Jounal of ces a Wind Eacey (SSN 2310-3608) {Gopriht © by The Interaional Sit of Oboe and Polar giners VoUT No. Febrny 2014, pp. 41-10 butp:s/www.isope.org/publications Lift of a Rotating Circular Cylinder in Unsteady Flows Stefan Carstensen* and Xerxes Mandviwalla DHL, Ports and Offshore Technology Department, Hérsholm, Denmark Luca Vita and Uwe Schmidt Paulsen DTU Wind Energy, Technical University of Denmark, Roskilde, Denmark A cinder rotating in steady current experiences lit known as the Magnus effect In the present study, the effect of waves on the Magna effect has been investigated This situation i experienced with the novel, floating offshore vertical axis wind tarbine (VAWT) concept called the DEEPWIND concept, which incorporates a ralaling spar buoy and thereby utlizs seawater as a roller bearing. The a prior assumption and dhe results suggest thatthe lt in waves, to a frst approximation. may be represented by a formulation similar tothe wellknown Morison formula a, The force coefficients are experimentally found to depend primarily on the ratio between the surface speed of the cylinder and the outer low locity: INTRODUCTION ‘The DEEPWIND project is a collaborative research project ‘on future deep-sea wind luzbine technologies, which ie partially {funded by the European Commission (EC) through the ub Frarae- work Programme (FF7). The DEEPWIND concept was described in Vita etal, (2008) and is illustrated in Fig. 1 I consists of a ‘VAWT (ie. Darrieus rotor) asthe energy-capturing device rigidly fixed on top of « long upright tube connected to the seabed with ‘mooring lines. The submerged part ofthe tube acts as a rotating spar buoy and ullizes seawater asa roller bearing instead of bav= ing a main bearing between the rotor and the support structure. ‘This solution solves some of the structural problems experienced bby VAWT in the past, A conrect description of all forces on the rotating spar huey is paramount in order to be able to simulate ‘the proposed wind turbine concept and to assess the performance. ‘The flow around the rotating spar buoy may be broken down into two classical topics within hydrodynamics: (1) flow around a circular cylinder in current and waves, and (2) circular cylinder With rotation in steady current. The flow around a circular eylin- der in current and waves exerts a resultant in-line force (drag) and ‘cross-flow force (lift) on the eylinder (Sumer and Fredspe, 1997). (Cylinder diameter, surface roughness and inclination all inivence the resultant force, as well as current speed and wave motion [A rotating cylinder in steady current experiences an additional {force perpendicular to the motion (lft) known ax the Magnus effect (Batchelor, 1967; Hoemer and Bors, 1975). Furthermore, ‘the wall shear stress on the eylinder surface exerts a resultant fri tion opposite the eylinder rotation. Theodorsen and Kegier (1945) five a detailed account of revolving cylinders in stagnant fluid ‘Although the flow around the rotating spar buoy may be broken down into classical topics within hydrodynamics, litle is known, thout the combined effect of current, wave and rotation, The SISOPE Member [Received November 12, 2012; revised manuserptcecived by the edi tors aly 30, 2013. The orginal version (prior fo the Gna reveed rmanuseipt) was peseted at he Twen'yseeond Intemational Oshore fd Polar Bogiecting Conference (ISOPE-2012). Rhodes, Greece, Sine 11-32, 2012 KEY WORDS: Rotating einer, waves, il, Magnus effet. Boating ver teal ane wind turbine, DEEPWIND concept ‘observed behaviour of the inertia coefficient (in-line force) of a horizontal cylinder beneath waves was attributed (Chaplin, 1984 see Sumer and Fredspe (1997) fora review) to the steady, reer culating streaming that builds up around the cylinder asthe cylin der is exposed (0 waves, However, in this case, the circulation is linked to the orbital low around the eylinder and is disrupted bby the formation of separation