimensions of the DUTSES model The computational results do not however indicate a large added resistance due to the air cushion (see Section 4). This enables comparison to the realistic hulls. This ensures a very stable RPM of the fans. This version of the model has almost no buoyancy. Section 4 presents and compares the of the air cushion. The goal of the new model experiments is to get more Two versions of the model are tested. Therefore new model tests have been carried out in order A flexible membrane is mounted on top of the air cush- to get more insight into the magnitude and origin of the ion. When the cushion excess pressure be- This section presents a description of the new model ex. 12-5 The added resistance due to the air cushion should be by far the largest contribution to the total added resistance.100 m creased the RPM of the fans in order to compensate for T20 0. The present computational method only computes the added resistance due to the air cushion. An sions of the model. this diaphragm technique. These experiments are a part of the of the added resistance of the hulls. He in. This version of the model will be referred The model was partly derived from the target vessel of the to as the model with plates. The difference between the results of this experimental results of Kapsenberg[26] that were car. which is effectively removed by first part of the results is included in this paper.200 m Only a small part of the weight of an SES is carried by the Lc 3. version and the version with plates provides an indication ried out at MARIN. This enables resistance of the air cushion. A diaphragm added resistance of SES. The next section describes the can be used to obtain a correct scaling of the stiffness new model tests. This membrane is called diaphragm.549 m der. This model will be referred to as indication of the added resistance due to sinkage can be DUTSES. The forces that act on these plates are very made to measure the three added resistance components small (in head and following seas). The fans are mounted di. Therefore only a very realistic situation. comes negative the seals will collapse. Moulijn[27] also presented a more elaborate discussion on the topic of scaling of air cushion dynamics. This is not a the time this paper has to be delivered. The results of the model with hulls rectly on the model. a measurement of the added resistance of the air cushion alone. Therefore the Ps 306 Pa major part of the added resistance must be attributed to the air cushion.525 m resistance due to sinkage is relatively small too. where the model 3. KapsenbergflO] showed that the added Bc 0. ^plates 0. first version are very slender. B hulls 0.

10 The model with plates is subject to two types of experi. It served free sailing experiments the heave and pitch displacement as a study object in the HYDROSES project. quency range the computations and experiments agree .22m wide and Pc 11. Next to this the hori. tional and experimental results for the HYDROSES target vessel.2 1. The RAOs follow from a harmonic anal- Presentation and comparison of ysis of the computed and measured time signals. Finally this section as a function of the wave frequency.4 0. the excess pressures in the cushion and stern seal seal. During captive experiments the vertical and horizontal The HYDROSES target vessel was designed as a forces in the oscillator legs are measured. presents the main dimensions of the vessel.11 kPa about 2. kA = 0. kA = 0. is Table 2: Main dimensions of the HYDROSES target installed at one end of the tank. It is equipped zontal and vertical seal connection forces are measured. The model carries out an imposed oscillatory heave or pitch motion while it is Figure 2: Heave RAO of the HYDROSES target vessel towed in calm water. This tank is 142m long. The computations were carried out on full scale.00 m of Technology. The forces in the oscillator legs are measured and the model is restrained in it's mean position while it is towed in waves. 4. The carriage can be vessel equipped with an oscillator for forced oscillation experi- 1.00 m u 144. with a tree-lobe bag-type stern seal and a finger-type bow Further.05 kA = 0.15 — experiment + ments: wave force measurements and oscillation experi- 0.00 m The experiments are carried out at the #1 towing tank of B 35. They manifest them self most prominently as This section presents and compares the computational sinkage and higher harmonics. In the higher fre- presents a discussion of the results for added resistance.4 to the carriage by the oscillator legs. Results for several levels of wave steepness are shown. Again the forces in wave frequency [rad/s] the oscillator legs are measured. as was shown in a previ- and experimental results.2 Experimental setup i-/pp 153. plena.85. In the experiments the model was equipped with a di- erence wave is measured. The model with hulls is subject to three types of experiments: wave force measurements. First it presents computa.6 0. while the model is towed in waves. Finally the ref. All results are presented on full scale. —i 1— ments. The carriage is suitable for speeds up to 7m/s. which can generate both regular and irregular waves. to the bow and stern seals are measured. During the oscillation experiments the model is also connected 0 0.30m deep.12-6 3. oscillation ex- periments and free sailing experiments. The effect of the non-linear cushion dynamics on the heave RAO is not very large. During the free sailing experiments the model is free in heave and pitch.2 0. A hydraulically driven flap-type wave maker.2. aphragm in order to obtain a correct scaling of the air cushion dynamics. Figure 2 presents computed and measured heave Re- sponse Amplitude Operators (RAOs) as a function of the wave frequency. During the wave force measurements the model is connected to the car- riage by the two legs of the oscillator. ous paper by Moulijn[16]. and it sails in head waves.01 kA = 0. 4. Then the computational and experimental results Figure 3 presents the computed and measured pitch RAO for the DUTSES model are presented.00 m the Ship Hydromechanics Laboratory of Delft University Bc 26. Table 2 and the resistance are measured.1 The HYDROSES target vessel During the experiments several variables are measured.8 1 1. The lift to weight ratio of the vessel is about 0. The com- the results putations and experiments agree well. the flow through the fans and the wave height close The vessel's speed is 45 knots.8- ments. while during car/passenger ferry by FINCANTJERI in Italy.

Figure 6 presents the measured and computed RAO for The Heimholte resonant frequency of the air cushion is the wave force in heave direction. but in the lower frequency range the computational resistance.72 m/s.05 kA = 0. The experimental results have not are in good agreement. The computational results also indicate a results are larger than the experiments.2 The DUTSES model Figure 4 presents the computed and measured RAO for the cushion excess pressure. but still not really give some interesting new insights into the topic of added significant.4 0.15 experiment 1. The computations and experiments computational results.4 0.8 1 0.6 0. The target vessel computational results are much smaller than the exper- imental results. been analyzed thoroughly yet. kA = 0. It is not clear non-quadratic relation of the added resistance with the whether the computational method over predicts the pitch wave height. and are presented on model scale. 1 I —i— 1 1 kA = 0. while in the low frequency with plates. In the high frequency range the experimental resistance of SESs. or whether the resonant frequency is shifted. The effect of the non-linear cush.01 kA = 0. aft very sensitive to for instance the behavior of the fan.4 resistance is proportional to the incident wave height wave frequency [rad/s] squared.10 kA = 0. All results are for the model version results are somewhat larger. There even does not seem to be any correlation between the computed and measured added well. The latter is towed in head waves at a speed of 3.15 located in the high frequency range.10 kA = 0.8 1 1.6 0. In the case of conventional ships the added 0 0.01 kA = 0. The pressure was measured This section presents preliminary results of the new at two locations: one at the fore side of the cushion and model experiments.4 0.05 E 400- kA = 0. resonance. The scattering of the experimental results in- dicates that this relation does not hold for an SES. The Figure 4: Cushion pressure RAO for the HYDROSES measured added resistance is negative in some cases. The discrepancy between the computed and the measured This discrepancy is probably caused by the relatively added resistance was the reason for the new experiments.6 0. The model range the computational results are greater. * t + + + + + ■K- ££± + + ? -200- kA = 0.01 + + 4- kA = 0.15 experiment + 1 i 1 1 0. This resonance is exp. Nevertheless these results ion dynamics on this RAO is larger. 4. discrepancy is caused by the large computed pitch mo- tions which cause air gap modulations under the seals.fore exp.10 + + -400. The effect of the non-linear cushion dynamics on the pitch RAO is negligible. MARIN measured added re- sistance values up to two times as large as the calm water resistance. 12-7 "T. Figure 5 presents the computed and the measured added resistance divided by the wave height squared as a func- tion of the wave frequency. although the scattering is much less severe.4 wave frequency [rad/s] wave frequency [rad/s] Figure 3: Pitch RAO of the HYDROSES target vessel Figure 5: Quadratic added resistance operator for the HYDROSES target vessel kA = 0.5. It displays results for . harsh linearization scheme of the hydrodynamic prob- lem. These results are also compared to one at the aft side.2 0.8 1 1.05 kA = 0. and is therefore difficult to predict computationally (see ref- erence [16]).2 1.

It was measured dur- that the heave force is quite non-linear in the wave height. compared to the total added resistance as presented by Figure 7 presents the RAO for the cushion pressure dur. Figure 8. model are of the same order of magnitude. A . Generally the correlation is better for higher added resistance of the plates appears to be relevant when waves.04 measurements^ is wave am- to wave height squared.04 x 2 3 4 5 6 wave frequency [rad/s] 2 3 4 5 6 wave frequency [rad/s] Figure 7: Cushion pressure RAO for the DUTSES model Figure 9: Quadratic added resistance operator for plates of the DUTSES model several wave heights. A . Both experimental and computational relation with wave height. The linear way. The other A = 0. The results are not di- tance of the plates. Contrary to the results for the plitude). This assumption is approved data indicate that the added resistance is not proportional by one of the A = 0. The added erators for the added resistance of during (captive) wave resistance of the plates was assumed to have a quadratic force experiments. by subtracting the added resistance of the plates from the ion and at the aft side. the air cushion with the plates is neglected. resistance of the air cushion. the results for the DUTSES prove a quadratic relation (see Figure 9).04 measurement does not ap- HYDROSES target vessel.02 + exp. ing wave force experiments without cushion pressure.0. The cushion pressure was The added resistance of the air cushion can be calculated measured on two locations: at the fore side of the cush. The correlation is not fully satisfac- tory yet. Figure 10 presents the thus calculated experimental added The computations do not include the usual added resis. The scattering of the experimental data indicate as the plates are extremely slender. This way of computing the added results for the heave force. The calm water resistance .12-8 experiments calculations 2 3 4 5 6 3 4 5 wave frequency [rad/s] wave frequency [rad/s] Figure 6: Heave force RAO for the DUTSES model Figure 8: Quadratic added resistance operator for the DUTSES model exp.0. resistance of the air cushion implies that interaction of Figure 8 presents measured and computed quadratic op. These results are similar to the total added resistance. The computational data appear to behave in a much more Figure 9 presents the results of these experiments. This added resistance should be small vided by wave height squared. ing wave force experiments.

In some cases an extra towing force was where the experimental results were calculated from the added because the model could not reach the desired seal forces and the cushion pressure. Fx\.02 A = 0. It This section presents a discussion on the topic of added would also explain the statement of Ehrenberg [9] that an resistance. pc is the ent ways.03 A = 0. (positive in forward direction. If this is the reason for the large added resistance. There are however many differences between is propulsed by means of surface piercing propellers. catamaran. The MARTN model was self-propulsing by cushion excess pressure. while the model was rigidly con- nected to the carriage during the new experiments.Fxb -pc-Ad. both experiments. Ad is the deck area and 775 is means of model water-jets. the water-jet.04 1 1 1 -i 1 1 1 r ~ i i experiments calculations 2 - + t _i i i i i_ 234567 234567 234567 234567 wave frequency [rad/s] Figure 10: Added resistance of the air cushion.775 (6) to the hulls.01 A = 0. This is not however the case. The added resistance of the determined from a propulsion increase of the water-jets. pitch and surge during the MARIN experiments. The . the experiments follow from the total added resistance minus the added resistance of the plates of the vessel was about 27 N. 4. 12-9 A = 0. and experiment. more simple and the air cushion agree quite well (compare Figure 10 with more reliable than the method adopted at MARTN. air cushion follows again from Equation 5. resistance follows from an increase of the towing force. The added resistance due to the air cushion can also be calculated from the horizontal seal connection forces.3 Discussion The large speed loss of SES might also be caused by a decrease of the thrust of the water jets. There is an obvious difference between the SES has much less speed loss in waves than a comparable experimental results of MARIN and the new experimen. The computed and mea. according to the author. The DUTSES model is always towed. which follows from a measurement of the flow through Figure 11 presents the added resistance of the air cushion. This would ex- plain the large added resistance measured at MARTN. as his experience follows from an SES which tal results. This is not very likely as the realistic hulls where Fxs is the horizontal stern seal connection force are still very slender. The results of both meth. The Figure 10). The MARIN model has realistic side hulls instead of The resistance of the air cushion follows from: plates. force experiment than during a comparable free sailing able agreement. cushion pressure amplitude is much larger during a wave sured added resistance of the air cushion are in reason. ac) the major part of the added resistance should be attributed Ä( -Fxs . ods for calculating the experimental added resistance of This method is. The added are in reasonable agreement. added resistance during wave force experiments. The model was free in heave. The added resistance was the pitch displacement angle. The added resistance was measured in two quite differ- is the horizontal bow seal connection force. Again the results speed. The magnitude is about the same. MARTN results might be affected by a decrease of the efficiency of the water-jet propulsors. hence the minus sign). Therefore one might expect to find a larger the trends do also agree.

1993. Cobblestone Effect on SES. [4] A J. and for [8] O. 1996. pages 211-247. hydrodynamics of surface effect ships. Schiffbau der Universität Hamburg. J. In Transac- ion pressure are concerned. Steen.12-10 A . The results of the computational method that was pre- sented by Section 2 are in good agreement with the ex.1981.04 T" i 1 r I i experiments calculations ■ to i i i i l "234567 234567 234567 234567 wave frequency [rad/s] Figure 11: Added resistance of the air cushion.D. and S. Ulstein. Nakos. [2] P. PhD the- sis. Bentson. periments.M. PhD thesis. Ehrenberg. and providing a lot of equipment for use in the new experi. R. Acknowledgments [7] D. AIAA Paper No. and ses in a seaway. that are in progress at the time this paper is written. The use of pressure distribution to periments. the added resistance due to an air cushion is relatively small. PhD thesis. Davis. 1992. [3] LJ.1974. PhD thesis. Minsaas.E. References [9] H. Faltinsen. 1991. Doctors. [5] S. Institut für [1] R Kaplan and S. Norway. The Norwegian Institute of Technology. 1991. Ulstein. A simplified representa. for added resistance are however much smaller than the added resistance that was measured at MARIN. Das Verhalten von Luftkissenkata- maranen (SES) im Seegang. Seakeeping analysis of surface effect MARTN for their financial support of my PhD-project. Speed loss and operability of catamarans ments. Dynamics and perimental results of MARIN. In First International Confer- ence on Fast Sea Transportation (Fast'91).01 A = 0. New ex. Kaplan. The added resistance of the hulls and the added re. In Intersociety High Per- which is in reasonable agreement with the computations. In AIAA/SNAME Advanced Marine Vehicles Confer- ence. Helmers. Arlington. Nonlinear effects of a flexible stern seal bag on cobblestone oscillations of an SES.B. formance Marine Vehicle Conference and Exhibit There were however many differences between both ex. 1993. Zhao. sistance due to sinkage will follow from the experiments The Norwegian Institute of Technology. Nestegäard. as far as motions and cush. A. The computational results tions ofSNAME. the experiments follow from the seal forces and the cushion pressure 5 Conclusions tion of the vertical plane dynamics of ses craft. Trondheim. S0rensen. (HPMV'92). . It can however be concluded that of Technology. VA. T. volume 89. which were carried out at Delft University model the hydrodynamics of air cushion vehicles of Technology. MARIN is also acknowledged for providing the experi- mental results for the HYDROSES target vessel.03 A = 0. The Norwegian Institute large added resistance. Davis. and P. ships.0.D.02 A = 0. Modelling and Control of SES in the At this stage it is not possible yet to say if SESs have a Vertical Plane. J. Trond- heim. 1995. In Fast Conference. K.74-314. lead to a much smaller added resistance and surface effect ships. [6] T.J. I would like to thank the Royal Netherlands Navy and Sclavounos.

