thomson, d.g. and coton, f.n. and galbraith, r.a.m. (2005

Thomson, D.G. and Coton, F.N. and Galbraith, R.A.M. (2005) Simulation study of helicopter ship landing procedures incoporating measured flow data. Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering 219(5):pp. 411-427.

http://eprints.gla.ac.uk/3359/

Glasgow ePrints Service http://eprints.gla.ac.uk

A simulation study of helicopter ship landing proceduresincorporating measured flow-field dataD G Thomson�, F Coton, and R Galbraith

Department of Aerospace Engineering, University of Glasgow, Glasgow, UK

The manuscript was received on 17 March 2005 and was accepted after revision for publication on 27 July 2005.

DOI: 10.1243/095441005X30351

Abstract: The aim of this article is to investigate the use of inverse simulation to help identifythose regions of a ship’s flight deck which provide the safest locations for landing a rotorcraftin various atmospheric conditions. This requires appropriate information on the wind loadingconditions around a ship deck and superstructure, and for the current work, these data wereobtained from wind tunnel tests of a ship model representative of a typical helicopter carrier/assault ship. A series of wind tunnel tests were carried out on the model in the University ofGlasgow’s 2.65 � 2.04 m wind tunnel and three-axis measurements of wind speed were madeat various locations on the ship deck. Measurements were made at four locations on theflight deck at three different heights. The choice of these locations was made on the basis of pre-liminary flow visualization tests which highlighted the areas where the most severe wind effectswere most likely to occur. In addition, for the case where the wind was from 308 to starboard,measurements were made at three further locations to assess the extent of the wake of thesuperstructure. The generated wind profiles can then be imposed on the inverse simulation,allowing study of the vehicle and pilot response during a typical landing manoeuvre in theseconditions. The power of the inverse simulation for this application is demonstrated by aseries of simulations performed using configurational data representing two aircraft types, aWestland Lynx and a transport helicopter flying an approach and landing manoeuvre withthe worst atmospheric conditions applied. It is shown from the results that attempting toland in the area aft of the superstructure in a 308 crosswind might lead to problems for the trans-port configuration due to upgusts in this area. Attempting to perform the landing manoeuvre inan aggressive manner is also shown to lead to diminished control margin in higher winds.

Keywords: inverse simulation, wind tunnel testing, deck landing

1 INTRODUCTION

The current world political situation has necessitatedthe development of rapid reaction forces usuallyrequired to be deployable from ships. The require-ment for helicopters (combat and transport) to beable to operate from ship decks in extremely difficultconditions (and possibly at night) has thereforeemerged. The situation is made more difficultdue to the airwake around the ship interacting withthe helicopter rotors (main and tail). The nature ofthe airwake is determined by the geometry of theship, and hence, each ship and helicopter

combination has its own operational limits ofpermissible wind speed and direction for safe heli-copter operation. This is often presented on a ship–helicopter operational limits (SHOL) diagram whichis constructed by test flying the helicopter to andfrom the ship with the prevailing wind from variousdirections [1, 2]. Figure 1, reproduced from reference[2], shows a typical example of a SHOL diagram.Apart from the expense involved in doing these testsfor each aircraft type likely to operate from the ship,it does of course mean that both the helicopter andthe ship must be available for the tests. Therefore, itis clear that simulation methods, which mightreduce the number of test flights required, would beadvantageous. Lee et al. [3] describe one suchmethod which uses CFD generated flow-field data

�Corresponding author: Department of Aerospace Engineering,

University of Glasgow, Glasgow G12 8QQ, UK.

SPECIAL ISSUE PAPER 411

G01005 # IMechE 2005 Proc. IMechE Vol. 219 Part G: J. Aerospace Engineering

and link this to a simulation model of the helicopterwith an optimal control model of the pilot. In thisarticle, a simulation method is demonstrated, whichcan be used to indicate wind speed and directionswhich will cause degradation of control marginfor a given helicopter and ship combination. Thismethod is therefore a powerful tool for the shipdesigner to ensure the basic design of the flightdeck is ‘helicopter friendly’ and also as an aid inreducing the number of flights necessary to constructa SHOL diagram.

The method used here is known as ‘inverse simu-lation’, which involves specifying the manoeuvrewhich is to be flown and computing the control strat-egy and resulting response of the simulated helicop-ter. This is a technique which is finding increasingapplication [4–6], but much of the work in develop-ing it and applying it to practical problems has beenperformed at Glasgow [7–10]. The main focus of thiswork has been the development of the helicopterinverse simulation package, Helinv. This packagefeatures a wide range of manoeuvre descriptions(the inputs to the simulation), allowing simulationof many operational scenarios and a non-linear,generic rotorcraft model. Owing to its genericconstruction, the helicopter model has a level ofsophistication which allows valid simulation of arange of single main and tail rotor configurations.Allied with the wide range of manoeuvre descrip-tions, this makes Helinv a powerful tool for flightdynamics research.

The main advantage in adopting the inverseapproach is that it is possible to set precise

performance goals, then use the simulation to deter-mine first if the simulated vehicle is capable ofachieving the goals, and if so, what performancemargin remains. Placing this in the context of thecurrent application, the first stage is to define theflight path trajectory of interest – a mathematicaldescription of a landing manoeuvre is required.Inverse simulation can then be used to determinethe control inputs required by a subject helicopterflying the manoeuvre. This exercise in itself isuseful as it is possible to vary aircraft and manoeuvreparameters to optimize the aircraft design or flyingstrategy adopted. The real power of the method isapparent when an atmospheric model is applied tothe simulation. There is then the potential to simu-late landing manoeuvres in the most severe atmos-pheric conditions. The atmospheric model couldbe a simple step or ‘1-cos’ gust profile, but for the cur-rent study, data from wind tunnel tests of a model of atypical aircraft carrier shape have been used to gener-ate realistic wind profiles. The wind tunnel tests wereused to identify areas on the flight deck where poten-tial difficulties on landing may arise due to local windconditions caused by adverse aerodynamic effectsfrom the ship airwake. Inverse simulation was thenused to simulate a typical ship-borne helicopterlanding through this wind profile.