vorices in the wake. Rotating cylinders have alto been considered as wave energy converters ‘(Chaplin and Revzler, 1996). Chaplin and Retzer (1996) used lin ‘ar potential theory to determine the Magnus lift force of a deeply submerged horizontal rotating cylinder beneath waves, It is seen, although expressed as a torque acting on the wave energy device, thatthe Magnus lit force equates to the KuttaJoukowski theorem, swith a time-dependent velocity, By extension of the theorem of Blasus, this solution is actually found asa special ease ofa cylin der maving in an unbounded ideal fluid (Milne-Thomson, 1962), “The Kutt-Joukowski theorem, known from steady flow around a rotating body, relates the it force tothe density ofthe Muid times the cigculation times the speed of the fluid, Fy = TU, How. ‘ever, experimental results with rotating cylinder in steady curents (Hoemer and Borst, 1975) have found the lift to be dependent of, viscosity (or the Reynolds nuzmber) ‘The present study addresses the combined effect of wave, cu rent and rotation through an extensive experimental campaign concerning lit of a rotating cylinder in pure oscillatory low. The Fig. 1 Ants’ illustration of the DEEPWIND concepta2 ‘experimental results are analysed end supported by numerical sim- Ultions t0 address possible scaling effects. EXPERIMENTAL SETUP ‘The experiments were carried out ip a 35m long, 3m wide and L-m deep current fume with a catiage running on rails on {op of the fume. The water depth of the fume was maintained at (07 m, and the model spar buoy, a section of a circular eylinder, was mounted under the carriage, as illustrated in Fig, 2. Current could be directed through the fume. while waves were simulated by an oscillatory movement ofthe carriage, The global coordinate system (X,Y, Z) was arranged so X was the steam-wise direc- tion, Z was vertical, positive upwards, and Y was inthe transverse direction according to the right-hand cule, ‘The current velocity was measured with an ultrasonic current ‘meter (Minilab SD-12), while the carriage position along the length of the fume was messured with a later distance-measuring device (SICK DTS0-P2113). Daring post-processing, the position signal was dilferentiated with time to give the velocity of the car- rage, U = dX/dt. Figure 3 shows a close-up of the motel. The cylinder. whieh ‘had a diameter D= 160 mm, was split in thzee parts, The mid dle part, which had a length of 0.4 m, was the measuring section, ‘while the bottom and top parte were dummy sections. The two Fig. 2 Schematic drawing of model submerged rotating cite lar cylinder mounted via ball bearing and support srame to the carriage Fig. 3 Submerged rotating circular cylinder Lift of a Rotating Circular Cylinder in Unsteady Flows ‘dummy sections were rigidly fixed toa drive shaft, while the mea- suring section was suspended in two 2-component force gauges (DHI-205/30). One force gauge was placed inside the bottom ‘dummy section and the other was placed inside the top dummy section, One end of each ofthe two force gauges was rigidly fixed to the drive shaft, while the other end of the force gauges held the measuring section, A small gap between the top and bottom «ap of the measuring section allowed for sight movement of the ‘measuring section relative tothe drive shaft so thatthe displace~ ‘meat could be measured by the stain gauges onthe force gauges ‘The measuring section was made rough by zluing sand, wilh « smcan grain size dsp =0:3 mm, onto the eslinder surface, A S-em band at the bottom of the tap dummy section and at the top of the bottom dummy section was similarly made rough. The re tive roughness expressed with the equivalent sand roughness was ‘estimated (0 be k,/D=4- 10 ‘The cylinder drive shaft was mounted through bull beatings to an adjustable support frame, which allowed adjustment of the ‘vertical postion ofthe cylinder inthe water column, The support frame consisted of to box