78-756. [27] J. (FAST 93).. Yokohamajapan. 12-11 [10] G. In Proc. [17] G. F.P. Durkin and L. Lift fan stability for ses. 1993. E. [19] G. [18] J. Masset and J. ternational Conference on Fast Sea Transportation 1962. Gerritsma and W.P. On the motions of high speed surface- effect-ships in waves. Liibeck- Travemiinde. [16] J. [24] T.1994. A rational strip the- [11] G.K. 9(47): 101-109.1995. McHenry.C. F. Ship Wave Patterns and Motions by a Three Dimensional Rankine Panel Method.C. Morel.F. Blume. 1995. Washington. Delft University of Technology. RINA. Dynamic response Technical Report 169s. Delft University of Tech- nology. E. Liibeck- Travemiinde. Third International Conference gineering. LUehr. [26] G. Netherlands Ship Research of lift fans subject to varying backpressure. Ship Hydromechanics Laboratory. 1990. In NAV'94 pro- ceedings.M. [21] J. surface effect ship (ses) air cushion dynamics: An innovative methodology for theoretical modelling validation. Seakeeping behaviour of a ses [23] W. and ship motions. . Delft University of Technol- ogy. Scaling of aircushion dynamics. Germany. scale effects. In HPMV '92 Conference.J. Moulijn. Second In. NAV'94.K. Kapsenberg. Non-linear motions of surface effect ships. University of Michigan. 1997. I. Tech- nical Report 1151.5-7 October 1994. Moulijn. manufacturing and results. Hydrodynamic analysis of surface effect ships: Experiences with a quasi-linear model. on Fast Sea Transportation (FAST'95). Hinchey. Morel. [14] J. P. Schiffstechnik. The impulse-response function in different wave directions. Journal of Hydrospace Tech- nology. 1969. In International Conference on Air Cushion Vehicles (ACVs). [20] J. Large Ship Hydromechanics Laboratory. A model for flexible bag stern seals.1978. Kapsenberg. Sullivan. AIAA Paper No. A test rig for the anal- yses of a large surface effect ship seals dynamics: design. In Center TNO.C. Korbijn. Kapsenberg. 1998. Trondheim. O. Cummins. Added mass and damping coef- ficients for a large ses including an appreciation of [13] P. Kapsenberg and P. Masset. namic responce of an air cushion lift fan. Tech- nical Report 1051-O. and G. Motions of surface effect ships. 1991. Germany. In Fast Conference. Technical Report 1082-O. [12] J. Department of Naval Architecture and Marine En- tios. In Proc. Technical Report 013. Tuck.A. 1996. Japan. 1(2): 13-35. Model tests for ory for ship motions -part i. T. Yokohama. [25] J. 1993. Ogilvie and E. [22] D. Beukelman.C. J. and A.K. [15] K. USA.Witt. Dy. London. volume 1. Moulijn. In Second In- ternational Conference on Fast Sea Transportation (FAST93). Gosselin. Ship Hydromechanics Laboratory. volume 1. Moulijn.C. M. Kaplan. and MJ.K. Lee. AIAA/SNAME Advanced Marine Vehicles Confer- ence. Analysis of the resistance increase in waves of a fast cargo ship. Nestegärd. Third International Conference on Fast Sea Transportation (FAST 95).F. In Proc. 1995. 1997. 1992. In Proc. Nakos. PhD thesis. F. a large surface effect ship at different scale ra.

=0. Since the lowest investigated subproblems. The influence of main parameters characterizing the stern seal bag is Uy .4 m. [2]. represents a comfort problem for passenger Head sea. . This SES. air leakage underneath the seals and through louvers has a node approximately midships. 45 knots. Fig. [1] demonstrated the importance of the spatially varying air pressure |ncjdent waves lowest eigenfrequency. [2]. 1 Full scale measured spectrum S (f) of A problem with a Surface Effect Ship (SES) is high vertical accelerations at bow of 35 m vertical accelerations in small sea states. It is demonstrated that both the spatially varying pressure underneath the flexible stern seal bag and the impact between the bag and the water are important for the vertical accelerations of the vessel. 10 1. The phenomenon is often referred to as the cobblestone effect and is a resonance effect due to approximated by a sinusoidal function. 2 gives an overview over resonance phenomenon occurs at high frequencies relative to the resonance frequencies for the rigid body Air flow Heave & Pitch accelerations motions of displacement ships of similar length. The accelerations. wetted length and hydrodynamic [5] and [6] give simplified introductions to cobblestone pressure oscillations. Ref. Faltinsen Department of Marine Hydrodynamics Norwegian University of Science and Technology N-7034 Trondheim. 2 Investigated physical effects. 13-1 COBBLESTONE EFFECT ON SES Tore Ulstein and Odd M. The wave length is roughly twice the ship length. ' Resonant spatially uniform and nonuniform dynamic cushion variations are then important. and published in RTO MP-15. [3] and [4] showed that Hydrodynamic loads on bag structure : acoustic standing wave effects must also be included. It can be Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". Important damping eigenmode is constant in space. it affects mainly heave mechanisms of the cobblestone oscillations are due to accelerations. Flexible stern seal bag. f=frequency. The second eigenmode. [2] documented by full scale measurements Fig. transportation. «Hz) discussed. Due to the frequency of encounter effect there are 1-D wave equation: air cushion pressure waves with sufficient energy in small sea states that Flexible bag motions Air leakage: excite these resonance oscillations. The Air flow into bag Fan flow two lowest resonance frequencies in the air cushion of out of ■*"T*k« a 30-35 m long SES are approximately 2 Hz and 5-6 Hz. The eigenfunction leakage height and for the dynamic air cushion pressure is constant in outflow air velocity space for the lowest eigenfrequency and represents acoustic wave resonances for the higher eigen. Air leakage: leakage height/outflow air velocity frequencies.. 1). 5-8 October 1998. which the air flow into the air cushion through the fans and the corresponds to the lowest acoustic resonance frequency.. //. The Netherlands.3-0. A nonlinear time domain solution is used. INTRODUCTION Fig. The nonlinearities are mainly due to the flexible stern seal bag behaviour. The modal the compressibility of air in the air cushion. onboard a 35 m long SES the importance of the two lowest eigenfrequencis (see Fig. Norway SUMMARY Sa«) • Wave induced vertical accelerations (cobblestone [(m/s2)2s] oscillations) of a SES in small sea states is studied. It means cobblestone effect is excited because the water waves that the second eigenmode affects mainly the pitch dynamically change the air cushion volume. held in Amsterdam.

a simplified but rational -pdtydt. the normal velocities at the stern validation is needed. describes the air flow by the wave equation Three pressure sensors were used in the air cushion and one accelerometer on the vessel as part of the ride f* 2 . and the leakage below the air bag. A added mass. The air cushion is then Fig. it will have a negligible effect on the acoustic resonance mentioned above.vssVc|> 0 (3) control system. For encloses the air cushion volume. The nonlinearity in F™ is mainly caused by nonlinear effects related to the flexible stern seal bag. inside of the bag and the air cushion. Here the X. F/1 is the linear hydrodynamic excitation force on the side-hulls and F™ is the nonlinear force acting on the SES due to integrated unsteady excess pressure in the air cushion in direction i. The give the correct signals to the louver system. First a global flow model is studied. In order to linear dynamic pressure pc can be expressed as construct a ride control system.1 Global Air Cushion Model The air cushion is analyzed in two steps.t) • n on the surface T that The cobblestone oscillations do not Froude scale. The normal vector n instance resonance periods are approximately to T is positive out of the air cushion. air leakage and the velocities of the bounding surfaces are of the air cushion is used. The opening and closing of accounted for. approximated as a box. and goes through the center of gravity (CG). Full scale longitudinal direction. and seal regions are averaged over the height of the air cushion. while it has its moment of inertia around the 7-axis of the SES. reason is simply that the acoustic pressure component M is the structural mass and L is the structural mass h has a small amplitude midships. maximum absolute value at the ends of the cushion. The Finite Element Method has been (M +A3*)f13 +533f|3+C33TI3 +A3A5fj5 used to solve the problem. The placement of (2) the louver system is essential. For instance at the stern of the air cushion 2. The hydrodynamic loads on the hull are expressed by strip theory ([8]). The global flow is later connected to a the vent valves control the air flow from the air cushion detailed model around the bag. vs is the velocity of for the behaviour of the SES. The t is the time variable and dot stands for time derivative. This makes sufficient to consider a one-dimensional flow in the comparisons with model tests difficult. THEORY there is a combined effect of the flow between the Head sea longcrested waves are assumed. mathematical model is needed ([4]). normal velocity V(x. The boundary condition at the free water velocity. 3 shows a proposed ride control system by [7]. damping and restoring coefficients of the louver system consisting of two vent valves in the front two side-hulls. By properly filtering the signals from dt the measurement units and using a mathematical model Here $ is the velocity potential. 3 Ride Control System. The motions velocity U of the SES and is fixed to the mean of the air bag have an important effect on the lowest oscillatory position of the vessel. Air flow through fans. Ref. A right. This is not an important part of the cobblestone analysis. [7]. Bjh and Cy are hydrodynamic infinite frequency Fig. Ay . surface follows from the kinematic free surface The F-axis points to the port side and the Z-axis is condition. 2. This study focus Formally we can write that dtydn is equal to the on the nonlinear effects of the flexible stern seal bag. The equations of boundary condition at the rigid part of the wetdeck coupled heave Cn3)and pitch (T)5) motions are follows from the heave and pitch velocities. Eq. Both goes through CG. The upwards. Boundary conditions have to be imposed. The normal formulated in the time domain as velocity at the fan is expressed by the steady fan characteristics. Since it is proportional to the ship length. For instance if the louver +55*VC53*l3 = FfryFf® system is placed midships. (3) is combined with the (1) boundary conditions and written as +B£t\5+Cf5T\5=Fr\t)+F3ac(f) .13-2 (/5+A^)tj5+fi5*f|5+C^Ti5+A53fi3 that are part of a ride control system. Only incident waves are considered. The global model and cause damping of the cobblestone oscillations. the control system can sound in air and V2 is the Laplacian operator. the motions of the handed coordinate system that moves with the forward air bag. acoustic spatially varying natural mode of the air axis points upstream in the direction of the forward cushion. is used.

[9] verified the numerical method by comparing with the Finite Difference Method and by d\ + (T0+EA)dr]tEA using an analytical solution based on mode MJ\n=T0. 13-3 pvW 2 2 (4) approximated by 3 cable segments with constant radius of curvature R.d$ segments 1 and 2 are constant and equal. A linear variation is used over segment are finally written as each element. ds2 R ds R 2.. between the pressure and the air density. In the static case the bag is not touching the difference pressure across the cable.r) = Sa.){<b(t)} where {$(t)} is a vector cable segments are linked together with the boundary containing the node point values of the velocity conditions. AP02=AP03. The dynamic contributions from gravity will appear as restoring terms and can be neglected r]t(s. 4) and APu is the unsteady plane. * ' < difference pressure over each segment. (11) ~Ph / <&jds+fp(s>'nn>t)$j(s)ds .2 Local Air Cushion Model at the Stern MJ\=EA *\_EA*\n (7) It is now focused on the details at the stern bag seal. There are fNT(^-V(x.+Cjfi. respectively and integrated over the length Lcs of all the cable segments. The structural water. n„+AP„ (6) superposition. The two equations are multiplied with <j>J and cf>j. A ds2 R ds 2-D problem is solved in the longitudinal plane of the SES. The bag structure is therefore modelled as a weightless but not massless Here aß) is the principal coordinate of vibration mode cable. The first term in Eq. (6) and (7). The studied bag geometry is a typical 2-loop bag number i and ({>" and $' are the mode shape functions configuration (see Fig. We can write procedure.t) ■ n)dY = 0 J dn equal static difference pressures across segments 2 and 3. 4 Static geometry of 2-loop flexible bag seal. (4) is rewritten by Green's The equations of motions for the flexible bag are theorem as derived by a linear perturbation of the static solution.t) = Zfl. The aerodynamic pressure force acting on the bag is an order of magnitude larger than the gravity Tl„(5. A finite number of modes is used. the equations of motions can be 1-D elements in the longitudinal direction. but the leakage height underneath the bag is boundary value problem is solved by "dry" mode assumed to be initially zero in the time domain solution superposition. Here NT is the transposed of the trial function matrix N. This implies spatially uniform static v. are the motions in the transverse and the bag. The two coupled equations of motions are added and give M/i.-WtiW (9) compared to the tension terms. Eqs. = [APtfds (10) where segment 2 A Fig. Compatibility gives the relationship between tension fNT^^dVc=fNT^-dT-fWNT-V^dVc (5) and the motions in the transverse and the longitudinal vc r " vc directions. The bag s is the longitudinal coordinate along the cable structure is modelled as a cable in the cross-sectional segments (see Fig. 4).(0<j>» (8) force. It is assumed set up separately for each segment. (8) and (9) are substituted into Eqs. This pressure can be related to the pressure in longitudinal directions. Since the radius of curvature R is constant The analyzed domain is broken down into Nel small for each cable segment. Afterwards these that <$>(£„t)=N(E. A is the cross dimensional area of unit compressible and an adiabatic relationship is used width and T0 is the static tension in the cable segment.e. The static bag geometry is in the transverse and longitudinal directions. i. The air inside the bag is assumed the material. The pressure is assumed spatially constant inside Here nn and n. Mm is the structural mass per the air cushion by using continuity of mass flow in and unit length of the cable. E is the elasticity modulus of out of the bag. The equations of motions for a cable potential of the air flow. The tensions in .