In section 2, the inverse simulation package Helinvand the helicopter mathematical model embedded init are described. The enhancement of the simulationto include a ship landing manoeuvre and atmos-pheric disturbances are presented in sections 3 and4 along with the presentation of a typical simulationresult. Section 5 of the article describes the windtunnel testing of a ship model and discusses theresults obtained. The synthesis of this data intothe inverse simulation is presented in section 6. Thesimulations are repeated for a transport helicopterin section 7 which demonstrates how this techniquemight assist in the preparation of SHOL diagrams.

2 THE INVERSE SIMULATION PACKAGE HELINV

The conventional method of solving the Eulerequations of motion is to use numerical integrationto obtain vehicle response to given inputs. Thismethod is well known and understood unlike theinverse method. Consequently, it is appropriate thata brief description of some of the basics to beincluded here (a fully detailed account of themethod is presented in reference [6]).

2.1 The inverse algorithm

Essentially, the rates of change of the states in theequations of motion are calculated by numerical

Fig. 1 A typical SHOL diagram (reproduced from

reference [2])

412 D G Thomson, F Coton, and R Galbraith

Proc. IMechE Vol. 219 Part G: J. Aerospace Engineering G01005 # IMechE 2005

differentiation (forward differencing). Effectively, thedifferential equations of motion become algebraic inform and can be solved, in the case of Helinv, by aNewton–Raphson approach.

The simulation is initiated by establishing themathematical description of the manoeuvre ofinterest. This is in the form of time histories offour variables: three positions (xe, ye, ze) and head-ing (c). This is easy to understand when one con-siders that the helicopter has four controls; hencefour states may be constrained: main rotor collec-tive pitch (u0) influences vertical motion (ze),longitudinal cyclic pitch (u1s) fore and aft motion(xe), lateral cyclic pitch (u1c) sideward motion (ye),whereas tail rotor collective (u0tr) influencesheading (c).

There are seven equations of motion, the six Eulerrigid body equations

m _U ¼ �m(WQ� VR)þ X �mg sin Q (1)

m _V ¼ �m(UR�WP)þ Y þmg cos Q sin F (2)

m _W ¼ �m(VP �UQ)þ Z þmg cos Q cos F (3)

Ixx_P ¼ (Iyy � Izz)QRþ Ixz( _Rþ PQ)þ L (4)

Iyy_Q ¼ (Izz � Ixx)RP þ Ixz(R2 � P2)þM (5)

Izz_R ¼ (Ixx � Iyy)PQþ Ixz( _P � QR)þN (6)

and the engine torque equation

€QE ¼1

te1te2

½�(te1þ te3

) _QE � QE þ K3(V�Vidle

þ te2_V)� (7)

where te1, te2

, te3, K3 are the time constants and gain

of the governor and Vidle is the angular velocity of therotor in idle.

Once cast in algebraic form, these equations aresolved at regular, discrete time intervals throughthe manoeuvre for the seven unknowns: u0, u1s, u1c,and u0tr (the control displacements), the roll andpitch attitudes (F and Q), and the rotorspeed V.The pilot’s stick displacements (h0, h1s, h1c, andh0tr) are then readily obtained from the knowncalibration relationships with the blade displace-ments. Of course, in calculating the control displace-ments, a wide range of other computations takesplace, including power, torque, attitude, etc., givinga large amount of information on the aircraft’s per-formance as it completes the manoeuvre. In thisstudy, the control angles (i.e. blade angular displace-ments) are converted to stick positions for ease ofinterpretation.

2.2 The helicopter mathematical model

The mathematical model used by the inverse simu-lation Helinv is known as HGS (helicopter genericsimulation) [11]. The main features of HGS includea multi-blade description of the main rotor withquasi-steady flapping assumed, dynamic inflow, anengine and rotorspeed governor model, and look-up tables for fuselage aerodynamic forces andmoments. The question of the validity of the resultsis also important – if any meaningful informationis to be derived, then the mathematical model mustreplicate the actions of the real aircraft. In the caseof Helinv, comparisons have been made using trajec-tory data from manoeuvres flown by real helicoptersto drive the inverse simulation. The computed statesand controls are compared with those recorded inthe flight tests to establish the validity of the simu-lation. The results have shown good correlation fora range of manoeuvres [7]; the validity for lowspeed manoeuvres such as the landing manoeuvresused in this work are particularly good. The HGSmodel is generic in structure, representing singlemain and tail rotor helicopters by a series ofbasic configurational parameters. It is thenpossible to simulate a wide range of different rotor-craft by developing appropriate data files forspecific types.

3 INVERSE SIMULATION OF DECK LANDING

Figure 2 shows a schematic representation of a typi-cal landing manoeuvre. It is assumed to begin withthe helicopter keeping station with the ship from aposition off its port side. The ship is travelling in adirection due north at a velocity Vs, with the helicop-ter positioned at a distance s to the side and height habove the landing point. The first phase of themanoeuvre involves a sidestep to a position overthe deck; the helicopter is then stabilized and finallythere is a vertical descent onto the deck itself. Toobtain an inverse simulation of this manoeuvre, itis necessary to derive appropriate time histories ofthe aircraft position, relative to an earth fixeddatum, and its heading (xe(t), ye(t), ze(t), c(t)). Asthe manoeuvre consists of a series of phases, it isnatural to derive the manoeuvre model in thesame way.