profiles (placed ontop ofthe caiage), four M20 threaded steel rods and two 200 mmx 400 mm 15 mms steel plats assembled with nuts and washers. The ball bearings holding the deve shaft were positioned inthe cent of each of the steel plates ‘The motor contolling the eylinder rotation was placed next (0 ‘the eylnder dive shaft between the two steel plates of the support frame. The motor gearing was adjusted to give a rotational speed ‘ofthe metor driveshaft berween 20 rpm and 110 rpm depending fon the voage supplied to the motor. The gearing between the rotor drive shat andthe cylinder drive shaft was 32/32 = 1 i. the angular postion of the motor is identical to the angular posi- tion of the eylinder(fotation around the vertical axis, Z). This allowed the angular postion of the cylinder and the rotational speed to be measured atthe motor drive shaft with two contns- ‘ous potentiometers (Contele WAL305). During postprocessing, the two outputs were combined to achieve a 360° measurement range, and a crossing-analysis of the position signal was used to determine the rotational speed ‘The output from the two force gauges, (Fy: Fy) and (Fs Fy) respectively given inthe local coordinate system rotating with the cylinder, was re-aleulated during post-processing to the global coordinate aystem (Figs. 2 and 4) using the angular position of the eualr eid. lee Te ]=[8G) cose) | ATE []-a[P22]-[ee AEE] © Osstaary Dow U~UpsinaevT) (Cross foros) Fx ine fore) Fig. 4 Definition sketch for asillatory low with cylinder rotationJournal of Ocean and Wind Energy. Vol. 1, No. 1, February 2014, pp. 41-49 4B where Fy isthe resulting in-line force, Fy isthe crost-low (lif) {force and Rz isthe rolation matin. NUMERICAL MODEL ‘The mumerical model solving the incompressible Reynolds- averaged Navier Stokes equations (Eg. 2) used the OpenFOAM toolbox version 17.1 Gnipiwwwcopentoar org): au uy tap, 8 (au my) a 95, pas, abs) a] o ‘combined with the local continuity equation: au, o me where u; are the mean (phase-resolved) velocities, x, are the Cartesian coordinates, pis the pressure, v is the kinematic vis~ cosity, 7, is the Reynolds stress tensor and Fy, represents body forces used to drive the flow (see below). To close the system of ‘equation, the SST k-w-turbulence model was used with standard tuning coefficients (Menter, 1994). The numerical model is essen tially similar to that described and validated by Fuhrman et al (2008). ‘The ahove equations ate tolved in twa dimensions, subject to the following boundary conéitions: The cylinder surface is con- sidered as a friction wall with no-sip boundary conditions, ic all velocity variables are set to zer0 or tothe rotational speed at the surface of the cylinder. Velocities at the inlet and outlet are imposed « zero gradient boundary condition. The turbulent kinetic ‘energy, Ky, is imposed zero gradient boundary conditions except at the inlet boundary. Here a fixed value is imposed to specify the turbulence in the incoming flow. Wall roughness on the cylinder is handled with standard wall function for the eddy viscosity, v, and the specific dissipation rate of the turbulence, ap. ‘The equations are solved using a PISO algorithm, and the resulting Poisson equation for pressure is solved with zero gra dicot condition atthe eylinder wall and a fixed total pressure at the inlet and outlet. The computational grid (Fig, 5) is defined bby an inner (cylinder) radivs and an outer radius (combined inlet Fig. 5 Plan view of computational grid (N’ 54720) and outlet boundary). Within these, the computational cells are strotched witha progression of 1.03 inthe radial diction to pro Vide adequate resolution of the boundary layer forming on the cylinder ‘The flow is driven with an oscillatory body force, Fy, in the X-tirection (/= 1) “ where U is the freesteam velocity defined in Eq. 5, and Fup 0) is a global forcing term allowing any mean flux. ‘The CED simulations were performed in 2D. From