To get a correct representation of the equation.yXx(x) on y=CW (12) atmospheric pressure at the trailing edge. the equation that is satisfies no normal velocity on the last element on the lip is exchanged with an equation that prescribes <&y(x. is the length of bag segment number i. An outflow velocity can be found from the jet contraction coefficient by mass conservation. The potential on the free air surface can now be found from Fig. 5. The unknown coefficients in this function are determined from the approximation of the jet contraction coefficient and the fact that the tangent to the free air surface is equal to the tangent of the bag at the lip. When this velocity is known. We assume the tangential velocity and aerodynamic pressure are first the bag is not touching the water and p=0. 5 Air flow underneath stern seal bag. The assumption of quasi-steady flow and inviscid air may also then be questioned. Further ß=log|Jc. 5. since now both the velocity potential at the outflow Here £(*) is the air surface coordinate defined in Fig. The velocity potential (J) for important wave lengths in the acoustic air cushion the incompressible air flow inside the fluid domain is problem is much longer than the cross-sectional length represented by Green's second identity. surface and the tangential velocity at the free air surface The subscripts x and y denoted the x. It expresses that We will show how pml and p are evaluated.y)=vfl on y=CW (13) is dependent on the bag deformations. the . The horizontal uniform inflow problem is analyzed. The velocity potential water surface and is pointing towards the stern of the and the normal velocity are assumed constant over each SES. The flow is assumed quasi.?i)|M-cK3^^W) (14) dn{x) dn(x) However. the dynamic free surface condition defined in Eq. A 2-D continuous at the lip. The solution procedure is as follows. 27tcj)(x1)= I ß(f.. There are no point of the bag and the free water surface. i. viscous effects are neglected. the local Mach number is high. This implies that the tangential velocity at the separation point must be equal to Va (seeEq. The air is in viscid. 5. p(s. Since the constant through the air jet. pcal is the unsteady spatially varying air The subscript s denotes the tangential derivative at the cushion pressure underneath the flexible stern seal bag. The unknowns are the normal derivative of the velocity potential on the free air surface and the velocity potential on the inflow and the wall surfaces.t) is implies that the unsteady hydrodynamic pressure. Since Here S is the surface that encloses the air flow volume. When the field point used. The outflow velocity Va is assumed steady and incompressible. The *-axis is parallel to the undisturbed free discretized into straight elements. the problem is hydroelastic. Since the hydrodynamic pressure <M*.13-4 Here pb is unsteady pressure inside the bag. The kinematic condition on the free air surface velocity potential on the lip of the bag. The lowest point of the bag is domain. A set of linear equations follows from the potential (|) of the local air flow satisfies the 2-D Laplace discretization. are known. Green's second identity can then be used to The dynamic condition states that the pressure on the find the normal derivative of the velocity potential on the free air surface is equal to the atmospheric pressure.e.-J?| . The velocity element. an integral equation defined in Fig. This surface is described by an exponential function in the first iteration. This information is used as a basis velocity is assumed known and constant over the height for a 1-D approximation. of the air cushion. L(. A Kutta condition is prescribed at the separation point (lip).\. normal velocity on the "walls". First a jet contraction coefficient ka=^(x-~^)lhL (see Fig. the assumption of incompressibility is consistent from that point of view. the separation point of the The normal vector ft to S is positive into the air flow air flow must be chosen. 5) is guessed. dimensions of the air bag. (13)).y)=$x(x. To establish this equation the local behaviour of the tangential velocity of the air flow near the lip is used. The geometry is surface. A free air surface is generated based on this guess. when the air gap becomes very small. Origo is placed at the steady free water of Fredholm type is obtained.and ^-derivatives. The geometry and the coordinate system are approaches a point on the boundary. (13). free air surface and Va is the constant air flow velocity This is present when there is a gap between the lowest at the free air surface as indicated in Fig. The y-axis is pointing upwards. This free air surface.

and is not updated by the algorithm described above. incident irregular wave potential and <$>(x. The slope of the first (y=0) defined above (see Fig. 13-5 kinematic free surface condition is used to generate a force defined by the last integral on the right hand side of new free air surface (see below). The modal hydrodynamic hL0+\J2c(t)j)<( -|. Gravity can be neglected. .y.t) where 4>/x. This implies that the immersion of the bag is low. Since a more simplified method is beneficial in a ride control system.t) is the velocity potential for the water flow caused by the bag.f) is the Va is found by setting p equal to pc upstream. The water surface is analyzed as a water entry problem.. 1 %).. The air pressure is found from Bernoulli's $=0 equation. The lowest point of the bag is chosen.n ). (12) is rewritten into the following iterative scheme undisturbed water flow and the y-axis is positive upwards.nj/n2. The hydrodynamic loading will vo(*. $(x. will not be affected by the detailed behaviour of repeated until there is no updating of the free air surface the flow around the spray. The body boundary conditions are transferred It is used that £ =.t) = -Ux+tyj(. One incident waves are considered.y. hL is the value of h at the lowest point of the bag. The jet-contraction coefficient ka velocity potential is defined as is determined a priori by a 2-D analysis for a given bag. negative x-direction relative to this coordinate system. and the procedure is Eq. 6).y. relative vertical velocity between the bag structure and At the wetted surface of the bag the following vertical the water surface. that ensures a smooth flow at the separation HBVP point of the bag. ^x=nl^n+n2^s and to a straight horizontal line that corresponds to the x-axis (t.0-Ti3(r)-|tis(0 (18) . The origin is fixed at the lowest the free air surface.g. N. . the kinematic free surface condition defined in positive towards the upstream direction of the Eq. dynamic free surface condition is approximated by (j>=0 assuming a large forward speed of the SES relative to the on x < 0 and x > 2c(t). This together and a horizontal line defined by y=0. Here x\bag describes the unsteady geometry of the bag. It has L -Ut 0 7 2c(t) 0 / 2c(l) been found that the pressure distributions can be well 3i()/ay = v0(x. To obtain a stable and accurate point of the static bag configuration. Now both the velocity potential and its Simplified HBVP normal derivative are known on the whole enclosing (3/at. The total 2 element on the free air surface is set equal to the slope of the last element on the lip.x. (15) Since potential flow is assumed. Water impact loads (j) is found by a quasi-steady approach in the simplified The behaviour of a planing bag bouncing on the free Hydrodynamic Boundary Value Problem (HBVP).5PflVfl2 AW Fig."-C/4>„" must be determined a priori. The undisturbed free stream velocity U is in the Ji. free surface and the bag. The method has been verified against analytical $=0 <)> = 0 $=0 solutions based on conformal mapping ([9]).t) described by 8<t>/ay = v0(x. Here 2c(t) is the wetted length.o=vfv^ -u.dx -Zl dy (17) be much larger than the aerodynamic loading due to the relatively low pressure in the air cushion and inside the bag.v=n2<l)«~ni<i)t • Heren=(n. The x-axis is solution.. The vertical global motions of or that the error estimate obtained from the conservation the bag and the effect of the vertical motions due to of momentum is below a chosen value (e. An approximation with the free surface conditions leads to a square root of the wetted length is found from singularity in the hydrodynamic pressure at the spray root in the planing problem.t) P(*)-Pa _l{KhL)2 (16) 0. The wetted velocity is prescribed length of the bag will vary strongly through the impact 3ri 3*/ on the free water surface. N.y.t)+ty(x. (11). The body boundary conditions can therefore be that is the vertical distance between a point on the bag transferred to a straight horizontal line. A right-handed local xy- important point is how the kinematic free surface coordinate system that moves with the forward speed U condition is rewritten into an iterative scheme to update of the vessel is used.u a/ax) 4> =o surface. 6 Hydrodynamic boundary value problems Here h(x) is the local vertical distance between the water (HBVP) for water entry of bag.. 1-D approximations have been studied.y. the separation point 4>. The reason is the Kutta condition.

) 2m pressures have a similar behaviour. These frequencies presented first. 7 =1. CG and FP.0 and 1 at AP. Numerical studies by [9] showed that the errors using the simplified solution are not important.5 m forces and moments on the side-hulls are negligible. A modified Pierson Moskowitch wave spectrum Fig.00104 . The peak period is chosen so that there are significant vertical accelerations in the with many peaks. L is the length of the air cushion and the flexible bag is located at X= -LI2. Head sea. 7 shows the unsteady air cushion pressure when the with peak period 7 =1. 7p=1.) AP (no hyd. except that the 8m accelerations around 12 Hz is negligible. Fig. so the correspond to the two first spatial pressure resonances importance of each effect can be quantified. Around 12 Hz there is significant response also at CG. 6. 8. Air of a SES are studied at AP. 8 Spectra of vertical accelerations. there are two main spectral peaks organized so that the simplest simulation model is at approximately 6 and 12 Hz. The SES Head sea.) dp Linear bag fan slope (two fans) -0. This represents a more complicated procedure. This is caused by amplification and frequency domain around the first spatial pressure cancellation effects in the volume pumping of the air resonance of the air cushion. RESULTS Unsteady air cushion pressures and vertical accelerations Fig.[Hz] 3.Hm=0. A more correct way to define the HBVP is shown in Fig. Main particulars of SES box is equal to -1. 7 Spectra of air cushion pressures.l5 m. f. excit. . 8 Air cushion beam (Bc) illustrates that the linear hydrodynamic wave excitation Mean side-hull beam 1.[Hz] Longitudinal position of fans (Xf) 12m Fig.0 m Mean air cushion pressure (pc0) 510mmWc Mean fan flow rate (two fans) (Qj0) 2. The spectral are presented in terms of response spectra as a function of values around 6 Hz are nearly the same at AP and FP and nearly zero at CG. Rigid stern seal. The results are effect is disregarded. Air cushion length (Z.8 s. The results of a rectangular box. Air Forward speed (U) 23 m/s leakage neglected. Head sea longcrested waves are assumed. 1 and -1 at AP. This spectral period stern seal is rigid and air leakage underneath the seals is neglected. Different effects are then added.8 s is used.) 28 m Vertical accelerations at AP. The spectra have an oscillatory behaviour is realistic for small sea states. CG (two fans) c^)S m2s/kg - - FP (no hyd. Here AP is leakage neglected. Mean side-hull draft 1. The reason is that the second eigenfunction for acoustic Weight (M) 140 000 kg resonance is equal to -1. excit. If the cancellation energy at the lowest eigenfrequency. CG and FP.00063 m2s/kg dp '.\5 m.8 m2/s . [10] and [11] solved this HBVP and did also include pile-up of water upstream the wetted part of the bag.13-6 Here hw is the vertical distance between the lowest point of the bag and the free water surface at t-0 and £ defines the incident wave profile. eigenfunction for acoustic resonance in a rectangular Table 1. Rigid stern seal. CG at X=0 and FP at X=LI2.8 s. The accelerations and the air cushion Air cushion height (//. is described in Table 1. located at X= -Z72. with length L=28 m. FP • AP Linear air cushion fan slope -0. The reason is that the first frequency of encounter. CG and FP are presented in Fig. CG and FP. There is negligible wave cushion due to the incident waves. Pile-up of water is neglected. Hin=0.

Since rL is less dependence e e and neglecting damping.=1.e"""' is the velocity potential of the air flow in the air cushion due to the motion of the bag. 8. dVC dVc dp _ *C These mode shapes are the first four used (see Eqs.3 kg/m. the absolute value of the ratio is less than 1. some differences are observed. eigenvalue problem can be defined Fig. The solution of the 1 Slate si :Nv^ i Slat* snapo eigenvalue problem is t|r. This is a well known result for standing = 5500 Pa.. This can be expressed stern seal bag. 8 is compared Second one notes that the spectral peaks of the vertical with Fig. the flexible bag is deformed pressure and the air cushion pressure at the bag. 10 shows that the flexible bag increases the accelerations in the region just above 10Hz. T =1. cosrx' where wc=vgr and ' ' ! 3 3 |— ■ Mo *. the resonance frequency M =4. This leads to an equivalent horizontal velocity at shape function of the pressure waves. H1/3=0.8 s. 10 Spectra of vertical accelerations. =f|r. Static cushion. dVJdpc can be estimated by a .| \'ZL Maae » hape4(< ■"5™K tanrL+^ ^L=0 (20) HficL dPc :X ^i_y . This approximately 6 Hz down to about 5 Hz. (8) u'e 'H(Bc dt dp dt and (9)) to describe the deformation of the flexible stern .. When dVJdpc is zero. Bag length Lfc=3. the following than 7i. Since height #.u. Head sea.4m.15 m. the static and the first four mode shapes of as a 2-loop flexible stern seal bag are represented in Fig. dVJdpp-0 for a flexible stern. then the variation in the height of the air cushion.. e ' is the dynamic pressure in the air cushion at x'=L. the ratio of specific heat of air Vertical accelerations at AP.6-106 N. pc0 is the steady mass density of air in the air 8 10 12 cushion. If Fig. * S 19 |Hi|) —^ static analysis of the bag. Hc and Bc are the cushion height and beam.. 10. Air volume Vcdue to bag deformation. First one accelerations at AP and FP in Fig. Assuming harmonic time of tyhsp between AP and FP is cosrL. The air leakage underneath the seals is neglected. due to the assumption of uniform pressure over simplified solution of i|/. at x'=L is a longitudinal velocity u 'averaged over the . 13-7 In order to illustrate possible deformation patterns of the cross-section of the air cushion.a seal bag structure. 9 Static configuration and first four mode relation of the acoustic pressure waves inside the air shapes of flexible stem seal bag. The ratio illustrate this. may be explained by the mode pressure. *. Bag pressure waves in a long tube with rigid ends. is reduced relative to rigid ends.™.^^^ Here r is the wave number of the standing pressure waves in the air cushion. CG and FP with a flexible ys and the steady excess pressure pc0 in the cushion.* • • . 10 are reduced with notes that the main spectral peak is shifted from approximately 40% and 20% relative to Fig.JI-. It can be evaluated by leakage neglected. Here pc = -pc0my|j.Bc. dVJdpc is a measure of the change in air cushion Fig. AP02 n=l. phenomenon. ^ b £4=0.Hc. If the previous the bag. 10. A simplified model can the longitudinal direction is given by cosrx'. = 1« (H*[)|-v^ «P. This is due to the reduction is explained by the flexible behaviour of the flexible behaviour of the stern seal bag.16m. The due to the unsteady air cushion pressure so that the differences between the vertical accelerations at AP and deflected volume is in phase with the unsteady air cushion FP presented in Fig.2.. stern seal bag are presented in Fig. This -^ 0 increase is mainly explained by the volume change of bsB-\s t e^bsp ™bspjcx the air cushion due to the deformation of the flexible at X =Q 1W = ° ' (19) stern seal bag. is used. .L/vs=nit. difference pressures: AP01= 500 Pa. (20) is the dispersion Fig.bspj' at x'=L dpc Hßc ■FP AP CG Here i is the complex unit and i|/fa . 10. 9. Flexible stern seal noting that the right hand side of the boundary condition bag. Eq. rL=co<. The mechanism stern seal bag and the coupling between the bag may be explained as follows. The actual value depends on L.aPs.

11. a small increase in the immersion of the bag results in large increase in the wetted length.13-8 The effect of a constant leakage height underneath the When vertical accelerations presented in Fig.#1/3=0. Constant leakage heights: 0 m (bow). lHzl Fig.0Nm2. The reason is important reason is reduced damping due to smaller additional damping caused by the air leakage underneath leakage height underneath the flexible stern seal bag. These values of the static air leakage height are assumption of a linear elastic model. 10 around the first the first spatial pressure resonance frequency. the stern seal bag.02 m (stern). variable leakage height underneath both the bow and stern Water impact on bag. Head sea.8 s. Improved convergence was achieved by introducing a bending stiffness term -EId*r\Jds4 on the right hand side of Eq.#. #1/3=0. On the other hand the spatially varying air pressure is in a way forcing the bag structure to reduce the leakage height underneath the bag. Variable air leakage.=l. Head sea. 13 Spectra of vertical accelerations. underneath the seals is dependent on the relative motion between the seal and the water surface. These used through the remaining part of this section.1 m before air leakage at the bow deformations of the bag structure and not contradict the occurs. Hm is reduced from 0.s. An spatial pressure resonance frequency. Many modes are needed to describe the dynamic behaviour of the bag during water impact. This causes large impact loads that punch the bag structure out of the water. larger than 3 Nm2.1 m and 0. 10 and 11. seal are presented in Fig.15 Fig.„=0. Spatially varying Vertical accelerations at AP. Spatially varying pressure controlled that the global response was independent of underneath flexible stern seal. Head sea.15 m.1 m. . Variable £7=0 were used. '. 12 Spectra of vertical accelerations. 0. deformations are mainly caused by the interaction between the spatially varying pressure underneath the flexible bag and the hydrodynamic impact loads from the water. CG and FP accounting for pressure underneath flexible stern seal.. It was air leakage. one observes an increase of The vertical accelerations in Fig. El was set equal to 4. Since the bag structure is close to being horizontal in the impact region. The effect of air leakage accounted for. The hydrodynamic impact forces from the contact with the negative static leakage height at the bow means that the water.Hm=0A5 m. 13. compared with Fig. (6). The spatially varying pressure underneath the bag forces the flexible bag to reduce the leakage height. 12.15 to 0. 12 are stern seal bag can be studied by comparing Fig. 11 Spectra of vertical accelerations. The static leakage height (hw) is set equal underneath the seals is included together with to -0. There is also a change in the excitation of the air cushion pressure caused by the flexible bag motion. .r. Flexible stern seal bag.10 m in these relative motion between the bow skirt and the water calculations. The mode shapes corresponding to Fig. This is necessary to avoid too large surface must be 0.8 s. this particular choice of El as long as El was small and 7=1. The spatially Vertical accelerations predicted by the complete model varying pressure underneath the flexible bag is also are presented in Fig. 11 are decreased with approximately 85% in the vertical accelerations around approximately 10% relative to Fig. The variable leakage height r =1.02 m at the bow and the stern.