3.1 The sidestep phase

The origin is located at the starting point of themanoeuvre, and the sidestep phase is assumed totake a total time t1. The helicopter moves sidewardspicking up velocity until it reaches a maximumvalue _ymax at a time 1

2 t. The sideward motion is then

Helicopter ship landing procedures 413


arrested and the helicopter is brought to a hover overthe landing spot. It has been shown [12] that thismotion is best modelled by considering the sidewardvelocity profile, _ye(t). From this description, thefollowing boundary conditions are proposed

t ¼ 0 _ye(t) ¼ 0 €ye(t) ¼ 0 y��

e(t) ¼ 0

t ¼1

2t1 _ye(t) ¼ _ymax

t ¼ t1 _ye(t) ¼ 0 €ye(t) ¼ 0 y��

e(t) ¼ 0

The zero acceleration conditions at the beginningand end of the sidestep ensure that the aircraft is intrim at these points, and the third-order derivativecondition is simply to ensure a required level ofsmoothness. The simplest mathematical function

that satisfies these seven conditions is a sixth orderpolynomial

_ye(t) ¼ �64t

t1

� �6

þ192t

t1

� �5"

�192t

t1

� �4

þ64t

t1

� �3#

_ymax (8)

This may seem to be an over simplification; however,previous work [12] in validation of this assumptionshows a very good fit between this ideal velocity pro-file and those recorded from flight test. The sidewardvelocity, _ye(t), and distance, s, are related by thesimple expression

ðt1

0

_ye(t)dt ¼ s (9)

Substitution of equation (8) into equation (9) allowsan analytical solution which yields

t1 ¼50s

23_ymax

(10)

Hence, it is possible to define the lateral positionof the helicopter, s, and the maximum sidewardvelocity, _ymax, calculate t1 from equation (10), andhence generate a time history of sideward velocity,_ye(t), from equation (8).

3.2 Stabilization phase

It is assumed that once the helicopter reaches itsposition over the landing point, there will be aperiod of time, ts, taken to stabilize the helicopter.In practise, this may be several seconds, butfor simulation purposes only a few seconds arerequired.

3.3 Vertical descent phase

From the stabilized hover, the descent is initiateduntil some maximum vertical velocity, _zmax, isreached. The descent is arrested and the aircraftbrought to rest on the deck. If it is assumed thatthe maximum vertical velocity occurs half-waythrough this process, then it is clear that the samevelocity profile as used for the sidestep will be suit-able. If the descent takes a time td and the timebase t� is defined, where

t� ¼ t � (t1 þ ts) (11)

Fig. 2 Schematic representation of deck landing

manoeuvre



then the vertical velocity profile is given by

_ze(t) ¼ �64t�

td

� �6

þ192t�

td

� �5"

�192t�

td

� �4

þ64t�

td

� �3#

_zmax (12)

As with the sidestep phase, it is possible to calculatethe duration of the descent from the height, h, andthe maximum vertical velocity, _zmax

td ¼50h

23_zmax

(13)

Specifying h and _zmax allows td to be calculated fromequation, (13) and hence, the descent velocity timehistory, _ze(t), is defined from equation (12). The vel-ocity profile for the complete manoeuvre is thereforeas follows.

3.3.1 Sidestep phase (0 , t , t1)

_xe(t) ¼ Vs

_ye(t) ¼ �64t

t1

� �6

þ192t

t1

� �5"

�192t

t1

� �4

þ64t

t1

� �3#

_ymax

_ze(t) ¼ 0

3.3.2 Stabilization phase (t1 , t , (t1þ ts))

_xe(t) ¼ Vs _ye(t) ¼ 0 _ze(t) ¼ 0

3.3.3 Descent phase ((t1þ ts ) , t , (t1þ tsþ td))

_xe(t) ¼ Vs

_ye(t) ¼ 0

_ze(t) ¼ �64t�

td

� �6

þ192t�

td

� �5"

�192t�

td

� �4

þ64t�

td

� �3#

_zmax

Defining the manoeuvre in this way allows its sever-ity to be prescribed by varying the parameters s, h,_ymax, and _zmax. For example, holding _ymax constantand decreasing the distance s forces the helicopterto accelerate and decelerate more quickly to achievethe defined manoeuvre in a shorter time.

Finally, it is necessary to define the heading of thehelicopter to complete the set of four constraints.

In still air conditions, it is sufficient to make theassertion that C ¼ 0, that is, the aircraft heading isidentical to that of the ship. In conditions of ambientwind, the pilot may, for example, choose to point thenose of the aircraft into the wind, effectively perform-ing the manoeuvre with zero sideslip (i.e. V ¼ 0). Togive the flexibility required to model different flyingtechniques, it is necessary at this stage to simplysay that

C ¼ f (t)

and the form of f is dependent on the strategyadopted by the pilot.

A typical manoeuvre description is shown in Fig. 3where the parameters used were Vs ¼ 10 knots,s ¼ 20 m, _ymax ¼ 10 knots, h ¼ 10 m, _zmax ¼ 5 m=s,and ts ¼ 1 s.

The _ye(t) plot (‘ydot’) shows an increase from zeroto the maximum value, _ymax, of 5.148 m/s (10 knots)and back to zero over a time of 8.44 s (as calculatedfrom equation (10)). There is then a 1 s stabilizationperiod (ts) and the vertical descent begins at 9.44 s(‘zdot’ plot). The maximum vertical velocity, _zmax,of 5 m/s is clearly visible and the aircraft touchesdown after a further 4.34 s, as calculated by equation(13), giving a manoeuvre time of 13.78 s. The _xe(t)plot (‘xdot’) shows a constant value of 5.148 m/sthroughout the manoeuvre, which is equivalent tothe 10 knots velocity of the ship. Integration of thevelocities generates the flight path plots as shown.