flow axound non-roaling circular cylinders, it is known that the vortex shed- ding in the turbulent wake regime occurs in cells along the length ‘of the cylinder: Therefore, the maximum resultant force acting on, the cylinder over the length of a single cel (as calculated here) ‘may be larger than the force acting on the cylinder over its total length. However, for small Keulegan-Carpeater aumbers, the cor relation Iength of the vortex shedding wil be several multiples of the cylinder diameter (Sumer and Freésge, 1997). ‘TEST CONDITIONS ‘Table 1 and Table 2 summatize the test conditions for pure ‘oscillatory flow experiments. Note that U,. in the tables is the amplitude of the free-stream velocity defined by: U=U,sin@aft +e) 6 where f= 1/T is the wave frequency (T' is the period of the oscillatory motion) and g is the phase lag. The amplitude of the ‘orbital motion is calculated from: UT ao o u Ta Ke wpe N os) ) om) 0 om 0 020° «17400722 32-10" 8085 023-300 016 43-36-10" 30-774 026° 214 Old 344110 29-75 028 247 01943 44-10" 974 032 2d 022 43 SLO 28 1 032° «17402642 62.10" ITS oad 455032125 70-10" 64 1 9, SIS 0.36 141 70-10" 22784 O44 576 040 158 70-10" 62 1 oat 636044 174 70-10" 754 044 697049 19170-1062 1 oat 757 053-208 70-10" 30-764 044 820 0ST 224 70-10" 62 1 as 880061 241 JO." 277 046 21404662 F410" 19-65 050 402 098 «12479-1065, 1 053-247 069 $285.10" do~69 056 © 402125140 89.10" 2 1 062 402154 156 99-10" 60 1 ‘Table 1 Test conditions physical model experiments: pure oscil latory flow; cylinder diameter D= 160 mm. N is the number of tests performed in the range reported for the cylinder rotation, ope.44 uo Ta KC Re afte N @ @ @ oO 0 fom) lie 551006810 15,20. L678 20-21 9.6-10° 15,40 19798380320” 30, 228 1040)” eT 151 41-42 10-10? 10 1 2q7 18S T4780" 10 1 228 2200 ks 1 266 198 85 88 16-107 10 1 296 200 96 99 18.10" 10 1 338246 134138 2.0.10 10 1 ‘Table 2 Test conditions, numerical modelling: pure oeillatory flow; eylinder diameter D= 6,0 m. isthe number of tests pet= formed in the range reported forthe cylinder rotation, «0/27. ‘The Keulegan-Carpenter number and the Reynolds number ate defined as: een? Ua * DID ° Rea wee ®) respectively, where » is the kinematic viscosity (0.01 em? =~) ‘The water temperature was 20 °C. The cylinder rotation may be expressed as the ratio of sur speed and the free-stream velocity (the speed ratio) oP oR e wT, where R is the cylinder radius and o is the angular frequency of the cylinder In ‘Table 1 and Table 2, the range of angular frequency tested has been indicated. ‘A number of validation ests were performed for the physical ‘model experiments with steady current: The current speed was in the range U, = 0.06~0.35 m/s, and the angular frequency of the eylinder was in the range w/22 =20~80 rpm. The Reynolds number for the current, defined as: bu, ao) was i the range 1:10*~6- 10, and the speed ratio, R/U, . was between O and 5; RESULTS Steady Current For clockwise rotating circular cylinder in steady ow (see sketch in Fig. 6), the velocity of the upper surface ofthe cylinder isin the same dicection as the free-sweam velocity, Separation is thereby delayed on the upper surface, whereas it occurs earlier fon the lower surface, As a result, the pressure distribution on the cylinder surface is altered when the rotation is present. Pressure is reduced on the upper surface and increased on the lower surface, causing a positive net lift force, This effect is known as the Mag- nus effect. Rotation in the opposite direction reverses this effect and causes a negative lift force. Lift of a Rotating Circular Cylinder in Unsteady Flows 1 2 3 4 5 6 oRU, Fig. 6 Lift and drag of «rotating cylinder as a function of speed ratio: Magnus force. Circles and squares: present data: lines: Hocrner and Borst (1975), data band of Cy, ateibuted to Reynolds number, The nondinensional mean lit fre (oF it coefien) is seta by Fy You? a and the mean dag coefficient, Cp, is defined