Full water impact. #1. an A main reason is that the spatially varying pressures analysis is made with the same model and data as caused by the air leakage reduce the leakage.15 m.3=0. (see also Fig. 7>1. An important parameter from resonance of the air cushion (=5 Hz). wave loads are negligible in this region due to the The complete model is used.4 (mls2fs at#1/3=0. simulation model. 13-9 To evaluate the effect of the hydrodynamic impact flexible stern seal bag is "following" the water surface. One possible explanation is that the accelerations motion of the stern seal bag induced by the water impact affects the air cushion pressure similarly as a piston at the end of a long tube and may in this way cancel a part of the Location: AP pressure variation in the air cushion. Spatially varying pressure bag configuration as function of time. The air leakage introduces damping and therefore reduces the vertical accelerations. it illustrates that vertical accelerations are nonlinearly Fig. 14.4 (mis 2)2s at at AP for three different height to length ratios of the #„3=0. 12 and 14 are obtained by the same physical model. The previous analysis has in particular focused on the largest differences between Figs. 13. 15 Airgap underneath lowest point of static air leakage. Data and model accelerations at FP at the highest peak of the spectrum as in Fig. 16 Spectra of vertical accelerations at AP dependent on Hm.6 Hz. = 0. 12 shows instead a value of flexible stern seal bag. One possible explanation for the of the lowest point of the bag. Fig. A probably more H„ /L„ = 0. Since the results in Figs.8 Tp-\. spectral value of the vertical flexible stern seal bag. The uniform pressure resonance The airgap underneath the lowest point of the static bag frequency is approximately 1. Variable Fig. Fig. Fig. 15 illustrates that the reduced accelerations is that increased height to length . between the flexible stern seal bag and the water. The results are presented in Fig. underneath flexible stern seal bag. [Hz] Fig. w/o bagdeformation with bagdeformation '.44 important reason is that the hydrodynamic impact punches H„ IU. The seal bag. 16 shows also transfer functions are used.15 m when the corresponding spectral value is 2. but with different values of Hm . that the frequency region of the uniform pressure resonance is affected. oscillations of a SES can be overpredicted if linear Similar results are obtained at FP. This means that cobblestone is increased. 13).75 impact so that an air leakage is recovered. Spectral values for a design point of view is the height to length ratio of the vertical accelerations are increased about 20% relative to flexible stern seal bag. This described in Fig. Head sea. 15. the response is due to between the two curves is due to the dynamic deformation nonlinear effects.10 m. should be approximately 5.1 m. 14 Spectra of vertical accelerations. but without the effect of water behaviour is a characteristic feature of a flexible stern impact. the vertical accelerations are reduced. The significant difference assumed wave spectrum. Since the linear configuration are presented as function of time in Fig. 13. 16 shows vertical the results in Fig. If the vertical accelerations had a for different height to length ratios of linear behaviour. No Effect of dynamic bag deformations. 13 and 14 occur in the dynamics of the flexible stern seal bag and its effect on frequency region around the first spatial pressure the vertical accelerations.60 the bag out of the water immediately after the initial H„ lit = 0. 13. Fig. When the height to length ratio 3.i s.4 (mls2fs for #1/3=0.

Dr. 4... "Nonlinear Effects of a Flexible Stern Vertical accelerations have a nonlinear dependence on Seal Bag on Cobblestone Oscillations of a SES". pp. Marine Hydrodynamics.M. S0rensen. S. underneath the bag is mainly important in the region where the ratio between the leakage height and the height at the 3. O. Steen.ing. 1993. No. Ship Res. studied... In J.. 1981. this implies that the length of this region is reduced and the effect of the spatial pressure variation as 4. "Hydroelastic diffraction of the incoming wave system by the presence Analysis of a Flexible Bag-Structure". M. "Dynamics and ACV39-ACV47. 8. 25- the present case where the height to length ratio is 41. Sullivan. Ship Res. Thesis.. Faltinsen. Norw. Bentson. of oscillations. Univ. Gosselin. 1992. Dr. Science and Techn. Schiffstecknik. Boston. Norway. USA. scale results. F.. O. the theory needs to be validated by full 12.M. field. S. 20th Symp. 3.M. of Engineering Cybernetics. O.. This Dynamics in the Vertical Plane". SNAME. 200-210. VA. ASME. Vol. Faltinsen. Ulstein. significant wave height. 39. The analysis assumes that the incident waves pass Norway. "Cobblestone actual position is most influenced by the motion of the Oscillations of a SES with Flexible Bag aft Seal". E. M... O. follows. of the side hulls. Trans. Bd 40. S.. 1970. University of California. Trondheim. Hydrodynamic impact on the bag matters and reduces the vertical accelerations. "Modelling and Control of SES underneath the bag reduces the air leakage. T. the air leakage. bag.l. It is shown by a case study that: 6. "Sea Loads on Ships and time domain. Marine Hydrodynamics. Salvesen. Since the cobblestone oscillations are difficult to study in model scale [1]. HPMV'92. Cambridge University Press. bag deformations are important for cobblestone 1996. USA. T. of Science and Techn. J. pp. Arlington. O. pp. Norway... The effect of the spatial pressure variation 1993. S0rensen. A quasi-steady analysis of the fans is used. O. are not important. Davis. This reduction may be explained as of Science and Technology. Dr. Dept. Univ. Norw.13-10 ratio of the flexible stern seal bag reduces the effect of the 2.. [12] reports that dynamic fan effects may significantly reduce 11.1993.A. Faltinsen... oscillations. "Cobblestone Effect on SES". Faltinsen. "Ship Hydrodynamic wave induced loads on the side hulls Motions and Sea Loads". Increased height to length ratio of the bag reduces the vertical accelerations. pp. Tuck.M. 3 in Advances in Marine The flexible stern seal bag reduces the lowest spatial Hydrodynamics (Ed. 9.. 250-287.M. pp. March 1995. P. Computational resonance frequency of the air cushion.. CONCLUSIONS Cobblestone oscillations are theoretically studied in the 5. "Two-Dimensional the spatially uniform pressure response in the air cushion. 1995.M. "Hydrodynamics of High Speed Vehicles". 1994.J. Steen. Vol. Hinchey. This includes interaction with the steady flow Santa Barbara. P. Focus is on the influence of flexible stern Offshore Structures". increased.. problem can be simplified. Trans. that is the region near the lowest point of the bag. seal bags and how the associated aero-hydroelastic 1990. Southampton. The elastic Mechanics Publications. Vol. cushion pressure and bow seal should be on Naval Hydrodynamics. References Flagship Section.. Thesis. Sept. "Dynamic Response of an Air Cushion Lift Fan". undisturbed through the air cushion. effect is important and increases the cobblestone Dept. Trondheim. Conference and Exhibits. Faltinsen. 1996. 211-247. Ohkusu)..J. Steen. Faltinsen.. 71-84. Dynamics in the Vertical Plane". S. Kaplan. Norw..J. Hydrodynamics of Surface-Effect Ships". 1. Unsteady Planing". No. 78. spatial pressure variation underneath the bag caused by Thesis. Faltinsen. 40. T. Univ. pp. "SES a result is reduced. Ch. .ing. Trondheim. SNAME. A.O. 89. A. The spatially varying pressure caused by air leakage 7. Proc. Vol.M...ing.. Dept. Ulstein. N. Ulstein. O. J. The influence of the change in air cushion volume and leakage area due to 10.

the 40 000 tons displacement CdG-CVN is chosen for the control laws (task allocation for the actuators. with landing runway. permitting non-stop handling highly trained crew. the three aircraft main are discussed in this paper: operational requirements in terms operations (handling.01. DCN and control of the rudders while the fins and Cogite system the French Navy took special care of these weather automatically reduce the ship motions.79 .Fax +33. However. take-off and landing) become of performances. and published in RTO MP-15.62. The ship must be a real airport. un Systeme de masses The SATRAP is mainly designed to ease the most dangerous mobiles en transversal pour COmpenser la Glte (Cogite) . instead of up to sea state 4 for the Clemenceau class former CV. active lifting surfaces le vent ou le chargement). C.62.fr emel : hardier@cert.05. To improve the seakeeping performances of the french nuclear architecture du SATRAP.1 Main aircraft operations coordonne. l'appareil ä gouverner. called SATRAP.LAMBERT DGA/DCN INGENIERIE . assurent une stabilisation automatique des mouvements du Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". due to ship motions. which reduces the ship motions while ensuring navigation capability. I 'action des differents organes: deux paires d 'ailerons stabilisateurs. roll. +33. Cogite has evolved into a system which also maintenance devices. +33. even on calm seas : so.01.25. the shipboard feedforward-feedback controllers). aircraft operations such as take-off and landing.59.. .KUMMER G.15 Tel. Steadying and Piloting Automatic System). The following points constraints. la DCN a which controls "horizontal" ship motion. different from the classical oil l'homme de barre a toute liberte d'action sur l'appareil monovariable laws used for a classical piloting or stabilizing ä gouverner. commandes predecessors Clemenceau and Foch. par l'intermediate d'un calculateur centralise.16.40. plate-forme.15. Operational requirement which controls three subsystems: two pairs of stabilizing for the Charles de Gaulle aircraft carrier fins. DCN designed a new system Pour ameliorer la tenue ä la mer du porte-avions nucleaire called SATRAP (Systeme Automatique de Tranquillisation et Charles de Gaulle (en comparaison des porte-avions de Pilotage. a set of rudders and a moving-mass system (athwart ship) for heel compensation.fr Abstract navire. a multivariable structure. parking area.e.. aircraft operational capability of a 90 000 tons US Navy CVN. Les points suivants sont abordes dans ce papier: besoins operationnels au niveau des performances.40. The Netherlands. etc. Ce Systeme 1.System Control & Flight Dynamics Department 8 boulevard Victor 2 avenue Edouard Belin 00303 .77 . where the helmsman has full At the early stages of the Charles de Gaulle design. tandis que les ailerons et le Systeme Cogite system. platform motion control system. navire en compensant la gite durant les manoeuvres. le SATRAP.e. the Cogite system. improves the operational capability of the ship by limiting shipboard aircraft operations remain complex and need a the heel during sharp turns.PARIS ARMEES (France) 31055 TOULOUSE Cedex 4 (France) Tel.59. ce qui facilite et accelere la mise en oeuvre des aeronefs. the ship is on straight course. This paper presents the results of operations in rough weather conditions ? the studies which led to the development of a performing mode of operation for SATRAP.. 14-1 Heel compensation for the Charles de Gaulle aircraft carrier : principles and control structure S.Fax +33. displacements). Ce papier As the horizontal ship motions are physically coupled. qui stabilise le navire et assure les fonctions d'un pilote automatique traditionnel. hangar. i. principes des lois de commande aircraft carrier Charles de Gaulle. which means aircraft operations up to sea state 5. 5-8 October 1998. Resume To achieve this requirement. Originally Aircraft carriers are especially designed to integrate aircraft conceived for situations where the stabilizing fins would be and ship constraints. with enough speed to lorsqu 'ils sont inefßcaces (gites permanentes induites par generate sufficient wind over deck . During these Concu ä I 'origine pour suppleer I 'action des ailerons phases. i. the präsente le resultat des etudes qui ont abouti ä la mise au rudders and fins control laws of this "steadying system" have point d'un fonctionnement " transparent " pour le SATRAP. le Systeme Cogite permet (rudders and stabilization fins) give the best trade-off desormais d 'augmenter les capacites operationnelles du between cost and efficiency.27. DCN has developed a boucle ouverte et boucle fermee). In fact. what about and preparation of aircrafts.25. even if otherwise saturated (steady heel from wind or heavy weights smaller. traffic control.64 emel : dcning@cedocar.25.05. Thus.Suface Ship Department ONERA . SATRAP system architecture and principles hazardous. yaw and mis au point un Systeme de contröle des mouvements de lateral displacement at the arresting gear area.HARDffiR. held in Amsterdam. required to reach. so. in rough weather conditions. with respect to its (repartition des taches entre les actionneurs. This system involves a centralized computer 1. Clemenceau et Foch de Vancienne generation).

It increases the safety by giving the OOD by a high yaw rate. the heel control is essential for CdG-CVN (figure 1): . changing the giration parameters while turning is very demanding. system was designed. For the CdG-CVN. Indeed. most of the time the ship sails with zero heel. an increase of ship reaches its limits. and if the OOD Heel control is also useful to reduce the mission preparation decide to switch to manual steering.full automatic course changing with the mass transfers pre- system instead. a better solution was chosen. heel constraints could phase.by choosing a high rudder angle to quickly reach the parameters. the ship turns into a classical ship and the righting moment could have been chosen. but when the heel compensation device To achieve Foch CV heel performances. The course changing delay is an . to get the maximum righting lever arm. it must be always possible for the OOD to correct easily and safely the giration . it was decided to use the heel compensation system for reducing heel during girations in . DCN chose to develop an active heel compensation . just under the flight deck.also due to the larger flight deck beam and to the increase of 3 4 5 6 shipboard aircraft weight (11 tons Super Etendard versus Steady heel f) 20 tons Rafale). as m Foch-CV shown by aerodynamical scale model testing. Even during straight courses.e.real time computation of mass transfers permitting manual or operational constraint which has to be taken into account for automatic course changing. which allows to carry out take-off and landing operations. however. involving solid mass transfer. rudder angle and wind speed.2 Reducing the mission preparation stage along the manoeuvre . but in this case aircrafts have to be tied prediction algorithm (predicted over about one min) has been down during the course changing. due to the heel induced developped too. and allows aircraft handling. Jk CdG-CVN without Cogite . In this mode. So the second option was chosen. aircraft handling create higher heeling Figure 2 moment (x 1. an aircraft carrier has to change course no more be ensured. while course keeping Wind or course changing. two ways are available to decrease unknown in advance. information in advance for changing the giration parameters.roll less than 3 degrees (max. 1. low frequency part of the roll) less than 1 degree in straight courses. and as wind characteristics (speed and induced heel. the OOD has the ability to change the giration parameters at any time.3 Operational course changing management Aboard a classical ship. The ship would have been difficult keeping the heel operational limits : to stabilize or unacceptable fins surfaces would have been induced. But high stability module induces unacceptable architectural constraints. .by choosing a small rudder angle which results in a low according to the wind.14-2 To achieve heel requirements. which means a high heel increases. without degrading the order to decrease the course changing duration while stabilization fins dynamic efficiency. value). this delay : As an aircraft carrier has to find quickly the best heading . The associated criteria are : Heel compensation . In the first design step. but the aircraft launch. direction) can also vary quickly. and much There are several means to carry out course changing while higher natural roll motion. to come to head wind. maximum heel reached during the manoeuvre will be On Clemenceau class CV. a specific heel control carrying aircraft handling (see figure 2). This solution allows to check that the heel won't be above the operational limit all 1.8) than on former French CV. With an automatic heel compensation system. computed (before course changing). Most of the time. the wind heeling moment • CdG-CVN with Cogite is bigger on Charles de Gaulle than on former Foch CV. transversal stability module. After an optimization stage. the Officer of the Deck (OOD) can learn quickly how to associate the giration steady heel with ship speed. it was made of Speed 25 knots sea water ballasts and transfer pumps located in sponsons.due to the larger flight deck beam.heel (i. which is the third mechanical component of SATRAP. and a heel operational heading. the behaviour of Charles de Gaulle Figure CVN becomes non-linear.