The flight path parameters generated in theexample shown in Fig. 3 have been applied to theinverse simulation Helinv with representative datafor a Lynx helicopter (derived from Padfield [13])applied to the helicopter mathematical model. Theresults are shown in Fig. 4. The manoeuvre isinitiated by a pulse of lateral cyclic stick of around5 per cent of its travel. The resulting roll-time historyindicates a maximum roll angle of around 108. Half-way through the sidestep, the roll direction isreversed to bring the helicopter to a halt directlyabove the landing point. A second pulse of lateralcyclic pitch is applied to achieve this. After the 1 sstabilization period (i.e. at 9.44 s), there is a rapiddrop of collective of almost 20 per cent to initiatethe descent, followed by a rapid increase of around35 per cent to arrest the downward motion. Finally,the collective is returned to its initial value to com-plete the manoeuvre. As the manoeuvre involvessideward and descending flight, there are only smallinputs to longitudinal cyclic pitch, and subsequently,very small changes in pitch attitude are predicted.There are, however, more substantial changes inpedal position (tail rotor collective). During thesidestep phase, the input of tail rotor collective isnecessary to balance the sideward force generated



by the lateral cyclic pitch, whereas during the des-cent, large inputs of main rotor collective have a bear-ing on the torque applied to the fuselage, and hence,there are large inputs of tail rotor collective to providethe anti-torque necessary to maintain heading.

The results shown in Fig. 4 are in still air con-ditions. In section 4, the changes necessary to thesimulation to allow the inclusion of wind effects aredetailed and results presented.

4 DECK LANDING IN PREVAILINGWIND CONDITION

The HGS model had to be modified to enable theinfluence of prevailing wind on pilot strategy andvehicle behaviour to be evaluated. This study isparticularly important as prevailing wind can signifi-cantly alter the power required and control marginsof the helicopter, which when operating in confined

spaces such as a ship flight deck can be crucial to thesafety of the helicopter. A prevailing wind can alsoaffect pilot strategy and is particularly importantduring the low speed phases of a landing manoeuvrewhere the pilot, for example, may find himself apply-ing large degrees of sideslip to counteract the influ-ence of a strong cross-wind. The modificationsrequired for the mathematical model necessary toallow simulation of such conditions are presentedin section 4.1. In this work, it is assumed that theaircraft is instantaneously immersed in the gust,that is, the effect of the rotor disc penetrating thegust is ignored.

4.1 Including the effects of atmosphericdisturbances into the equations of motion

The aerodynamic forces and moments acting on ahelicopter are generated by the relative motion ofair over the vehicle. In inverse simulation, it is

Fig. 3 Parameter time histories for modelled deck landing



usual for the vehicle’s trajectory to be expressed withrespect to some Earth fixed axes set, and previousinvestigations have assumed that the air surroundingthe aircraft does not move relative to this axes set.The velocity time history of the helicopter may thenbe given by

Vh;g ¼ Vh;a (14)

where Vh;g and Vh;a denote the velocity vectors of thehelicopter with respect to the ground and air,respectively.

Considering the influence of the prevailing wind, itis assumed that the wind velocity field is constant inthe region in which the helicopter is immersed.Consequently, there are no significant wind speedvariations over the rotor. Let Va;g denote the velocityvector of the air with respect to the ground; then,equation (14) can be rewritten as

Vh;g ¼ Vh;a þ Va;g (15)

As the velocity of the wind with respect to the groundis known, then the velocity of the helicopter withrespect to the air can be easily determined from

Vh;a ¼ Vh;g � Va;g (16)

The specification of the wind velocity is simplifiedif the horizontal components (in the earth axes xe –ye

plane) and vertical components are considered sep-arately. The wind velocity in the earth axes xe–ye

plane is defined in terms of its absolute velocitycomponent, Vw, and angle between the in-planevelocity vector and the x-axis denoted by Cw, asillustrated in Fig. 2. With this definition, the winddirections are

Cw ¼ 08 tailwind

Cw ¼ 908 wind from port side

Cw ¼ 1808 headwind

Cw ¼ 2708 wind from starboard

Fig. 4 Inverse simulation results for landing manoeuvre from Fig. 3, Lynx helicopter



The specification of the wind velocity is completedby defining the vertical component of wind, Vwv

, andthis is chosen to be positive downwards. Hence, thethree wind velocity components of the vector Va;g

can be obtained from

Va;g ¼

�Vw cos Cw

�Vw sin Cw

Vwv

24

35

Therefore, once the earth components of wind areknown, the velocity of the helicopter with respect toair expressed in Earth axes may be determined fromequation (16). It is then a simple matter to transformthe resulting velocities through the Euler sequence(C, Q, F) to determine the velocity of the airflowwith respect to helicopter in vehicle body axes.The aerodynamic components of wind velocityatthe rotorcraft centre of gravity, Ua, Va, and Wa, arethe sum of the inertial velocities, U, V, and W, andthe wind components Ua,g, Va,g, and Wa,g. Thus

Ua ¼ U þUa;g

Va ¼ V þ Va;g

Wa ¼W þWa,g

The fuselage angle of incidence and sideslip used todetermine the fuselage forces and moments can begiven by

aF ¼ tan�1 Wa

Ua

� �and bF ¼ sin�1 Va

Vf

� �

where the flight velocity is given byVf ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiU2

a þ V 2a þW 2

a

p. Hence, the aerodynamic

forces and moments for the complete aircraft maybe evaluated in the usual manner.