in the same way ‘based on the in-line force, Fy. Figure 6 shows the messuted lift and drag cocflicient ae a function of the speed ratio, R/U. [Experimentally measured lif and drag coefieient for Reynolds numbers between 40,000 and 660,000, reported by Hoerner and Borst (1975), hes also bees included in Fig. 6. The figure shows a good comparison between the present data and the data reported by Hoemer and Borst (1975). The discrepancy of lift coetficients at higher speed ratios is expected to be related to differences in ‘Reynolds number and cylinder roughness, as detailed below. The data band reported by Hoerner and Borst (1975) forthe lift icient is attributed tothe range of Reynolds numbers, Specti- cally the negative lift is reported o occur for small speed ratios in the transition (or eritical) Reynolds number range of 10° to 510%, ‘The present results are obtained for a rough cylinder, Cylin- der roughness has the elfect of reducing the transition Reynolds number range (Achenbach and Heinecke, 1981). Because of the reported transition Reynolds number range and the lack of infor- ‘mation about roughness conditions, itis concluded thatthe data reported by Hoerner ang Borst (1975) are for smooth cylinders ‘This may explain why the present results indicate a negative lift {or small speed ratios although the current Reynolds numbers are smaller than those listed by Hoemer and Borst (1975). Note that the highest current Reynolds number in the present experiments will be associated with the smaller speed ratios. For subcritical Reynolds numbers. the drag coefficient takes the value obtained inthe ease of smooth cylinders imespectve of cylinder roughness (Achenbach and Heinecke, 1981). Generally, Re, is in the sub- eritical regime, and a good comparison between the present data and the drag coefficient reported by Hoerner and Borst (1975) is observedJournal of Ocean and Wind Energy. Vol. 1, No. 1, February 2014, pp. 41-49 45 Potential flow theory predicts zero drag force (Cip = 0) and Lin force given by: TU, =pQroR'\U, (a) R the KuttsJoukowski theorem. Hence, potential ow theory pe its & it oetiient given by pareR WU. eRU? wT 7 ama (3) ‘Thie theoretical value is much higher than experimental results suggest (Fig. 6) a discrepancy that is primatily due to viscosity However, the measured variation of Ci, with respect to the speed ratio is (o zome extent in accordance with the theoretically pe dicted Finear variation = at least over a certain range of speed Oscillatory Flow ‘The forces on a rotating cylinder moving in an ideal Muid in ‘the postive direction of the X-axis ae given by (Milne-Thomsoa, ay, as) ‘where A= 7R?, The in-line force isthe hydrodynamic mass force ‘of the Morixon formation, while the cross-flow force exprestes the KuttaJoukowski theorem, ‘Figure 7 shows the measured ensemble-averaged cross-flow force at two different KC numbers: KC = 4,3 and KC = 140. ‘The two examples aze representative of the situations KC < 10 and KC > 10, respectively. Also included in Fig. 7 are the measured velocity and accel- ‘zation. The graphs show that the cross-llow force variation is ‘proportional t the velocity for KC « 10, as the Kutta-Toukowski theorem suggests, while Fis proportional to the velocity squared, LUlUI, for KC > 10, Particularly for KC «10, there is a nots able phase shift between the velocity and the cross-low force. ‘The phase shift is most clearly seen as the phase difference in the zero-crossing between the cross-flow force and velocity. An, ‘explanation for thie phenomenon may be that the inertia com= pponent of the pressure around the cylinder (Sumer and Fredsoe, 1997) has a non-zero contribation to the crossflow force. The ‘non-zero contribution from the ineria component to the cross flow force would suggest an asymmetry between the upper part ‘of the cylinder andthe lower part ofthe eylinder originating from, viscosity effects in combination withthe ciculation or boundary