Sie sind auf Seite 1von 23

J Mar Sci Technol (2007) 12:160182 DOI 10.

1007/s00773-006-0238-1

ORIGINAL ARTICLE

Efcient calculation of slamming pressures on ships in irregular seas


Ole Andreas Hermundstad Torgeir Moan

Received: July 4, 2006 / Accepted: November 7, 2006 JASNAOE 2007

Abstract An efcient method for calculation of the slamming pressures on ship hulls in irregular waves is presented and validated for a 290-m cruise ship. Nonlinear strip theory was used to calculate the shipwave relative motions. The relative vertical and roll velocities for a slamming event were input to the slamming calculation program, which used a two-dimensional boundary element method (BEM) based on the generalized 2D Wagner formulation presented by Zhao et al. To improve the calculation efciency, the method was divided into two separate steps. In the rst step, the velocity potentials were calculated for unit relative velocities between the section and the water. In the next step, these precalculated velocity potentials were used together with the real relative velocities experienced in a seaway to calculate the slamming pressure and total slamming force on the section. This saved considerable computer time for slamming calculations in irregular waves, without signicant loss of accuracy. The calculated slamming pressures on the bow are of the cruise ship agreed quite well with the measured values, at least for time windows in which the calculated and experimental ship motions agreed well. A simplied method for calculation of the instantaneous peak pressure on each ship section in irregular waves is also presented. The method was used to identify slamming events to be analyzed with the more rened 2D BEM method, but comparisons with measured values indicate that the method may also be used for a quick

quantitative assessment of the maximum slamming pressures. Key words Slamming Bow are Cruise ship Experiments Validation

1 Introduction Slamming pressures are important in designing plates and stiffeners in the bow are area of ships, and the slamming loads will normally also contribute to extreme hull girder load effects. It is important to be able to assess the magnitude of slamming loads in a practical and sufciently accurate manner. Design slamming pressures are usually obtained from formulas given by the classication societies. However, these formulas are empirical and therefore not necessarily valid for novel designs; hence, there is a growing need for direct calculation methods. Since the pioneering works of von Karman1 and Wagner,2 many methods have been developed for the calculation of the slamming pressure on a body that impacts the water surface with a prescribed velocity. Most approaches are based on potential theory assuming two-dimensional (2D) or axisymmetric ow. Some work has also been done on three-dimensional (3D) potential theory formulations (Chezhian3). Finally, Reynolds-averaged Navier-Stokes equations (RANSE) solvers have been applied to 2D as well as 3D problems (Schumann,4 Sames et al.5). Good agreement with drop tests has been reported for these latter methods (Muzaferija et al.6), but they require signicant computational resources. Hence, there are still no 3D slamming methods suitable for practical design calculations.

O.A. Hermundstad (*) T. Moan Centre for Ships and Ocean Structures, Norwegian University of Science and TechnologyNTNU, NO-7491 Trondheim, Norway e-mail: ole.hermundstad@marintek.sintef.no