4.2 Inverse simulation of deck landing with aprevailing wind

Referring to Fig. 2, it is necessary to define a windfrom a direction, Cw, with constant velocity, Vw.Figure 5 shows inverse simulation results for a West-land Lynx helicopter flying the landing manoeuvre,described by Fig. 3, in still air (as in Fig. 4), in aconstant 40 knot headwind (i.e. Cw ¼ 1808 andVw ¼ 40 knots) and in a constant 40 knot tailwind(i.e. Cw ¼ 08 and Vw ¼ 40 knots). It is assumed thatthere is no vertical component of the wind.

Fig. 5 Inverse simulation results for landing manoeuvre from Fig. 4 – fore/aft wind cases



The effect of flying the manoeuvre in a constantwind is quite clear – all the controls are offset fromtheir still air values. The additional 40 knots airflowover the rotor gives a lower collective pitch to gener-ate the same thrust, and hence, the requirements onengine torque and power are also reduced. The offsetin longitudinal cyclic pitch indicates a forwardmotion of the stick to maintain fore/aft position rela-tive to the ship in the 40 knot headwind and an aftmotion for the tailwind case. Lower engine torqueallows a reduction in tail rotor pitch, and hence, ashift in the lateral cyclic stick (less side force isrequired to counteract the tail rotor thrust).

There is nothing in Fig. 5 to suggest that landing ina constant headwind would cause a problem. Thereis ample control margin, with all the controls wellwithin their limits as are engine and tail rotortorque. This is not a surprising result as landinginto the wind is always the most desirable option.The power of an inverse simulation is that thismanoeuvre can now be repeated with any windintensity and direction applied. The simulation

performed for Fig. 5 was repeated with the samewind intensity (40 knots) but with the focus onbeam winds (i.e. Cw ¼ 908; 2708), and the resultsare plotted in Fig. 6.

A beam wind has two significant effects: it influ-ences sideforce on the fuselage and the tail rotorinflow and hence thrust. Consider the case of the40 knot wind from the port side (the chained line).As indicated in Fig. 2, it is assumed that themanoeuvre is initiated from the port side of theship. The initial sideward motion is aided bythe wind, and hence, a lower collective position isadopted when compared with the no wind case; how-ever, in attempting to arrest the sideward motion asthe ship deck is approached, the additional drag hasto be countered by increased collective. This con-trasts with the starboard wind case where additionalcollective is required to initiate the manoeuvre andless to achieve the ‘hover’ over the deck. The sideforceproblem due to the beam wind will of course havean effect on required lateral cyclic position. Themain problem with the beam wind is the effect it

Fig. 6 Inverse simulation results for landing manoeuvre from Fig. 4 – beam wind cases



can have on tail rotor thrust. Consider the case ofthe Lynx where the main rotor rotates in an anti-clockwise direction when viewed from above, genera-ting a torque in the clockwise direction on thefuselage. The tail rotor provides a thrust to port togenerate the anti-torque moment (in an anti-clockwise direction) to this. A wind from the portdirection (chained line) in addition to the portdirection of motion can cause a reduction of theinflow and hence thrust generated, thereby leadingto higher tail rotor collective pitch displacements.The value of 0.5 for tail rotor collective correspondsto the pedals being centralized, whereas a value of 0corresponds to full left pedal. Figure 6 shows thatwith a starboard wind, there is plenty of controlmargins available on the pedals, whereas with aport wind causes a larger left pedal input, taking itclose to its limits.

Another feature of these results is that the enginetorque exceeds its maximum value between 2.11and 3.52 s (coinciding with maximum collectivepitch) in the starboard wind case. The simulation ispredicting that the Lynx would not be able to flythis manoeuvre under these conditions – a reducedmaximum sideward velocity or increased lateraldistance from the ship at initiation of the manoeuvrewould be necessary.

The use of inverse simulation to investigate theeffect of a constant wind on deck landings has beendemonstrated. Clearly, the situations of a beamwind are more critical than head or tail winds.Further, only constant winds were considered inthis section. The real issue for helicopter and shipdesigners is the effect of flying the helicopter in air-wake generated by a wind impinging on a ship super-structure. There is no limitation on the profile of thewind distribution applied to the inverse simulation,and so, it is possible to apply realistic atmosphericsituations to the simulation. In the following section,results of a wind tunnel trial using a representativeship model are presented. These data are then usedto generate realistic profiles for application to thesimulation.

5 EXPERIMENTAL MEASUREMENT OFAIRWAKE AROUND MODEL OFAN ASSAULT SHIP

A series of wind tunnel tests were carried out ona model helicopter carrier/assault ship in theUniversity of Glasgow’s 2.65 � 2.04 m Argyll windtunnel. The 1/145th scale ship model was madefrom wood and mounted on a ground board abovethe floor plane of the tunnel. Testing was conductedat a wind speed of 13.2 m/s giving a Reynoldsnumber of 1.27 million (based on hull length), and

three-component velocity measurements weremade using a TSI IFA-300 Hot-Wire constant temp-erature anemometer system with DANTEC 55P61cross-wire probes. These probes have 5 mm platinumplated tungsten wire sensors with a length to dia-meter ratio of 250. The measuring volume of theprobes is approximately 0.8 mm in diameter and0.5 mm in height. The test wind speed was chosento allow the measuring system to resolve velocitiesover the full range of flow conditions expectedaround the ship.

A series of preliminary flow visualization tests wereconducted on a smaller scale model of the ship in theUniversity of Glasgow’s 0.91 � 0.91 m smoke flowvisualization facility. This study provided qualitativeinformation that was then used to identify the mostappropriate regions of the ship flow field to site thehot-wire sensors, i.e where the most severe windeffects were most likely to occur. An example of thevisualization data is shown in Fig. 7 where a flow

Fig. 7 Flow from 308 to starboard

Fig. 8 Schematic representation of probe measurement

positions



from 308 to starboard is shown to produce a substan-tial recirculation region in the wake of the shipsuperstructure.