layer separation, Following this line of thought, aa equation for the cross-flow force may therefore be formulated as: A= pAC pot +p AC 06) Where the coefficient, Cy, has been introduced to handle tietional effects, ie, in potential flow Cy =? and Cy, = ‘Expressing the KuttaJoukowski theorem, the frst term in Bq, 16, as in the steady flow around a rotating cylinder: pAC aU = 4pC,DUIU an ives another formulation for the crost-flow force, namely $eCrDUI+ PAC cs) 20 2 oxen v — a 3 é og 3 aE reel —— petites a ‘90 480 aS 9S TB 225 2+ (de) “a (wKc- 40 20 z €o 2 2 3 ay é nessa rss 4 yo S045 90135 180 225270 ats 6 (de) Fig. 7 Ensemble-averaged crost-flow force at different KC num- bers: (a) KC=4.3, and (b) KC = 140 [Note that the velocity-squared term in Eq, 11 is written in the orm UIU to ensure that the force is always in the dtection of, ‘the velocity. rom Eq. 17, it is found that Cy can be described by: cll oR Ce or as) In the steady flow around a rotating cylinder, it isnot important ‘which formulation for the cross-flow force is used, because the velocity is a constant. In oscillatory low, it is. however, important ‘hich formulation is used if Cy and C, are defined as constants lover one wave period, ‘The formulation with C; (Eq, 16) asrumes that the cross-flow force is a function of the velocity. U, with a small conection for the hydrodynamic mass, while the formulation with C, (Bq. 18) assumes that the cross-flow force is a function of the velocity, squared, U|U|, Figure 7, as previously mentioned, shows which {formulation should be used considering that KC — 4.3 is repre sentative ofthe low regime KC = 10 and KC = 14.0 is represen tative of KC = 10, Table 1 shows that there are zero teats with 4 KC number in the range §2< KC < 12.5. The change from ‘one flow regime tothe other may therefore occur anywhere in the range §2 7) in pure oscillatory flow around a non-rotatng cylinder. Vortex shedding is therefore the46 ‘most likely cause forthe change in low regime in oscillatory flow around a rotating cylinder, 2s wll be further substantiated below. ‘A tolal of 58 tests were performed with oscillatory flow around a rolating cylinder. For each test, the lift coeslicient, C, (or the KattasJoukowski coefficient, C,). and the hydrodynamic mass coeflcient, C,., have been determined using the method of least, squares For KC < 10, the formblation with Cy (Eq. 16) has been used Figure 8 shows the lift, expressed with Cy and Cy, as a function fof the speed ratio. Cy is in the order of 1 when oR/U, = 05. 1 pradually decreases with increasing speed ratio and attains the value 0.5 for R/U, =4, a value thats substantially smaller than the potential flow solution, C= 2. Even at small speed ratios, such as w/U,,=05, the measured Cy is only half of the value predicted by the potential Now solution. For w/U, < 0.5, the Kautta-oukowski coellicient is expected to increase with decreas- ing speed ratio, although with an upper limit given by the poten- tial Now soluson, Cy = 2. The inertia coefficient. Cy, appears to attain a constant Value of 02 over the entire range of speed ratios tested (05 « R/U, = 4). if the scatter in the data points is neglected, Figure 7a includes the predicted cross-flow force, ‘The predicted cross-flow force is calculated from the determined force coeticents using Eq. 16. The figure shows a good agree- rept between the measured and predicted cross-flow force. There ae, however, some soft fluctuations in the measured cross-flow force that are not captured by the proposed formulation. If such fluctuations are located around the phase of maximum accel tion (wt =0 and wr = 180), it may have a large impact on the determined inera coefficient and may explain the scatter in the data point observed for C. For KC = 10, the formulation with C, (Eq, 18) has been! used, Figure 9 shows the lift as a function of the speed rai. ‘The number of data points with KC > 10 ate limited, However, the lif