. Figure 4 . developped and built by NFM.2. developped and built by ACH Engineering. To avoid failure. COGITE system architecture and functions 3.steering control (course changing and course keeping) lower speeds being suited for smoother transfers. 0.a redundant high level control and supervision unit which takes ship motion data from the "Integrated Navigation The "solid mass" choice induces 20% weight. 30% cost. but a redundant device located in the Auxiliary Navigation Room allows to operate SATRAP if the bridge is 3. take-off may become impossible and landing crash may occur. the .four 12 m2 stabilization fins. In case of failure. 14-3 SCC : auxiliary navigation room control and supervision unit SCC : fore computer and inertial navigation system Stabilisation fins SCC : aft computer and inertial navigation system Rudders Figure 3 2. This system is usually operated from the design with high reliability constraints. from side to side (figure 4).2 Reliability of Cogite device destroyed. 0.6 and 0.steadying horizontal ship motions Four speeds are available (0. the SATRAP system is essential for allowing this paper.8 m/s) : the last one allows a complete transfer in less than one minute. and so do .the heel compensation device.1 Mechanical principles To ease control of the sub-systems. Thomson ISR developped and built this controller subsystem. . all of them acting on This SATRAP subsystem is equipped with twelve physically bound phenomena. . SATRAP has been designed to comply with a very high reliability and availability requirement. bridge. with ä 12°/s deflection rate.heel compensation (in straight courses or during course changing). The SATRAP system is fitted with the following subsystems (see figure 3) : ©^ . System" (inertial system) and wind data (speed and heading) 40% volume and 50% electrical power reduction compared from the "wind network". with a 10°/s deflection rate. It is also more precise and easier to control orders. the CdG-CV to perform aircraft operations. SATRAP is a very integrated independant 22 tons moving masses (trains). . and which elaborates subsystems with the liquid solution.two 19 m2 rudders. running athwart system which fully performs three functions : ship under the flight deck.4. called "COGITE" by Technicatome which developped it. SATRAP system : 3. which integrates the control laws described in On rough seas.

therefore. motor cooling.14-4 the Cogite subsystem. no desirable. The reconfigurations can happen even load conditions. permitting to derive a continuous estimation of the heeling moment acting on the roll axis.b i I Cogite righting rudders ^ ship heeling moment ^ pseudo-derivation deflection modeling (hydrodynamic displacement ft -+. and the care and roll stabilization for preventing any ill-timed motion of Cogite (which implies a safety filtering with regard to the lowest frequency seas). Principles of heel compensation performances (residual heel less than one degree). be improved by introducing a feedforward component wind. ceaseless displacements of the masses. Cogite response time makes it unsuited to oppose the roll . which will allow to anticipate the foreseeable likewise the outcomes of ship manoeuvres (course-changing heel motions induced by the wind and rudders deflection for instance). because of the controllers structure redundant state. It can effects of the different disturbances acting on the ship. This feedforward control action makes use of of actuators. and from the frequential complementarity of available. compensating the heel variations induced by different ship electrical power . disturbances (sea waves). and nominal performances are still these constraints. insensitive to the first failure. The .system designed to survive to the first failure. compensation system). . delay comes from the integration of these effects through the ship own dynamics. overheating and wear separate rooms.the fins devoted to the control of high frequency predictible moving-masses speed can thus be deduced. i.. the two actuators.very secure manual mode in case of several serious troubles. to avoid loosing all the devices electro-mechanical constraints are not consistent with contributing to a given function.the moving masses of the Cogite device devoted to the characteristics : compensation of the whole low frequency motions. thanks to the following .. and using them all over the frequency range would necessary for preventing the ship from heeling. for any part of It should be observed that a classical water-ballasting system the system and associated ancillaries. on the other hand.a compensation ^/&\ controller ^££ moving-masses lowpass speed SHIP stabilization controller fins deflection bandpass measured roll Figure 5 .b). On the roll axis. during a transfer and are fully automatic. In fact.e. their static efficiency is low. The specific actions of the fins and Cogite . and by means of appropriate filtering.aerodynamic) measured wind Figure 5. and is available for manually redundant processors).e. (figure 5. i. namely : a simplified ship modeling. In case of damage for one electro-mechanical element. / Figure 5. In fact. mechanical actuators. especially during the previously mentioned trends.high level of reliability and availability even in the fully are frequency uncoupled. However. and heel proportional-integral feedback for the . on-board masses displacements (aircrafts handling).e. which is certainly not aerodynamic effects on the heeling moment . by taking into account the desired operational 4. The produce a bias in case of steady heel : due to the limitation advantages of this approach are manifold : of control deflections. (roll rate and acceleration feedback for the stabilization system. the train is coupled to The control scheme of figure 5. according to the (helmsman or automatic piloting). SATRAP disposes of two types transient phase. for control units (3 still subsists on the CdG-CVN.two separate systems (two sub-groups of six trains) in two fast motions and. it should result in a disymmetrical .the control is directly issued from the hydrodynamic and reduction of the sea disturbances. To fulfil this requirement. the control system aims at minimizing the this classical feedback scheme is not sufficient by itself.a is derived from satisfying the adjacent one. sea. i.

v+Cl7/+Cv5sin8+Cvi. The modeling and model speed variations. 14-5 .m I9p+MgGMsmq> +AAf. and from wind tunnel the propelling model is fictitious and not linked with the tests (for the aerodynamic part). and M the ship mass : Mu. present and future informations. P. respectively of hydrodynamic. ..since we get a model with a ship state vector can be developed in the following form : (synchronously estimated) at our disposal. represent the masses and inertia of their corresponding axes. A. As far 5. it is easy to take advantage of them to predict the ship behaviour by f simulation. After synthesis from an ideal frequency shaping. over a short receding horizon (some secondes) . The goal is to allow the model heeling when the =P real Cogite device is saturated (beyond the righting /:. (p is a the roll. this FIR filter is < =Ip5LV: Ch. this is not crucial since using and the adimensional states as : a model is mainly motivated by the search for early starting or stopping of Cogite (more generally. robustness.The thrust Fxp is expressed in terms of u and propeller revolutions Th through the propeller advance ratio. but used to make sure of the relevance of the trends issued from implies to go further into the structure of the internal model the feedforward component (figure 6) : first. This question will be adressed in §5. l. I by denoting L and S the ship length and reference surface. which are measured on board by the wind angle. In the present case. and possibly of the difference between the model and ship heels. and therefore to determine the way the moving masses sensors. The coefficients involved in the aerodynamic forces are In the extreme. and to integrate this compensation shortage when the control is on the point of unsaturating the system. It is evently and H A Mji-M\r = FX +FX +F!' asynchronously readjusted for taking into account the nonlinear effects induced by the masses displacement limit Mvv+Mur = F?+F. aerodynamic. The hydrodynamic part . p Regarding the drawbacks. GM the athwart produced through it are "clean" ones.„2«|" \+C^2+Cxr2r2+C„/v possible to implement a non-causal derivator filtering to produce a moving speed without any additional phase lag F}" =Ip5V2 Cn. the hydrodynamic real one . In other words. the working of hydrodynamic and propelling parts).2v| with respect to the desired righting displacements. several ways are with the real ship. propelling type and induced by Cogite displacements. by using past. as confirmed by many simulations intending to check the robustness of the method.„ (1) capability). the propeller number The internal model is of the traditional manoeuvrability of revolutions is not measured. u and . these parameters are calculated by using the latest coupling existing between the different axes and of the estimation of the real wind and the course changes of the ship nonlinearities induced by the physical limits of the masses trajectory. C represent the forces (F) or the moments (M). where the superscripts H. without taking care of its fitting and consistency Following the structure of equations (1). it is C.bsm8+C/pp+C^ p/V described by the (finite) series of its impulse response coefficients. it would be irrelevant to leave the model internal state free.—rL v = = level change) . and do not require stability modulus. Later on. introducing an explicit modeling into the control scheme could prove to be detrimental to the the water density. the control orders the ship speed. anticipating any speed *=P L . so. (+ 20 m). but is used as a control modeling type. even a simplistic modeling could be useful to computed as a function of the relative wind parameters (speed predict how the ship is going to heel in terms of the steering and direction).the modeling naturally disregards the high frequency v correspond to the longitudinal and lateral components of disturbances induced by the sea . . permitting to compute the feedforward control < =lp5LV2 Cmy+C„rr+C„ssm8 action by straightforward convolution with the sequence of estimated displacements. this trend information proves to be V Tn P -Wi ' V unsensitive to the modeling errors. 5 the rudders angle. Consistency of the internal model state as the internal modeling is concerned. so.the AM. structure obey the following equations : . the different components of which have been variable for keeping the model longitudinal speed u close identified from sea tests on scale models (for the to the ship measured one.v+C. y the heading. it is only used for balancing some of the modeling coefficients will be improved from real sea tests to be errors and for ensuring the consistency between the ship performed soon aboard the Charles de Gaulle.p and I¥ the same kind of filtering as their feedback counterparts. motions. in view of the model. During the prediction stages with the internal have to start for opposing this effect. Mv. including added masses and inertia. In practice.. term squares with an estimate of the aggregate modeling error on the roll moment. However. V = /7cos<p This adjustment is calculated in terms of the moving masses v position (dcs). V the speed modulus.

It is used to be sure that the model the control trajectory stays in agreement with the ship one (especially The feedforward and feedback control components. were continuous variables. a servomechanism involves the up to now. build and periodically adjusted with recent signal samples.14-6 INTERNAL MODEL -<g>>. it can be stated that the rudders angle used during the simulation stages is associated either to the prediction (4 secondes) of these two components. permitting to smooth the decisions to be taken. this yaw rate has aforesaid 9 speed levels . and therefore it is a delayed model automaton. law Th SHIP AMip MODELING if wind compensation feedforward control zr. Since it is imperative not to introduce high However. ? 11 stabilization fins feedback law servo-controls SHIP JCS autopilot rudders or servo-controls helmsman TT wind sea T heading set point Figure 6 . and permits to recover this delay and to get synchronous obviously ill-timed startings which coud result from a too informations about the modeling state. The feedforward helmsman set point for manual steering. added for obtaining model and measured yaw rate with the variable AMV as a the final order to be sent to the compensation device. and the feedback control prediction is issued from variations of the model states. lowpass |^- <& AMW (t-t) r(t-T) P. the supervisor uses a very short-term component. a predictive simulation control output by means of a standard threshold logic. a closed loop simulation of an Auto-Regressive modeling.I.-i i compensation ^KX^J supervisor Cogite feedback law +x-> ' j automaton L. Its function is to quantify the which is servocontrolled. strict quantifying. It explains the appearance of a "supervisor" unit (delay T corresponding to the filter propagation time) in the scheme of figure 6. Problems induced by quantifying error on the yaw moment. this order has first to be quantified (following the frequency disturbances into the modeling. Then. law lowpass f^- P. control input. .D. or to a reconstruction component prediction is easily obtained by model of the autopilot orders from heading set points and internal simulation. in addition to the instantaneous orders delivered by the feedforward and To achieve describing the working of the feedforward feedback modules.I. while avoiding over-delayed decisions of speed level changes.the AMV term represents an estimation of the modeling 6.see § 3) before being used by Cogite to be filtered first. For achieving this. mentioned the yaw rate). To this end.

1-r1 0.0 70 l-P 140 2J10 1 260l 350 .« nniAr v Aft i 0 v VlrV^| » ^^J^ß\h fJVA 350 420 490 ' K.348 1.000 0 140 210 280 350 420 490 5E0 630 700 40 910 980 1050 1120 -70. 560 630 700 7Wf 840 910 980 \f"I 1050 1120 -0.50 tlOy—«o—' 560 630 700 770 ihuj 910 9B0 / 1050 1120 CO -r\ C_) > .174 -140..730 -1.3 0. i i i i Ü 70 140 310 ■ ' 281 350 420 490 560 630 700 770 B40 910 SBO 1050 1120 5 10 CO c_> /t i me i 70 140 210 280 ■j.. i p .Pn .615 4. 14-7 1-581 0 70 140 210 2B0 350 420 430 560 630 700 770 640 310 380 1050 1120 > 0-730 t I me 0.808 — Figure 7 . A A.581 \ H0-34B 70.174 0.r 420 490 560 630 700 1 770 840 Bid 980 -J tVrrie 1050 1120 J -0-6 3.000 ..003 CO U CD *o t i me =>c-1.

psi in degrees. a . supervisor too. The robustness issue is also addressed by means of (at the end of the simulation). . under some . the lateral component of the speed. -fins in degrees. the rudders angle. The notations used for for computing the feedforward control component.psip in deg/s. To cancel it. the ship moving forward at the initial speed of read as a series of course-keeping and course-changing 20 knots. The sea disturbances large hydrodynamic modeling errors. i. followed by emergency manoeuvring at the end of the time t=0) between 30 knots (at the beginning) and 20 knots sequence.(u-nmin) in m/s denotes the variations of the longitudinal by the lowest speed levels (± 0. feedback component is limited.rudders in degrees. and fluctuate during the test as a for the main simulator is different from the internal one used function of the ship encounter angle. posteriori converted into a speed command during the -phi in degrees. circumstances. an convenient lapse of time. the integrator of the closed loop must have the . with variable wind conditions (rear wind at stages. the yaw rate.e. capability of ordering a moving masses displacement. the displacement of the 12 moving masses of Cogite. the ship model used correspond to a sea state 5.v in m/s. because of the dead zone around 0 induced .14-8 On the other hand. . the roll of the ship (in dotted lines. a weak heel residue speed from its initial value of 20 knots. the heading.2 m/s). the deflection of the stabilization fins.vCS in m/s.dCS in meters. Validation of the approach by simulation available levels). the displacement speed (according to the 9 7. distinguishing the different plots of figure 7 are defined as : ZX&f^^l^ Figure 8 . This fonction is devoted to the estimate of the heel being superimposed in solid lines). could remain in steady state since the static gain of the . Figure 7 displays the results obtained during hard The manoeuvres carried out during this simulation can be manoeuvres.