On the basis of the flow visualization study, hot-wire measurements were made at four locationson the flight deck at three different heights. Themeasurement locations used in the test are shown(in full scale) in Fig. 8, and the measurement heightsare listed in Table 1 subsequently. In addition, for thecase where the wind was from 308 to starboard,measurements were made at three further locationsto assess the extent of the wake of the superstructure.In all cases, data were collected at a sampling rateof 1 kHz over a 4 s period. Prior to engaging in thefull test programme, data were sampled at discretelocations around the ship over longer periods (upto 20 s) to determine the range of significant

frequencies in the flow. On this basis, it was estab-lished that a 4 s measuring period was sufficient toprovide accurate estimates of mean velocity.

The mean values of the measurements taken atthese positions and heights are presented inTable 2. It should be noted that the velocities weremeasured relative to the flow direction, not the shiporientation. The sign convention of the velocitiespresented in Table 2 is as indicated in Table 3.

As accurate wind data are available for thelocations and heights shown in Fig. 8 and Table 2,the focus here will be on creating flight pathmodels which pass through these points and thensuperimposing the known wind conditions ontothem. This technique will be illustrated first byconsidering the most severe case identified: a windcoming from an angle of 308 from starboard(Cw ¼ 2408) across the superstructure.

6 INVERSE SIMULATION OF LANDING IN WAKEFROM WIND ACROSS SUPERSTRUCTURE

The most severe wind condition identified fromthe wind tunnel tests was for locations aft of the

Table 1 Measurement heights

above deck (full scale)

Height reference Actual height (m)

1 22.22 14.83 7.4

Table 2 Wind tunnel measurements of velocity

Runnumber

Wind direction(to starboard) (8)

Probeposition

Probeheight Uf (m/s) Wf (m/s) Vf (m/s)

37 15 1 1 13.516 22.148 0.46338 15 1 2 8.506 20.953 0.13939 15 1 3 12.476 21.059 0.75940 15 2 1 12.476 20.305 2.21841 15 2 2 12.959 20.394 2.80142 15 2 3 9.887 20.933 1.70543 15 3 1 13.621 21.123 0.47844 15 3 2 13.825 21.495 0.44345 15 3 3 14.050 21.616 0.95946 15 4 1 13.322 20.344 0.67947 15 4 2 13.376 20.300 0.28148 15 4 3 13.481 20.414 0.79849 30 1 1 14.008 22.127 �0.83750 30 1 2 11.731 21.515 �1.49281 30 1 3 8.229 20.129 �1.62052 30 2 1 14.477 22.691 1.39953 30 2 2 10.522 20.794 1.21982 30 2 3 10.079 0.882 1.37355 30 3 1 14.231 21.668 0.03156 30 3 2 14.495 22.704 �0.30683 30 3 3 12.303 22.100 �0.11558 30 4 1 13.597 20.738 0.24284 30 4 2 13.784 21.055 1.32285 30 4 3 14.017 21.324 1.34186 30 5 1 12.251 22.178 1.46387 30 5 2 6.648 21.516 0.01788 30 5 3 10.352 20.382 0.85589 30 6 1 13.562 20.617 2.81690 30 6 2 13.123 20.349 3.55191 30 6 3 11.723 1.472 2.29092 30 7 1 13.979 20.927 2.12793 30 7 2 12.333 0.678 2.31294 30 7 3 9.202 20.401 1.000



superstructure when a wind was present 308 from thestarboard side. Measurements of wind velocities atpoints 6, 5, and 1 have been recorded for a heightequivalent to 22.2 m and for two other heights atthe landing point 1. The aim first is to derive aflight path model which passes through these points.

6.1 Construction of flight path

Referring to Fig. 8, the lateral distance travelled frompoint 6 to point 1 is 30.81 m, i.e. for the manoeuvremodel(referring to section 3) s ¼ 30.81 m. FromTable 2, it is convenient to have the height abovethe deck as (h ¼ ) 22.2 m, and retaining the othervalues for the manoeuvre (i.e. Vs ¼ 10 knots,_ymax ¼ 10 knots, _zmax ¼ 5 m=s, ts ¼ 1 s), it is possibleto construct the flight path as shown in Fig. 9.

All the inertial data are now available for the simu-lation; however, it is also necessary to superimposethe wind velocity profile on top of the manoeuvre.This profile has been measured in a positional coor-dinate frame from the wind tunnel test but isrequired as a time history for the simulation. There-fore, it is necessary to locate the time at which themeasurement points are reached. These are indi-cated in Fig. 9, for example, having initiated themanoeuvre at point 6, point 5 which is a lateraldistance of 15.95 m (Fig. 8) is reached after 6.8 swhen using the model described in section 3.

6.2 Construction of wind profile

The atmospheric data for the simulation are requiredin relation to an earth fixed frame of reference,whereas, as indicated in Table 3, the wind tunneldata are measured relative to the flow direction. Atransformation is therefore required, which isderived as follows. Let Cf be the angle between thecentreline of the ship and the flow direction(Fig. 10) and Uf and Wf be the streamwise andcross-stream velocities of the flow. Referring toFig. 10, the transformation to the earth axes set(incorporating the downward component which isthe same in both axes sets) will be given by

Va;g ¼

_xw

_yw

_zw

24

35 ¼ cos Cf sin Cf 0

sin Cf � cos Cf 00 0 1

24

35 Uf

Wf

Vf

24

35

By selecting the appropriate runs from Table 2 andperforming the transformation given previously,with Cf ¼ 308, the wind data required can be con-verted to an appropriate form as shown in Table 4subsequently.