coefficient, C,, shows a clear wend to increase with the speed ratio, while the inertia coefficient again scatters around the value 0.2, ‘The Kutta Joukowski coetfcient, Cy (Fig. 8), has been re- calculated to a lift coefficient, Cy, using Bq. 19. The correspond ing lit coefficient is shown in Fig. 10. The lit coefficient is, as for KC > 10, a function of the speed ratio. Indeed, the data points for KC < 10 and KC > 10 coincide (his may, however, not be 2 KC<10 is 21% E | 2g é oP 2.00 4 gos 29 aw [. at Ef tates’ re os “o 1 2 3 4 3 6 on, Fig. 8 Lift expressed with Cy and Cy of a rotating cylinder in ‘oscillatory ow as a function of the speed ratio (2- 10 force coefficient oR, Fig. 9 Lift. expressed with Cand Cy, of a rotating cylinder in ‘oscillatory dow as a function of the speed ratio (10 < KC = 24) {force coefficient oRU, Fig. 10 Lift, expressed with C, and Cy, of a rotating cylinder in oxcillatory flow as a function of the speed ratio (2 < KC = 24) lift coeficient C4, at a speed ratio wh/U,=2, for oscilla tory flow around a rotating cylinder. This isthe same value as for steady flow around a rotaing cylinder. For «R/U <2, the lilt cocllcient in the unsteady case is higher than in the steady ease. And for R/U, > 2, the lift coefficient is smaller i the unsteady ‘ease than in the steady cate. Inthe steady case for an ideal Hid, R/U, =2 corresponds to te instance when the two stagnation points coincide to form a single stagnation point atthe bottom of ‘the cylinder surface, Above this value, the single stagnation point will move off into the fluid as cither a single stagnation point ‘or two stagnation points (Batchelor, 1967). This may possibly be related to the observed dillerence in lift coetficient between the ‘unsteady and steady case. SCALING TO FULL-SCALE ‘The expected dimensions of the full-scale rotating spar buoy relevant to this stady can be summarized as: a diameter in the ‘order of 6 m to 8 m, a rotational speed in the order of 10 xpm, to 15 rpm, a rollptch of less than 15 degrees, and a drat in the ‘order of 200 ma, The rotating spar buoy is considered to be subjectJournal of Ocean and Wind Energy. Vol. 1, No. 1, February 2014, pp. 41-49 "7 to waves with a relationship between wave height and wave period riven by: 128). = 8.0 = 8, Va @) wihere By = Bog"! is assumed tobe between 3.3 and 4.0 m= (10.3 5 is expected to be related to the scatter of data points. Given a larger dataset, i is expected to dsappeat Figure 12 presents a comparison between the ensemble. averaged cross-flow force measured in the experiments and sim ‘lated with CFD at the same KC numbers and speed ratio at in Fig. 7. The comparison is good, although the simulated lift coeflcient is larger than the measured lift coeficient for these two cases. Both the experiments and the CFD simulations show thatthe cross-flow force variation is proportional to the velocity, Tor KC < 10, a8 the KuttaToukowshi theorem suggests, while F; has a higher harmonie for KC > 10. In the previous section, it ‘was made probable thatthe change in regime was due to vortex shedding. Indeed, the CFD simulations show that vortex shedding ‘occurs for KC larger than approximately 10 and not for small KC numbers, as ilustrated by the flow pattems presented in Figs. 13 and 14, Inspection of the force time series simulated by CFD. reveals that KC~10 is representative ofthe change in How regime. However, the process by which the regime change takes place ‘might look similat to that presented by Justesen (1991) for oscil latory flow around a non-rotating cylinder (see also Sumer and, Fredsse (1997). The limited analysis of vortex shedding paterns land regimes presented herein hat been undertaken (© point out ‘that this aspect of the problem would be worth exploring further, ‘The CFD simulations predicted a ft coefficient, C,, of about 48 and 3.0, respectively, while C, in the experiment was mea sued to be about 46 and 29, respectively. The lift