13

J Mar Sci Technol (2007) 12:160182

161

In 1996, Zhao et al.7 presented a simplied 2D approach that is a generalization of Wagners method. In the generalized version, the body boundary condition is imposed on the real body surface and not on a horizontal line. This is a robust numerical method that is valid for quite general section shapes. Small horizontal and rotational velocity components are allowed in addition to the vertical velocity. The simplied method was extended by Zhao et al.8 to include ow separation from knuckles. Hermundstad and Moan9 discussed the difference between the two main approaches in slamming analyses; namely direct methods and k-factor methods. Examples of published work based on the two methods were given. A recent report, not referenced by Hermundstad and Moan,9 is the work of El Moctar et al.,10 who used a nonlinear strip theory to calculate ship motions and a RANSE solver to calculate the slamming pressure. Calculated bow are slamming pressures and the total forces on the bow section were compared with model test data for two different ships in head regular waves. The agreement was good, but the results were presented for very few conditions. Hermundstad and Moan9,11 and Hermundstad et al.12 presented a direct method for the calculation of slamming pressures. They used the simplied 2D boundary element method (BEM) that accounts for ow separation and which is based on the method of Zhao et al.7 Nonlinear strip theory was used to calculate the ship motions. They also included the effect of the steady wave elevation and accounted for 3D ow effects in a simplied manner. The signicance of these additional renements was demonstrated. For validation they used model test results for a 120-m Ro-Ro vessel in regular waves. Head waves and bow quartering waves were studied and the agreement with experiments was quite good. In the present article, a further development of the method of Hermundstad and Moan9 is presented and validated for a 290-m cruise ship in irregular seas. In irregular waves, there may be a large number of slamming events during a 3-h storm. In order to calculate the slamming pressures for all relevant events, a very efcient calculation method is needed. By dividing the slamming calculation method used by Hermundstad and Moan9 into two steps, a considerable increase in efciency was obtained. A method is also needed for identifying the relevant events among all the wave encounters. A simplied slamming calculation method has therefore been presented for this identication process. A description of the methods with validation is given in this article.

2 Theory 2.1 General A methodology applicable to regular waves was presented by Hermundstad and Moan.9 For irregular waves, the approach is similar, and the details of the theory will therefore not be repeated here. However, an overview of the method will be given, and some important modications will be described in the following sections. 2.2 Shipwave relative motions The ship motions were calculated using a nonlinear strip theory. The method used is similar to that presented by Wu and Moan13 and Wu and Hermundstad,14 except that the hull is considered to be completely rigid in the present case, and the only nonlinearities come from hydrostatics and FroudeKrylov forces. This is similar to the formulation denoted Nonlinear 1 in Wu and Hermundstad.14 The relative vertical velocity, V(t), was calculated by considering the vertical velocity of the ship relative to the vertical velocity of the incident water surface. When the ship had a forward speed, U, we included the component of U that acts in the section plane. Hermundstad and Moan9 presented a simplied method to include the waves caused by the steady forward motion of the ship. They showed that the relative vertical velocity could be corrected to account for the steady wave by the following formula: Vtot (t ) = V (t ) 1 + s d (1)

Here, zs is the steady wave elevation near the relevant ship section when the ships draft is d. By calculating the steady wave elevation for a number of drafts, the rate of change of zs with respect to the draft can be evaluated for all local section drafts. During a slamming event, the local draft may change from zero and reach a value that corresponds to the position of the weather deck or to the highest position where slamming is evaluated. We used the potential theory method of Zhao and Faltinsen15 to calculate zs(d). The effect of this renement will be demonstrated for a cruise ship. 2.3 Identifying critical slamming events The ship will encounter a large number of waves during a 3-h simulation. To avoid doing slamming calculations for each wave encountered, we needed a criterion for

13

162

J Mar Sci Technol (2007) 12:160182

selection of events that were expected to give signicant slamming. An estimate of the local slamming pressure on each section during each event was obtained by using the following equations: p= p= 1 2 V 2 4 tan 1 2 V 2 2 4 tan
2

for > 15 for < 15

(2) (3)

Equation 2 was derived for a wedge with deadrise angle b cutting through the water at constant velocity V, based on Wagners2 theory. The peak pressure is often written p = kV 2, where k is the slamming coefcient. According to this notation, Eqs. 2 and 3 correspond to using k = p 2/(4 tan2 b) for b > 15 and k = p 2/(4 tan b) for b < 15. In the present analysis, the terms in parentheses in Eqs. 2 and 3 were evaluated for each time step during the slamming event. The instantaneous value of the relative vertical velocity, V, and the angle, b, between the section surface and a horizontal line at the point of bodywater intersection were calculated. This gave a time series for the local pressure peak as it traveled across the section surface. If this local pressure never exceeded a given limit, pcr, during the event, the section was excluded from the slamming event. If all sections were excluded from an event, the slamming event was not considered. The process was repeated for each possible event. Depending on the choice of pcr, the number of slamming events to consider for further analysis could be signicantly reduced. The use of this method will be demonstrated later in this article. 2.4 Slamming pressure calculation with increased efciency The slamming calculation method presented in this article is in principle identical to the simplied 2D formulation presented by Zhao et al.,7 except for a few modications. The method was modied by Hermundstad and Moan9 to allow for a specied initial draft, and in the present work we have also included the hydrostatic pressure in the calculations. Moreover, the present method uses symmetry and antisymmetry to increase efciency when the hydrodynamic boundary value problems are solved. With these modications, the method is fast, and the analysis of a slamming event for one section takes 2030 s on a conventional PC. However, in a 3hour storm there may be of the order of 500 slamming events to consider, and several sections will normally be analyzed for each event. To accomplish this within a reasonable amount of time, the method needs to be even