The data from Table 4 are plotted in Fig. 11 andrepresent a useable wind distribution time historyfor the inverse simulation. The data presented inTable 4 and Fig. 11 is an average value taken overthe 4 s measurement period during the wind tunneltest. The question of how so few points are to beused to generate a time history was given extensiveconsideration. For example, it is possible to fit asmooth polynomial profile through the time pointsto generate a continuous profile. Given that thewind data are actually a mean value of fluctuatingturbulence, any form of smoothing would be unre-presentative, therefore, it was decided that simplelinear variation between the points would be themost appropriate course of action. The final stagein preparing the wind data is to scale it to an appro-priate intensity. For previous results, a 40 knot

Table 3 Velocity sign convention

Streamwise velocity, Uf Positive downstreamVertical velocity, Vf Positive downwardsCross-stream velocity, Wf Positive to the left

(looking downstream)

Fig. 9 Flight path for landing through points 6–5–1 Fig. 10 Frames of reference for wind data



(20.59 m/s) mean wind speed was used, and hence,for consistency, the wind velocities presented inTable 4 and Fig. 11 (which measured at a freestreamspeed of 13.2 m/s) were scaled by a factor of 1.56.

6.3 Inverse simulation results

With the flight path model and wind distributionmodel both available, it is possible to simulate aLynx helicopter performing the landing manoeuvrethrough the wind recorded in the wind tunnel exper-iments. The results are shown in Fig. 12, plotted incomparison with the still air results.

The results show that the presence of the winddoes enforce changes to the amplitude of the control

inputs (suggesting some increase in workload), butno gross change to the strategy adopted by thepilot. Although there is a clear offset in all of thecontrols, in each case there is still plenty of marginleft, indicating that the helicopter will still be control-lable in this environment. The rotorspeed doesapproach its limit during the initial descent phaseof the landing due to the presence of a smallup-gust at this time.

7 SIMULATION OF TRANSPORT HELICOPTERLANDING ON DECK AFT OFSUPERSTRUCTURE

Having repeated the aforementioned analysis forvarious locations on the deck, it was clear that thelikely airwake conditions for ship of this designwould cause little difficulty for small agile helicoptersuch as the Lynx. A significant feature of this aircraftis that it possesses a rotor which is stiff (i.e. hingeless)in its flap degree of freedom. This affords it superiorcontrol power and manoeuvrability over helicopterswith flap hinged rotors. Clearly, a larger, less agilehelicopter may not be as capable of landing safelyin the modelled conditions. To illustrate this, a

Table 4 Wind velocity information for landing

manoeuvre

Runnumber

Lateralreference

Heightreference

Time(s)

_xw

(m/s)_yw

(m/s)_zw

(m/s)

89 6 1 0.0 11.436 7.315 2.81686 5 1 6.8 9.52 7.214 1.46349 1 1 13.01 11.068 8.068 20.83750 1 2 18.06 9.402 7.177 21.49281 1 3 19.56 7.062 4.226 21.62

Fig. 11 Wind distribution from wind tunnel data



different data set representing a larger transporthelicopter of mass 8000 kg (based loosely on theWestland Sea King) was implemented in the HGSmathematical model. The inverse simulation of thetransport helicopter flying the flight path as shownin Fig. 9, with the wind conditions given by Fig. 10is shown in Fig. 13. This is directly comparable withthe Lynx results shown in Fig. 12.

The plot for the collective channel shows thatduring the vertical descent stage, the collectivepitch calculation estimates a value below zero.The wind profile used is shown in Fig. 10, and itcan be seen that at the time of the control limitbeing exceeded, the helicopter is experiencing anup-gust of around 2 m/s. In fact, the control limit isexceeded for only a fraction of a second; however,this indicates that the transport helicopter wouldnot be able to complete this manoeuvre in thedefined manner. It is clear that there are configura-tional characteristics that can affect the ability ofthe helicopter to perform the manoeuvre.

Another issue relates to the manoeuvre modelused. Although this model is representative of thetype flown and should lead to realistic strategiesbeing derived, there is the question of the parametersused – increasing or decreasing parameters vary, theseverity of the manoeuvre and influence the pre-dicted control strategy. For example, increasing themaximum sideward velocity in the sidestep phasefrom 10 knots to 20 knots increases the power andtorque requirements, and in the presence of a side-winds power and torque (engine and tail rotor)limits are quickly exceeded. The flight path para-meters used in the current work were selected togenerate relatively gentle manoeuvres well withinthe normal operating limits. If more aggressive pilot-ing was the norm, then the choice of parameterswould have to be reviewed to reflect this.

It should also be noted that there is an assumptioninherent in the helicopter mathematical model thatthe whole rotor is immersed instantaneously in thegust field. In fact, there are penetration effects

Fig. 12 Inverse simulation results for Lynx flying landing manoeuvre (data from points 6–5–1,

mean wind speed of 40 knots)



present (different regions of the rotor disc experien-cing different flow speeds as the helicopter pro-gresses through the manoeuvre). Such penetrationslead to load variations across the disc and generatehigh levels of vibration. A simulation such as theHGS model used here is unable to capture thiseffect, however, gust penetration can be incorpor-ated into individual blade models. Research intoinverse simulation using individual blade models(including gust penetration) is currently underwayat Glasgow.

Finally, the way in which the wind data are pro-cessed may have an influence on the nature of thefinal results. The simulation requires time historiesof the wind distribution, whereas the data providedfrom the wind tunnel were measured at discretelocations, with time histories of flow speeds at eachlocation. The mean value at each location wasused, and so, information on the rate of change ofwind speed and turbulence is lost. Ideally, themeasurement from the wind tunnel should berecorded in real time as the probe is traversed

along the flight path to be simulated. This wouldrepresent a technical challenge but would generatea more realistic model of the wind.