S. by resorting to many types of modeling S.Grandclement. too.Etude 2341 - paper has shown how the design stage has ended in defining Piece n° 11 . a number of simulations were performed to Specification technique de besoin . On the other hand. US Navy. Rapport DCN ING/AN/BS/ the feedback component is active alone. the Cogite system. Outside the zones where Cogite is Likewise. steering trouble) or sensors (loss of SATRAP V2 : Synthese de la comparaison des lois de informations from wind channel). PGn°3-Vl. this is Cogite. vol.Ferrand the feedforward action is useful for anticipating the speed Etude definitive de sürete de fonctionnement du Systeme demands for about thirty seconds.1995 translated into about 1. 1. G.Rapport DCSD-T 1/7997. an adaptation to others types of vessels. this Rapport du Bassin d'Essais des Carenes . 2 et 3 .42 benefits of automatic heel compensation. thus balancing the drawbacks induced by the discontinuous C. Kummer errors. and the control proves to be weakly sensitive to high frequency sea disturbances. Kummer PA CdG . the ship keeping the maquette de Castillon . Rapport Technicatome DI/SCE/ 9505250 . which permits to fulfil the required performances.A. within the framework of . compensation system.Lambert. Conclusions P. real tests. aboard the Charles de Gaulle.Jenn. as well hydrodynamic as aerodynamic.24 . in addition to the control structure which turns the device capacities to the best account.Le Pourhiet.1993 feedforward component in addition to the classical feedback control has settled the trade-off between transient performances and insensitiveness to sea disturbances. In the end. the use of a Ottawa. within the framework of CVX studies.Octobre 1996 stressing the transient stages of heeling. D.10th Ship Control Systems Symposium. M.Quintin stabilization and compensation systems are not available PA CdG : rapport de sürete de l'environnement navire (plu'NC).Lalumia. However. figure 3 depicts the roll motion when the H.O-Fevrier 1997 8. The latter curves PACDG/Sürete / RPS 2/19. in highly damaged situations where an extended loss of sensors signals prevents from safely using A. P. give an idea of the performances which can be in this new concept of compensation which improves the obtained in realistic conditions.Version 2. The robustness has been tested too. about this.Riot has been shown.5° gain on the transient heel. where both feedforward and feedback components developments for her next aircraft carrier. it must be highlighted that the working of the 2/7997. Juin 1998 Regarding future prospects. disposing of an internal optimisees et interruptibles . AN/BS N° 364/96 . V. will be held at sea next winter.Octobre 1996 system is provided with automatic reconfiguration facilities.Fevrier 1998 feedforward and feedback control components is completely uncoupled. take a great interest are active. the masses displacements are identical in the aggregate. as the requests rate of the safety and operational capability of the carrier-borne aircaft.20-Mars 1997 future heel on a receding horizon (of about one minute). G. Besides. They are considered confidently. Definition des lois de commande et du logiciel du Systeme Cogite du Charles de Gaulle -Rapport DCSD-T Moreover.Jung.Perdon Manoeuvrabilite du PAN .P. as far as they are recognized giration . remains below one degree. as the roll (phiFC) and Cogite control order when Attitudes de plate-forme. So. C.Rapport DCN ING/AN/BS/PACDG/SATRAP/ by the acquisition and processing system. and British Navy. In the present case.Rapport DCN ING/AN/BS/PACDG/SATRAP/ actuators.Hardier the feedforward action. G. G. why the control laws structure was Integrated control of warship's steering and stabilizing mostly conditioned by the actuators peculiarities and the system . or navy saturated (+ 20 meters in displacement).Hardier behaviour of the ship heeling.Hardier. but that adding E. in view of the high level of performances and reliability achieved by the Cogite device. the residual heel devices.Rapport DCSD-T modeling allows to provide the OOD with a prediction of the 1/7996. it J. it must be stated that the PG -n° 2 .Lambert. in both cases.girations stabilisees From the operational requirements of an aircraft carrier. desired performances.SATRAP : Girations optimisees interruptibles More generally. in many damaged situations of the actuators (moving masses S.Mars 1991 and realizing an innovating and high-performance heel compensation device. 14-9 Figure 7. References By comparison. Kummer locked or coupled.Rapport DCN ING/ check the performances in all operational conditions.Hardier Compensation de gtte du SATRAP : girations With regard to side advantages. and by SATRAP V2 : Procedures de comparaison des lois de simulating various unknown damage conditions on the giration . could be considered without too many difficulties.0 allow to quantify the contribution of the feedforward Juillet 1996 component : it is easy to see that. the Cogite system will continue to be Porte-avions Charles de Gaulle : identification de la piloted by the feedback structure only.

The objective of the first one. the holder of the world speed record in its This paper presents. been quite a deception for their builders and owners. Techniques Avancees holds the world Italian Navy and the Japanese one (photo ©). So. L 'Hydroptere sails reg. Williwaw (photo ©) is the only known other The objective of the second. either experimental or knots).sailing hydrofoils are said to have been tried in the 75-76. was built in First. stable manner but the speed of those boats have generally • The second deals with the problem of the general lon. it concentrates on the two main (photo ©). ning at medium speed fully level and it looks stable on its A model. this ship is run. held in Amsterdam. Never- sailing boat is Monitor (photo @). force and heeling moment due to the action of the wind The most successful entries were able to sail foil borne in a on the sails. The only speed record in its size category. field. recent well known American sailing hydrofoil is the very ularly at 35-37 knots and it has been recorded at 39 innovative Long Shot built at the beginning of the nineties knots. the well known French sailor. Techniques Avancees. consistently at more than thirty knots. it is to establish a new record for the east. but 1 don't have any documentation on those exper.e. with fully submerged main foils) sailing hydrofoil. Fargot french patent]. at least in France. He is the third second generation sailing. when a team of Dassault engineers discussed with Eric be of interest to extract the hull from the water. The first successful and well documented hydrofoil photo ©). From all I know on this boat. Hawaii. a very successful project (i. commercial. The boat speed. Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". foils have been made in England during the sixties and the • The first deals with the lateral stability of hydrofoils seventies. is even more sailing hydrofoil in the States during this period. inshore waters boats. and convinced him patent for an hydrofoil craft — a rowing boat— has been to try hydrofoils. design advances which explain the success of those two Most of the modern research work on small sailing hydro- ships. The photo © has been explains my permanent involvement in this very specific taken at Monaco in 1908. Eric Tabarly asked for the preliminary design of an Baker in the late fifties with the full support of the US Navy ocean going ship able to compete efficiently in the big (David Taylor Model Basin). for a time. HISTORY Sailors and naval architects have soon realised that the The technique was nearly unknown in France in the seven- wake was the main speed limiting factor and that it would ties. operation was soft (Eric Tabarly is shown on-board on iments. The deception was such in England that the enthusiasm for hydrofoil sailing has decreased a lot in the eighties. 15-1 New advances in Sailing Hydrofoils Francois Lefaudeux Delegation gönerale pour l'armement BP 20-92211 SAINT-CLOUD CEDEX 1. transatlantic races which were then very important events it is clearly apparent that a full aero-hydrodynamic model with a lot of interest in the general public. It was a ambitious. the Dassault engineers to work with a naval architect. and published in RTO MP-15. This ship is said to have sailed France. 2. As can be seen. landing". and. He asked The first successful hydrofoil motorboat has been con. small ocean going vessel and it crossed from Sausalito to ward crossing of the Atlantic Ocean . a for the US Navy are the well known fast hydrofoils patrol lot have been learnt from those experiments and they are boats made by Boeing and also made under licence for the quite successful. a brief history of hydrofoil category with 43. again. but by ocean going ships. foil borne at moderate speed. It was easy to handle and the transition to foil borne thirties. what is difficult is to establish a new all categories world speed record (now to understand is why there was no more interest for this belonging to Yellow Pages Endeavour with more than 46 technique and no follow-on design. gitudinal stability of an ocean-going first generation These bad results have long been a mystery for me: the hydrofoil in strong seas: how to avoid the dangers of fast design of these boats was clearly correct. starting with Monitor. mately one third scale of an ocean going vessel. . Tabarly was not interested by small applied for in 1869 [Mr. 5-8 October 1998. It took all the changes in the flow angle of incidence on the main foils experience accumulated those last ten years to understand with the risks of negative lift and subsequent "hard some of the reasons for this lack of performance. the dating from the end of the fifties. The Netherlands. The next development The two boats have been built and they sail regularly. this structed by an Italian. about the size of the English boats. After this introduction first being Force 8 and Philfly built in the seventies to hydrofoil sailing. but approxi- foils. Abstract has been devised and several very interesting innovations Two sailing hydrofoils projects are currently in progress in incorporated in the design. L 'Hydroptere. The first Tabarly. first. was quite limited.5 knots. the incentive was the speed week of Weymouth sailing ships which have to counter the strong lateral (in fact Portland harbour). Forlanini. designed and built by theless.

has finally been launched rudder. The Most sailing hydrofoils are of the first type. it The first one to be presented is the dependence of lateral succeeded in beating by more than two days the very old stability on the general shape of the main foils. namely the foils roots. they are now closely linked together erally " close hauled " and the resultant lifting force on the and they benefit of the same scientific and technical team. and sometimes a lot Techniques Avancees (the name is coming from the Engi.15-2 When we tried to develop the real ship we came to the con. speed at the time of the shot is about thirty seven knots. modifications in order to have an even faster boat. so.6 knots. has been boats have been designed: particularly appreciated. With Techniques Avancees as well as with L 'Hydroptere clusion that the material most widely used then. a lot of So we came out with an hybrid design — what is sometimes such loops were necessary to understand the physics call a foiler — that is to say a boat with small foils to assist behind the observed phenomena. Atlantic crossing race non stop. in the 1979 double ited and it is the goal of this paper to present two of them. numerous and are mainly used as fast passengers boats) are the starboard arm broke in pieces with no apparent reason at entering this category. ING BOAT. With fast sailing boats. and roll. sail has a lateral component greater. basin and wind tunnel tests have been also undertaken. moments (up to eighty kdn. Finally we redesigned several foils. at the fully mechanical. It took us a long time to understand the • second generation hydrofoils have fully submerged lifting real reasons of the accident. solicited parts of the ship. twist on longitudinal stability. the pilot has only to control the course with the The full scale boat. Most of the motor hydrofoils (which are quite June 95. there is now unanimity to consider that the best foils and. knots by a graceful day. about thirty knots. the horizontal fully submerged stabilising aft foil. To give an idea. the fast and very long towing tank of Centre d'Etudes Aeronau. beginning of December it was allowed to sail at up to thirty six knots when Alain Thebault rushed she up to thirty nine This lecture deals only with first generation hydrofoils. belonging to DGA. I allowed it to go progressively faster. it sailed during all the winter season 94-95. Starting in 1987 two new projects were independently Before to go in those details I will quickly explain the gen- developed in France. two main types of hydrofoils tiques de Toulouse (CEAT). building on our at sea experience. the distance between the forces generates important its size class with 42. one. then. was to finalise the idea behind Paul-Ricard of an sailing hydrofoil: ocean going sailing hydrofoil racer. during a race with the French Navy frigate Surcouf. theoretical interpretation of the problems encountered. Photo OO was also taken that day. the control loops being (photo ©). i. was not strong enough to build the most ence. we made some more the associated control loops. useful. apparent wind is gen- Due to circumstances. naval those performances were not obtained easily. new sea trials. Eastward Atlantic crossing record of Charlie Barr with the The second to be discussed is the importance of main foil well named schooner Atlantic. is of the second type. in the transversal stability but with no possibility to come foil borne. These ships have more or less the same reactions as in 1994. Such a geometry has not a self stability characteristic important pieces of the boat. has been sailing regularly since • propulsive forces and drag forces are not at all in the same 1991 and it has beaten in 1997 the world speed record for plane. rebuilt ship has been launched at the end of September 97 for instance. we made it quite more stronger and.e. as distant later- recorded. than the longitudinal. In normal boats. The second. To deal with those problems. GENERAL STABILITY OF AN HYDROFOIL SAIL- Crossbow II world record of about 36 knots. It finished Several aspects of hydrofoil sailing theory have been revis- second. At sea experi- aluminium alloy. five minutes behind the winner. This boat is Paul-Ricard (photo ©). Techniques Avancees"). I'Hydroptere. eral stability of an hydrofoil sailing craft. but Long Shot. The decision to go full scale has • first generation hydrofoils are designed with a specific been taken in 1992 after the scale model had proven its geometry and V shaped surface piercing foils so as to give ability to sail at more than thirty knots in waves of up to the ship self stability characteristics for flight altitude. • one is the strong lateral force due to the action of the wind on the sails. as two dozens of strain gauges were continuously disposition is with two main forward foils. One is an Engineers School students project "Techniques Avancees" (photo ©) whose ambition was to beat the 3. More interesting for the architects After experimentation with a lot of different foils configura- and engineers. it encountered the same day really harsh seas tions. After some modifications. pitch nearly one metre high. needs appropriate attitude and altitude sensors and and. supporting most of the ship weight and an and computations. neering School named " Ecole Nationale Superieure des this force is in the order of the boat displacement. greater. modifications to the boats and. the dream of a young sailor named Alain There are two main differences between an aeroplane and a Thebault. this allowed us to verify most of our assumptions aly as possible. Alain moments have to be countered either by classical righting Thebault and our team made a lot of sea trials with a scale moment of an archimedian hull or by the lifting forces differ- model about seven metres long (photo ©) and a lot of model ential on the foils.m for L 'Hydroptere) and these Before the drawing of its ocean going vessel. .

1 Lateral stability f other phenomenon Simplified theory depends on the size. we have been obliged to adapt a ver- Classical hydrofoils tical winglet at the tip boats have straight ■■■< of the main foil.—-■. force on the sail and. reason is easy to un- dicular to the wing derstand: with very middle axis.--.-. more difficult to S finally. the equilibrium equations. the transverse equilibrium is de- lateral component of the aerodynamic force induces an angle stroyed and there is an immediate hard landing! of leeway.-■. lift is perpen. 4. I) stipulate that J cal force greater than the two other forces I the displacement and are also passing | so it " climbs " out of through this point. the weight. namic resultant is exactly opposite to the sum of the two other forces. I call ! component can be this point M ! greater than the ship because it has some hydrodynamic resultant j.--. 3). Roll angle adjust itself so that the resultant . i condition the leeward foils theory (see I foil generates a verti- fig. the ampli. the leeway and Transverse equilibrium is far more complex to deal with and the flight altitude adjust themselves so that the hydrody- it is the subject of the next paragraph.. f specific to hydrofoils. hydrodynamic force when there is f charge by unity of no heel. the angle »■.. in that The simplest hydro. until to obtain equilibrium direction. the wetted This is obviously | useful surface de- true for the gravity p creases. sail. 15-3 On quiet waters this aft foil controls very precisely the lon. TRANSVERSE EQUILIBRIUM It can be shown that this equilibrium is stable if M is lower than P. aerodynamic ble. of the aerodynamic force and the displacement passes gitudinal trim.• r——:--.--r-r-TT -.--■•■ T~ . I phenomenon or an dynamic force which Fig. so the wide multihulls hy- two hydrodynamic : drofoils. this theory is quite simple.-. | near to 45°. that is to say under the aerodynamic center of effort. so changing the angle of incidence on the main To avoid this very annoying phenomenon we have added a foils. the twist. It says that the there is a violent side slip. ventilation. the two hydrodynamic forces All classical sailing hydrofoils have been built on that on the foils. This foils. For a practical reason. 8 the water. As for all mechanical equilibria one must have zero resultant and the equilibrium is unstable if M is above P. if there is a stall on the main foil surface. forces and zero resultant moments.-—. Mere we deal with four forces.~-. of the decreases brutally. Anyway. the transverse component of the aerodynamic theory. the lift of attack . wind one. as the dihe- resemblance with ( dral angle of the main the metacentre of a fj foil is always quite classical hull.T-^---. but it is | surface increases and. With this configuration of forces there is no When taking into account the presence of this winglet into roll effect.'~~-. as in the first theory. practice confirms this finding. 2): point M is still a the range of conditions in which the equilibrium is possi- fixed point in the median plane of the ship.. force and the weight converge at the point P which change This point can be explain graphically (fig. finally.-. through point M and.-J'. weight. with roll angle. increasing the down wind one and decreasing the up vertical winglet. the lateral forces converge at a aerodynamic force fixed point. the side force is maintained and the equilibrium is generally tude being automatically adjusted by the change in the automatically recovered before landing! flight altitude. through stall defend for the aero.--.--.-■■■••. etc. one can see that this winglet enhances considerably the stability of the equilibrium and A more realistic theory is as follow (fig.