coeficient predicted by the CED simulations is thus approximately 6% larger ‘than that predicted by the experiments for the eases shown, (KC #42; ORV, =19 2) Fy /49U,2 0 4 «9 13S 180 225270 O)KC#139; oRU, +09 s ‘ ae 2 dl a = 5. ws By ‘4s 0 4590135 0.7 10 0.8), according to Sumer and Fredspe (1997), Desp-water waves have an ellipticity inthe order of 1. IF the incline fore is affected by the eliptety, then eross-low force ‘may possibly also be impacted. The rlliptch angle i small, and itis assumed thatthe independence principle applies for KC <8 (Sumer and Fredsge, 1997). In the range 8 < KC < 20, the in- line force is influenced even for small rolpitch angles (Sumer and Fredsse, 1997), owing to the disruption of the transverse vor- tex street, The transverse vortex street or the beginning of vor lex shedding was, in the present study, found to influence the cross-flow force. The angle of alack due to the rellptch angle ‘may therefore possibly influence the cross-flow force in the range S 10, the cross-flow force variation in time has been, {ound to be proportional tothe velocity squared, UU, and it has ‘been found that it, to a first approximation, may be represented by a formulation similar to the well-known Morison formulation. ‘The lift coeftcients, C,, have been found to depend primarily on the ratio between the surface speed of the eylinder and the outer flow velocity, increasing with increasing speed ratio. Scaling to full-scale has been addressed theoretically, assisted bby CED modeling, and itis found thatthe experimental results axe representative of full-scale conditions ACKNOWLEDGEMENTS ‘The study is funded through the DEEPWIND project, which is supported by the European Commission, Grant 256769 FP7 Energy 2010 ~ Future emerging technologies, and by the DEEPWIND beneficiaries DTU(DK), AAU(DK), TUDELFT(NL), TUTREN- ‘TO{), DHI(DK), SINTER(N), MARINTEK(N), MARIN(NL), [NREL(USA), STATOIL(N), VESTAS(DK) and NENUPHAR®), REFERENCES Achenbach, E, and Heinecke, E (1981). “On Vortex Shedding From Smooth and Rough Cylinders in the Range of Reynolds Numbers 6-10" to 5-108," J Fluid Meck, Vol 109, pp 239-251, Batchelor, GK (1967). An Introduction to Fluid Dynamics, Cam bridge University Press, 615 pp. Chaplin, JR (1984). “Non-linear Forces on a Horizontal Cylinder Beneath Waves,” J Fluid Mech, Vol 147, pp 449-468, Chaplin, JR, and Retaler, CH (1996). "Prediction of the Hydrody- namic Performance of the Wave Rotor Wave Energy Device,” Appl Ocean Res, Vol 17, No 6, pp 343-347, Pubrman, DR, Fredspe, J, and Sumer, BM (2009). “Bed Slope fect on Turbulent Wave Boundary Layers: 1, Model Valida- tion and Quantification of Rough Turbulent Results," J Geophys Res, Vol 113, C03024, pp 1-16 Hoemes, SE, and Borst, HV (1975). Fluid: Dynamic Lif, Brick= town, NI: Hoemer Fluid Dynamics, 469 pp. Jusesen, P (1981), "A Numerical Study of Oscillating Flow ‘Around a Circular Cylinder” J Fluid Mech, Vol 222, ‘pp 157-196 doi:10,1017/S00221 12091001040. ‘Menter, FR (1994), “Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications” A/AA J, Vol 32, No 8, pp 1598-1605, Milne-Thomson, LM (1962). Theoretical Hydrodynamics, Mac- milan, 660 ppJournal of Ocean and Wind Energy, Vol. 1, No. 1, February 2014, pp. 41-49 49 Sarpkaya, T (1976) “Forces on Rough-walled Circular Cylinders Speeds, National Advisory Committee for Aeronautics, Report in Humonie Flow.” Coastal Eng Proc, Vol 1, No 1S. no. 795 Sumer, BM, and Fredsge,} (1997). Hydrodynamics Around Cylin- Vita, L, Paulsen, US, Pedersen, TP, Madsen, HA, and Rasmussen, Arical Sructures, Wotld Scientific, Advanced Series on Coastal (2009), “A Novel Floating Offshore Wind Turbine Concept Engineering — Volume 12, 500 pp. Proc 2009 Eur Wind Energy Conf, Marseille, France, Paper ‘Theodorsen, T, and Kegies, A (1945). Experiments on Drag ‘no. PO.276, hip/Iproceedings ewes.orglewec2009/proceedings! of Revolving Disks, Cylinders, and Streamline Rods at High index.php?page=zip.