8 CONCLUDING REMARKS

The construction of SHOL diagram is essential forsafe and effective operation of a helicopter from aship deck. This is a lengthy and expensive exercise,and consequently, any analytical aid which canreduce the number of flights required must be ofvalue. In this report, the use of inverse simulationto achieve this goal has been demonstrated. It isclear that the technique can differentiate betweenacceptable and unacceptable performances in par-ticular atmospheric conditions for specific helicopterconfigurations. Inverse simulation techniques arebecoming well established and accepted, and sothe real challenges in using this method are in ensur-ing that the helicopter model is representative andthat the airwake data are realistic. The current

Fig. 13 Simulation of transport helicopter landing aft of superstructure



study also included wind tunnel testing of a contem-porary helicopter carrier/assault ship design, whichindicated that the most critical condition occurswhen there is a wind coming from an angle of 308from starboard across the superstructure. In theseconditions, there is substantial recirculation regionin the wake of the ship superstructure which couldprovide landing conditions outwith the normal oper-ational boundaries of the helicopter. This work alsodemonstrates that wind tunnel data can be trans-lated into a form suitable for use with inverse simu-lation. It is then clear that inverse simulation hasthe potential to contribute to the construction ofthe SHOL diagrams and thereby assists in determin-ing the operational limits of the ship/helicoptercombination.

ACKNOWLEDGEMENTS

The authors wish to acknowledge the support ofHanjin Hervy Industries who funded this study.

REFERENCES

1 Gowen, T. E. and Ferrier, B. Manned flight simulatorshipborne handling qualities assessment using theproposed ADS-33 standard. In Proceedings of the 57thAnnual Forum of the American Helicopter Society,Washington, DC, May 2001.

2 Handcock, A., Lane, R., Johns, S., Howitt, J., andCharlton, M. Benefits of advanced control technology.In Proceedings of the 56th Annual Forum of theAmerican Helicopter Society, Virginia Beach, May 2000.

3 Lee, D., Sezer-Uzol, N., Horn, J. F., and Long, L. N.Simulation of helicopter shipboard launch and recov-ery with time accurate airwakes. AIAA J. Aircraft, 2005,42(2), 448–461.

4 Celi, R. Optimization-based inverse simulation of ahelicopter slalom maneuver. In Proceedings of the25th European Rotorcraft Forum, Rome, September1999.

5 Hess, R., Zeyada, Y., and Heffley, R. K. Modelling andsimulation for helicopter task analysis. In Proceedingsof the 57th Annual Forum of the American HelicopterSociety, Washington, DC, May 2001.

6 Avanzini, G. and de Matteis, G. Two-timescale inversesimulation of a helicopter model. AIAA J. Guid. ControlDynam., 2001, 24(2), 330–339.

7 Thomson, D. G. and Bradley, R. The principles andpractical application of helicopter inverse simulation.Simulat. Theory Pract., 1998, 6(1), 47–70.

8 Rutherford, S. and Thomson, D. G. Helicopter inversesimulation incorporating an individual blade rotormodel. AIAA J. Aircraft, 1997, 34(5), 627–634.

9 Thomson, D. G. and Bradley, R. The use of inversesimulation for preliminary assessment of helicopterhandling qualities. Aeronaut. J., 1997, 101, 287–294.

10 Thomson, D. G., Talbot, N., Taylor, C., Bradley, R.,and Ablett, R. An investigation of piloting strategiesfor engine failures during take-off from offshore plat-forms. Aeronaut. J., 1995, 99(981), 15–25.

11 Thomson, D. G. Development of a generic helicoptermathematical model for application to inverse simu-lation. Internal report No. 9216, Department of Aero-space Engineering, University of Glasgow, June 1992.

12 Thomson, D. G. and Bradley, R. Mathematical model-ling of helicopter manoeuvers. J. Am. Helicopter Soc.,1997, 42(4), 307–309.

13 Padfield, G. D. Helicopter flight dynamics, 2000(Blackwell Science, Oxford).

APPENDIX

Notation

h height of helicopter above shipdeck at start of manoeuvre

Ixx, Iyy, Izz, Ixz moments and product of inertiaK3 rotorspeed governor gainL, M, N external momentsm aircraft massP, Q, R roll, pitch, and yaw ratesQE engine torques helicopter’s distance to port side

of ship at start of manoeuvret timets stabilization timeU, V, W component translational

velocitiesUf, Wf, Vf flow components from wind

tunnel testsV velocity vectorVf flight velocity of helicopterVs constant velocity of shipVw mean wind speedxe, ye, ze flight path coordinate positions_xw; _yw; _zw wind velocity components in

earth axesX, Y, Z external forces_ymax maximum sideward velocity in

sidestep phase of landing_zmax maximum downward velocity in

descent phase of landing

af;bf helicopter angles of attack andsideslip

F, Q, C roll, pitch, and azimuth angles(the Euler angles)

Cf relative flow direction in windtunnel tests

Cw wind heading (from north)te1

, te2, te3

rotorspeed governor timeconstants

u0 main rotor collective pitch



u1s, u1c main rotor longitudinal andlateral cyclic pitch

u0tr tail rotor collective pitchV main rotor angular speedVidle angular velocity of the rotor in

idle

Convention for control positions

Collective lever 0 ¼ full down, 1 ¼ full up

Lateral stick 0 ¼ full left, 1 ¼ full rightLongitudinal stick 0 ¼ full aft, 1 ¼ full forwardPedal position 0 ¼ full left, 1 ¼ full right

Subscripts

a airg groundh helicopter



thomson, d.g. and coton, f.n. and galbraith, r.a.m. (2005

Documents