With strong winds. For instance. . change in leeway.5 ms"'. foil. OCEAN GOING HYDROFOILS SPECIFIC PROB- ing windward and this creates immediately a righting moment which can be at least twice as important than with LEMS It is first useful to describe the open seas waves to be no vertical winglets. as a result. X = 100 m H = 6m The use of such a vertical winglet has been for us a real V = 24 Kt. so we on the leeway angle. the point M is a fixed one. the sail area is greatly L'Hydroplere will needs winds in the range 20-25 knots. characteristics of ocean waves. decreased (reefs on the main sail are five metres apart). is asymmet- With classical straight foils. as most of the speed record ships. the M point is shifting leeward. For an established wind of 25 bined with a small foresail the aerodynamic centre can be knots. but this is not a problem. Photo 11 of £ 'Hydroptere illustrates perfectly this situation. 45° effect). particularly at its back where the speed is downward.. because this leeway angles modify the have decided to experiment as much as possible with the incidence angle on the vertical winglet faster than on the existing foils. it depends on the wetted surface of the main foil and virgule foils will be quite expensive to manufacture. encountered. com- This means sea state 5-6. 2 Lateral equilibrium Fig. (wave speed) breakthrough.6° (maximum slope) The orbital speed of the water is one of the most important increase in speed. ric). a tremendous ct= 10. with no going on . this can induce surging in the speed of the boat. so to be sure we have the best design before main foil (cos. this is not yet the case for L'Hydroptere due to cost With a vertical winglet the direction of lift on the foil is not considerations: her main foils are six metres long and new fixed.. the parameters of the trochoidal model are typically: quite low in these conditions. P is shift- 5. if there is an increase in roll angle. The classical model of sea waves is the so One can show that the increase in the stability domain is called trochoidal model by Gerstner. 3 lateral stability Standard theory virgule foils Techniques Avancees is equipped with a virgule starboard foil (this ship.a sea going hydrofoil like stant speed.15-4 hydvdynamlc resultant \ Fig. This is very important for . wave.(water orbital speed) more harder than before with. To make a record eastward crossing of the Atlantic L'Hydroplere. Horizontal speed is back- There is no theoretical differences between the straight main foil with a vertical winglet and the so-called " virgule " ward in the trough. the stabilising effect is the same. It is represented by such that the situation with the centre of aerodynamic effort below the M point can also be stable if the distance is not figure 4 Water particles are describing a circle of fixed centre at con- too big. allowing the crews to push the two boats far v = 2. The advantage of the The real problem is with vertical speed at mid height of the virgule foil is a significant drag reduction. forward at the crest.

Close hauled. Risks are limited to violent vertical The general aspect of such a foil will resemble the sketch of positive and negative accelerations and even to "take-off fig. may be. The bow can. but I don't also because of pitch rotation inertia of the ship. to adopt a nominal angle of attack more than a few dynamic angle of attack and. against the sea. this is what has been done for incidence (which needs an absolute pitch angle greater than the military patrol hydrofoils. trary to the objective of hydrofoils which is to reduce drag tive lift. much as possible the risk of apparition of a negative hydro. the two others are the vertical speed of the boat dence and thirty knots. The aft foil will modify the pitch to recover a positive incidence controlled foils. small as two seconds. the flight altitude decreases. this repre. Mechanical control solutions the wave slope). behind the crest of the wave. about _.This solution has of to pass in regions of negative lift on the main foils. or reaching in fol. the lift coefficient is about 1 and the itself when climbing up or down the waves and. For a ten degrees hydrodynamic angle of inci- of attack. Z5 The first idea to reduce this risk is to work with a great nom- a = tan :9. hydrodynamic angle of attack (fig. even.5° 15 inal angle of attack on the main foils. positive lift when the tip of the foil encounters a negative The risk. in consequence. negative incidence angles on • second. So. to have a significantly tapered foil. performance was very sensitive to the value of the nominal ing is unavoidable. "water line" and the root of the foil. If the course its own penalty: a momentary sharp increase in drag. but the risk will stay. real seas being quite complex. The needed foresee insurmountable problems to adapt the inner waters . 6). is to lose lift or. the best angle being 2° or less. As any sailor knows. finally. but this is. against the sea. himself having a positive vertical speed: the two The only other solution is to abandon the first generation effect converge to produce negative incidence angles and concept and to adopt the second generation concept with lift. problems are quite different going windward close hauled. efficiency of the aft foil in that situation determines its sur- sent a variation of the angle of attack on the main foils of face. 5. the wetted surface of as far as possible. • first. Time of reaction can greater surface to the high angle of attack parts than to the be a problem as period of encounter of the waves can be as nominal angle of attack ones. its lift per square metre on the foil would theoretically be: pitch angle. 15-5 Wind Effect of orbital Hem at the surface Surface current fig. alas. a hard land. angle of attack. The proposed solution is then two folds: lowing seas. not yet for ocean-going ones. He will landing. of a negative degrees would increase significantly the drag and be con- lift on the main foils. F= 1/2 p V2= 11 kdN.4 Supposing the speed of the boat is thirty knots. encounter there negative vertical orbital speed of water par- ticles. ship goes faster than the sea — and he will normally try to but a lot less than the one corresponding to a full hard do so— he will ascend the back of the wave. Experience has effectively shown that the foil increases and the negative lift increases. If a main foil is in the situation of nega. non This is not the only component in the variation of the angle practicable. they have been validated for orbital speed can be less aft of the ship than forward and inner waters boats. to twist significantly the foil between the nominal foils boats. enter the The goal is to recover as fast as possible an important global water when just beginning the wave ascent. The difference is even greater with fast hydro. giving thus a the main foils are not really anticipated. reaching in following seas.m-2 far too much compared with the limitations imposed by ven- The real problem with sea going hydrofoils is to limit as tilation and cavitation. but there can be a delay because vertical exists for sailing hydrofoils.

1998 . ising the platform. are clearly possible. CONCLUDING REMARKS turbine and foil design advances. capitalis- ing on light carbon construction.Of submerged main foils corresponding to that type of architec. either along the roll axis or the pitch axis. bly the next step for even faster ocean going racers. more advanced design than the PH from Boeing which are now nearly thirty years old. with a good radius of action at that speed and work can help to design better motor hydrofoils ships. lift repartition foil (indicative values not for real use) solutions to the high seas conditions. such systems have been computed. The only possible drawback of the fully As an example. fully flying hydrofoils cannot be thought of for the ture is an overall mean drag slightly more important than bigger types.lt is also. but even for very big ferries foils can enhance a with the surface piercing hydrofoils (due to the fact that the lot their performances. with an interesting armament capability are a real either for commercial or for military purposes. but this solution is proba. true hydrofoils fast patrol boats.Twisted virgule Fig. course. 6 Twisted foil. lower consumption gas 7. 5 . fast patrols boats of up to The developments presented here can seem very specific. of a of course. reducing hull wave drag and stabil- total surface is always wetted). the obvious solution for fast military patrol boats.15-6 FiG. possibility. I 200-300 tons capable of sustaining a speed of fifty knots in am in fact convinced that at least some of the results of this sea state 4. the market for fast ferries is booming. Frangois Lefaudeux January 26. In the military domain.

15-7 Bibliography photographies 1. anonymous 4. the boat that fly 3. Abbot & Albert E. Grogono 1987 8. the boat that flies 7. Grogono . Grogono.B. published by Daniel Charles in Les chasseurs de futur. Veber (Dassault) Francois Lefaudeux (DGA) Techniques Avancees crew Technical team (Techniques Avancies) Gerard Navarin Association Vecteur Vitesse (Ecole des Christian Colombeau techniques avancees students association) Hugues de Turckheim Francois Lefaudeux . Editions du Pen Duick Icarus. published in Icarus. 2.1991 Juanita Kalerhi. et 3 photographies published in Hydrofoil Sailing The forty knots sailboat 4.C. Francois Lefaudeux Theory of Wing Sections 11. 1972. J. Hydrofoil Sailing Editions maritimes et d'Outre-mer . Xavier Linck (ISBN 0-229-11803-8) 9. A Alexander. Francois Lefaudeux Adlard Coles limited. Von Doenhoff 10. Guillaume Plisson Dover Publications (ISBN 0-486-60586-8) L'Hydroptere crew Technical team (L'Hydroptere) Francois Brillant Alain de Bergh (Dassault) Michel Desjoyeaux Michel Fontayne (Aerospatiale) Jean Le Cam Vincent Lauriot-Prevost Alain Thebault (naval architect) Olivier Wroczinski Philippe Perier (Dassault) Jean Abribat (flight test engineer) Marc Van Peteghem (naval architect) Francois Mer (CEA) test engineer Maurice Prat (A6rospatiale) Andre Sournat (Dassault) Hugues de Turckheim (hydrodynamicist & naval architect) J.(ISBN 0-903238-00-4) ISBN 2 7373 0577 2 2. J. D. ©J. Flottes de Combat 1996 (photo Ships of the world 1994) Grosset & Dulap.6. Smith. Niggs 1. Ira H. New-York 5.

15-8 photo O photo © .

15-9 .

15-10 photo © photo © .

.-*^.V J. 15-11 photo © photo ^1P~ Sill 11 I...**&&> '*' ' -vie* .

15-12 photo © .

15-13 photo © .

15-14 photo© Photo ©0 .

5-8 October 1998. C. the fluid velocity Q is decomposed as the result of small immersion or large amplitude motions. A Stokes wave train possibly exists far from the Fn Froude number body. held in Amsterdam. Italian Ship Model Basin Via di Vallerano 139. is shown by computing breaking waves generated by a foil the flow field can be well described in terms of inviscid- slightly immersed with a large angle of attack. Introduction . Giesing and Smith (1967) gave one of the first by introducing x\ (i) = x — L U(t)dt. In this the proper time dependence for us can be inferred from (1) framework. The main motion of the r circulation hydrofoil is a translation but arbitrary oscillations in heave. Unsteady free surface flows with embedded concen- components are solved by an accurate as well as efficient trated vorticity are discussed by Casciola & Piva (1990) by method. hydrofoil arbitrarily moving beneath incoming waves. closer to %. cretizing the Euler equations and linearizing the wake dy- troduced and integral equations for the unknown velocity namics. with suitable evolution equations. V wave elevation P. 16-1 Unsteady flow around a hydrofoil beneath waves U. Once ity to handle highly nonlinear behaviours of the free surface large separation phenomena and cavitation are ruled out. Both foils with and without angle of attack are sented in [4].k=lbk<e sin(fc(a. a general time-domain model recovered. The solution is stepped forward in time by evolu.O vector position in the moving frame 2. 00128 Roma . In this case.ct)) both as control devices of the hull motions. out-of- design conditions render the flow unsteady. ambient Us = V{Y. p water density surge and pitch are allowed.P. rotational fluid mechanics in which a thin vortical layer mim- ics the hydrofoil's wake and a suitable unsteady Kutta con- List of symbols dition provides the mechanism for vorticity generation. the air-water interface. say at x\ = -co. both cases a (steady) Kutta condition provides the unique- face is presented. Moreover. ness of the (numerical) solution. c hydrofoil chord length The resulting model is able to capture qualitatively and quan- d depth of submergence from mid-chord profile titatively several features of the phenomenon and successful U stream velocity comparisons with experimental results are presented in §3. Free surface boundary n hydrofoil conditions which hold in the general rotational case are de- T free surface rived from the Euler equations and the nonlinear free surface w wake dynamics is fully retained.X2) such that the fluid at infinite depth (sp pressure coefficient is at rest. solutions Test cases are solved for which experimental results are for the fully nonlinear inviscid-rotational problem were pre- available. Coleman (1986) firstly erywhere except for a thin wake shed from the trailing edge. M. the wave elevation r\ and the velocity field us are expressed in terms of truncated Fourier series 1. where dimensional problem dealt with via inviscid models. considered and experimental wave profiles are successfully In this paper we present. 2). On that ground. Bulgarelli. In H the fluid velocity is q = U + us + u. §2.i . the pressure dis- gravity acceleration tribution on the foil satisfactorily compares with measured 9 G free space Green function data. using an integral representation for the velocity field coupled tion equations for the free surface and the wake. Greco. Paper presented at the RTO AVT Symposium on "Fluid Dynamics Problems of Vehicles Operating near or in the Air-Sea Interface". and published in RTO MP-15.ct »} waves.Italy SUMMARY ing boundary integral equations. M. frames TZ and TZf. Landrini. solved the unsteady nonlinear free surface problem by dis- A boundary integral formulation for the velocity field is in. fluid velocity in the moving frame In particular the waves generated by foils with zero and non- Q Us Stokes velocity field zero angle of attack are well reproduced and the related wave u perturbation velocity field drag is reasonably predicted.. The steady nonlinear prob- A general model for studying the unsteady free surface flow lem is solved in [9] by a suitable iterative procedure. The Kf inertial frame of reference velocity field is described through an integral representation n moving frame of reference derived from the Poincare identity. Lugni INSEAN. generally speaking.M Lifting bodies are largely employed in naval hydrodynamics = Hk=i ak cos(fc(:ci . We consider an inertial frame p pressure of reference TZf(x\. vehicle's motions and. for finite times. In generated by a hydrofoil moving beneath a wavy free sur.. In both cases. considered. Drag and pressure coefficients on the hydro. A frame of reference TZ(x. The abil. (1) 2 taining system of the entire vessel. which relates the two numerical solutions of the linearized steady problem by us. Consistently. fig. for the two-dimensional flow around a slightly submerged foil satisfactorily compare with the experiments. several authors attacked the problem and we limit with a time dependent velocity —U(t) = —U(t)e\ is also ourself in mentioning those papers relevant for the two.y) moving with respect to TZf Therefore. The Netherlands. Mathematical statement of the problem V unit normal vector to the boundary 2. . and as main sus. the sum of the known Stokes velocity field us and an unknown flow field around the hydrofoil is significantly affected by perturbation velocity field u. The flow field is assumed irrotational ev. as a while.1 The continuous problem T unit tangent vector to the boundary Notation and kinematics The unsteady motion of the a angle of attack fluid generated by a hydrofoil T-L advancing beneath the free 7 vorticity strength surface F is considered (v.

the velocity V and the acceleration A of a be confined in a narrow wake W downstream the hydrofoil point P e U in 11 f are which is modeled by a zero thickness vortex layer of local strength V{P. moves with a ve- flow is described by the Euler equations in the form locity 'defined' as w{£.t) = -U(t)ei+ V(0. Consistently. can be arbitrarily _(_^ei ^x(P-0) chosen. —. .t) are the corre- sponding quantities in TZ. t) are described Hydrodynamic forces For computing the forces through dP(t. the labelled by the Lagrangian coordinate £. 7W (f > t) °'ftne vortex layer according to the Kelvin theorem. Observing that us and U are solenoidal vector fields density ^TE of the 'nascent' wake element is determined and that ^ by an unsteady Kutta-like condition. from the Eu- locity ler equation written in a frame of reference fixed with respect to the hydrofoils. + ^x(P-0) of the fluid velocity across W while the normal component dU qv is continuous by virtue of mass conservation [10]. which follows from the tangential + n(t) x n(t) x (P -o) projection of the Euler equation. Hence. the pressure p(s. a wake point P(£.XTE.q) ■ Vil+S ■ -Vpa Dt J acting on the hydrofoil during its motion. The first equation states of the Euler equation in a body-fixed frame of reference Hb that P moves with velocity w. describes the time evolution of the density J(t. Jj is the rate of change of a quantity seen by an observer mov- ing with a wake point. ay£). (3) ^ Dt p at . which take into account the kinematics and the dynamics of the wake. wT. A zero-thickness vor- Dq Du . it follows Du x Dus v7 +. Wake evolution equations We consider the vorticity to tions of the foil. The + ti(t) x n(t) x (P -o) = ' dt Cl+o(P. appears a reasonable hypothesis.t) 0 (8) dt dt the introduced notation. the second equation. = dt + A(0) + dt gential velocity component uT of the fluid evolves according (12) to the second equation. regardless the actual value of wT.t) W where D/Öi = ö/dt + g • V is the substantial derivative. g -— = —Vp — tA\ (4) Vw = 0 Dtp Dt dT r-1 2-1 -w ■ -yTE (9) dt [3«] Body boundary condition In the frame of reference Tl. + p dr ~dt 2dr tors to T and Jj is the derivative following the motion of the point P(£.t) Lagrangian behaviour of the wake is described by (2) where O is the center of rotation moving with velocity dP(£. Anyway.t) dT(ij. t) of the free surface T. The tangential projection + «• dt (6) da 1 da + fa-T + g-r (11) In these equations v. dU tex wake cannot sustain a pressure jump and. Consistently with = w(£. where the apparent forces is fixed by the corresponding component qv — q ■ v of the fluid below. On physical ground the above condition Free surface evolution equations The free surface states that the vorticity created and convected in the bound- equations follow from the kinematic condition that the fluid ary layer leaves the hydrofoil at the trailing edge to enter into does not cross it and the dynamic constraint that the pressure the wake region. g is the gravity and p is the pressure of the fluid with den. the assumption of zero trailing edge loading the field equations are recast in terms of the perturbation ve. the standard impermeability condition holds in the form which relates the rate of change of the circulation around the hydrofoil.t) -f-1+A(0.{ (w . the normal component w ■ v is then considered. Equations of fluid motion In the moving frame 11.t) In (7).t) = \{q+ +g_). is atmospheric on T [21]. surge and pitch mo.] / p{s)vdl M -L p{s)r x ua (10) dt *„.t) +Cl(t) x (P-O) heave-surge pitch 7w = 7wfc = (q+ -q~) x v = \ir]k (7) -U(t)ei +v(P. In the present initial value problem it is assumed that the vor- sity p. in the limit of ~Dt Dt + Dt dt xw -J. Consistently. to the flux -w ■ iTB of vorticity -yTE = (u .16-2 In the general case with non-zero heave. These requirements are fulfilled if the motion of the free surface points P(£. In particular. The term dU/dt accounts the non-^inertial character ticity is continuously shed from the trailing edge where the of Tl. In (9).t) V(0. respectively. Dus . while the tangential one. r are the unit normal and tangent vec.t) dt to = quis + WTT flu. [qT] = [uT] is the jump of the tangential component A(P.v(P))-v = -(us + U)-u (5) 1 -1- (a-TE ~ aTß) inJected in W with velocity w = ?(aTE + where the unit normal vector v points into the body.t) has to be evaluated. t) and Vt{t) is the angular velocity. a{P. where Vq = 0 -T7 = —Vp + 0+—. the tan. a is the fluid velocity for an observer attached to the hydrofoil.t).t) and v(P.

The reference pressure p(so. Then. comparable except for a constant time dependent shift.Vp / v • v GdSp . Then mula the problem will consist in computing the disturbance u with respect to the uniform stream U(t). Again. while the and the wake evolution equations are integrated in time by a pressure distribution on the foil at different time instants are fourth order Runge-Kutta scheme. to some numerical difficulties. tegral of the vortex layer with uniform strength — [u — v] x v the unknown velocity components are evaluated by solving equals the total circulation T around the hydrofoil. and the in- For rotational flow fields. Dealing with incoming waves. -yw on W +P / (fap+9)-rde ■/' J Sn provides all the information requested to step forward in time is finally obtained. *■ GdSp + Vp x I UX V GdSp • the free surface geometry T can be updated by the first of (6). . (1998) for the p(s. the hybrid velocity representation • the body motion fixes in advance the normal compo- nent UU\H of the fluid velocity (cfr. In fact inside the body %. JH L JH • the knowledge of the velocity field on the hydrofoil allows computing the newly generated vorticity JTE> + Vp x / fWGdSP JW while W changes according to (8). We discuss this point later. it is assumed that the foil sud- «(F) = Vp L u ■ v GdSp + Vp x I uxy GdSp denly appears beneath a regular wave system ranging from Va X -fw GdSp xi = — oo to x\ = oo. is undetermined. the perturbation velocity field u is asymp- of the problem is completed by prescribing suitable initial totically zero far from the hydrofoil. u can be represented by the Poincare For- starts to move from rest in an initially quiescent fluid. say the the solution according to the procedure summarized in fig- pressure at the trailing edge. Let us now assume that at a given problem is here developed. under quite weak conditions. Although the where the first two terms take into account the presence of transient stage is non-physical.| —>• oo and. In this locity field case. the integral equations fol- lowing from (17) are free from eigensolutions and fast itera- V-w = 0 Vxu=C (14) tive solvers can be successfully applied. u(P) = /„. A different approach to the we consider only those cases for which the motion of the hy- drofoil is a priori known. waves and simplify the damping of outgoing signals (see The direct use of (15) to solve the kinematic problem leads Landrini et al. one can expedi. this procedure avoids the the free surface and the hydrofoil and the Biot-Savart inte- need of a numerical wave maker to generate the incoming gral represents the effect of the thin rotational wake W. both the free surface evolution equations loads (10) are not affected by its actual value. without solving the field equations. in fact. (1998)). a method introduced by Longuet-Higgins (1976) "H. Integral formulation of the velocity field In the mov- ing frame of reference 11. 16-3 per unit of mass are introduced and a = ar is the total fluid through a boundary integral equation approach.2 The numerical procedure drofoil exists [1] and that a singular value decomposition General description of the algorithm In the following technique has to be adopted in order to solve the discretized boundary integral equations [5]. (5)). (17) can be deduced by summing (15-16) and requiring Apparently.t) .p(s0. q —¥ us + U as |a. the normal component uv\? and the tangential component uT |^ are still required for a complete knowledge \u — V • u -k = — \u ■ v] x v r = / u ■ rdi (18) of the flow field at the time t0 + At. To the purpose we assume that the hydrofoil hypotheses [1]. eq. while the tangential component uT\j? of + V. Clearly the ure 3. Dealing with free and moving boundaries. In (17) the body effects are represented through a simple ently recast the problem in the form of boundary integral layer potential J aGdS and a vorticity layer distributed on equations.„< Vpp • u. / aGdSp + £vp x f kGdSp the fluid velocity can be prolonged by the second one. following Casciola & Piva (1990). The initial value problem The mathematical statement for finite times. and it is not possible to Ju prolong the solution further in time. In the velocity in this frame of reference.Vp x / v x v GdSp JH JH the fluid dynamic variables that can be prolonged up to the (16) time tg + At. the presence of the moving / JW foil will cause a perturbation u and a change in the wave (15) height T] with respect to the unperturbed r/s. Being the kinematic problem the body circulation explicited.t) •f ■I So ■ Ji — -p a I\ adt 1 2 2 S 'so (13) case of a circular cylinder beneath incoming waves. It can be shown. The source density a accounts the discontinuity in the and Faltinsen (1977) for fully potential free surface flows. a Cauchy problem can be formulated for the boundary geometry and some of V(P) = . By introducing a fictitious ve- time instant to the flow field is completely known. that an eigensolution related to the circulation V around the hy- 2. t). normal direction of the two vector fields across V. . The solution of the kinematic problem with boundary data uT on T uv on W. In particular. By integrating (11) along present paper the vorticity £ reduces to the wake vorticity the body surface the pressure 7 while the more general case with vorticity distributed over a finite area is discussed by Landrini et al.

The accuracy of the computation is even. Later.37. A wave profile shift is detected for the case Numerical treatment of the wake The wake is a vor. same number of crests are plotted for each case. the wake is replaced by cussion it suffices to recall that. refined enough to have results of comparable accuracy. mains are different. We start from figure 4 where. at first inserted upstream. modified free surface evolution equations are reported for increasing forward speed (from top down can be written by introducing a damping function p(x). An enhanced solver for (19) the numerical results with the experimental data available in based on fast summation of multipoles expansion coupled literature. if not stated otherwise. ing the forward velocity.0. in which i) the thickness and the chord are comparable and lution time. that usually sets in for long enough pares well with the numerical profile (solid lines). while the right hand side dealing with large deformations of the wake. by enforcing the conserva- tion of the total circulation and of the center of vorticity. ii) the free surface is far enough to assume the disturbance cal errors due to unbalanced edge singularities may occur. a fixed the final forward speed is used to reduce the oscillatory be- computational window is considered and those points drop. an algebraic system of the form implies that Tk stays constant during the following motions. Fi- Finally.590. steady regime: the wave profile is constant in a body-fixed uated by means of an iterative solver while for the influence frame of reference and the hydrodynamic loads reach con- of wake vortices a fast summation algorithm based on mul. in practical terms. The foil is placed at zero incidence with the trailing edge at tiplies uT and 77 in the dynamic and kinematic free surface x/c = 0. y/c = -1. zero Fn = U/y/gc = 0. the flow field can attain a quasi- self induced velocities. This drawback is avoided by artificial damping layers which Here we reproduce some of the Salvesen experimental con- force the tangential velocity component and the wave dis. stant values. respectively). both set of experiments. the troughs are eventually under- estimated. For clarity. by using a mathematically gener- the velocity components uT and uv and the wave height r\ ated profile. The non dimensional horizontal co- equations. He performed a wide computational domain has to be used and. Although cally determined. the Kelvin theorem (8) drofoil. the discrete parameters for each case are lutions for increasing domain lengths.928. Clearly. the and the maximum damping coefficient have been heuristi. In particular. on T one order lower in the perturbation expansion scheme. unphysical reflections and numeri. Therefore. the self-influence matrix 3. by introducing iV> points a set of point vortices with circulation Tk fixed at the time on the free surface and NH (second order) panels on the hy.Pk) x fe induced by T (19) the point vortices embedded in the flow field. respectively. [18]. the physical dimensions of the adopted computational do- tually maintained under control by comparing different so. Discussion [H -»• H] can be computed once and for all. implies a finite velocity the source strength. for not deforming bodies. the Biot-Savart integral in (17) is replaced by T^T U-+T Uv the sum of the velocity TkVpG(P.0. everywhere except within the damping regions. at least for long enough evo. Increas- evolutions. In these cir- T -> T cumstances. to prevent irregular motions and instabilities dis- ■v. In (19) the matrix coefficient A follows induced on near vortices and a zero self-induced velocity. In these circumstances it is easier to compare tipole expansion is employed.422. sight. A deeper discussion of the numerical method can non zero angle of attack. cussed by Krasny (see [12] and the literature cited therein). the computational effort becomes larger and larger as time passes. A the kernel is desingularized by convolution with a cut-off is obtained for the unknown normal velocity component and function which. vortices are r = {uT } + splitted or merged.16-4 Discretized integral equations For the purpose of dis.759. the measured wave height denoted by symbols com- The sawtooth instability. of shedding according to (9). [17]. as well as the After the transient stage.0. he developed a second order theory do not vanish at the edges. for the larger immer- turbances to vanish when approaching the outer portions of sion d/c = 1. ditions. body circulation T and the wake vorticity 7w. The solution of (19) is rapidly eval. For long time evolution. which mul. the wave forms T. A significant reduction has been obtained by { 7W } using multipoles expansion with fast summation techniques W^H (20) both for computing the self induced velocity and evaluating takes into account the known free surface contribution due the influence of the vorticity field on the other fluid bound- to the tangential velocity component uT. the hydrofoil. the influence of the aries. Fn = 0. at the expense of an approximation. a smooth acceleration up to in the frame of reference H of the free surface points. The length of the damping layers ordinate x/c is scaled by Fn2 so that. Hydrofoil at zero incidence In [17]. Finally.759. Consistently. we firstly discuss the flow around hy- to an iterative solver was developed by Graziani & Landrini drofoil at zero geometrical incidence and then the case with (1998). respectively. by assembling the self-influence and the cross-influence of An additional difficulty in modeling the wake appears when the free surface and the hydrofoil. In the computations. at the first order. Hence. be found there and in Landrini et al. respectively. ping out downstream are thrown away and new markers are Regular wave systems are obtained for all cases and. tex sheet continuously emanated from the trailing edge of The results reported in the right-plots of figure 4 for the . in general. (1998) for the specific problem at hand.37 shown on the left-plots. haviour characteristic of the transient stage. In particular the distance can increase or decrease too much. in order to handle the mean horizontal motion U(t) nally. [18] Salvesen gave the very first attempt to analyze the nonlinearities in wavy Numerical treatment of the free surface A truncated flows generated by a submerged foil. are removed by high order filtering procedures.

to determine the wave-drag. vortex shedding fTE is expected in the initial stage. gated. although RANSE computation takes the Hydrofoil with an angle of attack In this case the foil is bottom effect. the foil surface. 16-5 smaller immersion d/c = 1. This feature is repeated for hol- by the Salvesen formula D = \pgH2 is no more than a lows more downstream. the contribution —pj adl eventually disap- ments of the discrete solution. The improvement is appar. wave height is signif- numerical results obtained by direct pressure integration (de. captured. Con- The last point to mention is the experimental uncertainty. in the first computations performed the bottom of the d/c (bottom plot in figure 5) we also report the theoretical basin was not taken into account because of the deep wa- second order result (dashed line). Because of the smaller depth of submergence. the numerical grid coefficient Cp(s. In the RANSE code. the consistent) third order correction (dotted line) which nicely trough induced above the suction side is under-predicted by reproduces the experimental data and agrees better with the all methods. The reason for this disagreement has to be investi- Salvesen used the first wave peak which is apparently af.1626 and 1. which can explain the differences between the ex- pattern is here compared with data measured by Duncan periments and the reported numerical results only partly.2857. as Salvesen dis. Also a sort cous effects which varies with the immersi