more efcient. An essential feature of the hydrodynamic boundary value problems that are solved for each waterline of the section is that they are linear in the body velocities. Hence, we can solve the problems for unit velocities and multiply the velocity potentials by the real velocity before the pressure is calculated. In the original method of Zhao et al.,7 the pressure was calculated immediately after the velocity potential calculation, i.e., before proceeding to the next waterline. In the present method, we rst calculate the velocity potential for all waterlines and then store the results in a le. These stored results can later be used together with the real relative velocities obtained in any wave condition. For head seas, this procedure can be used without any loss of accuracy, provided that the section always emerges completely prior to each slamming event. If the section does not always emerge completely, it is necessary to calculate the velocity potential for a number of relevant initial drafts and store that information. In order to calculate the slamming pressures for a given initial draft, the stored results are simply read for the draft that is closest to the actual draft. The higher the number of drafts that have been included in the precalculations, the smaller is the loss of accuracy. In oblique seas, there will be several different initial roll angles. Hence, precalculations with a set of relevant initial roll angles need to be performed. If the different initial drafts and initial roll angles used during the precalculations cover the entire range of possible values, and if the resolution is sufciently high, then the precalculations need to be done only once for a given ship. The database of precalculated results can then be used with time-domain ship motion simulation data for any combination of sea state, ship speed, and loading condition. The efciency and accuracy of using precalculated velocity potentials will be demonstrated later in this article. 2.5 Processing of section geometry The above slamming calculation method can be used only if the length of the section waterline increases monotonically at a nite and nonzero rate as the section is being submerged. In other words, horizontal and vertical portions are not allowed, and thus sections that contain a bulb cannot be handled. Since slamming-exposed sections in the bow of a ship often contain a bulb, it is important to nd a rational way to modify a section so that the requirements are fullled. Zhao et al.8 modied the method of Zhao et al.7 to allow for ow separation. A separation point could be dened on each side of the section. In the present article, the method of Zhao et al.8 was used to automatically modify the slamming sections. An example is shown in Fig. 1. We assume that separa-

13

J Mar Sci Technol (2007) 12:160182


28 26
30 Original geometry Modified geometry 25

163

24 22 20 18

1 2 3 4

S2
20

16 14 12 10

15

R1
10

6 4 2 0 20 10 10 20

S1
0 0 5 10 15 20

Fig. 2. Body plan of the cruise ship, showing the position of the four panels. In this body plan, all four panels are drawn on the same side. On the real model, panel 4 was located on the port side, while panels 13 were located on starboard side Table 1. Main particulars of the cruise ship Parameter Length overall Length between perpendiculars Breadth molded Draft fore Draft aft Approximate value (m) 290 242 36 7.7 8.6

Fig. 1. Typical section with a bulb: original and modied geometry. Units: m. S1, rst separation point; S2, second separation point; R1, reattachment point

tion will occur at a point where the local deadrise angle of a smooth convex surface exceeds a certain limit, bS. Moreover, separation will occur at a sharp corner, i.e., where the change of slope from one segment to the next exceeds a critical value, bC. In Fig. 1 and in the rest of this article, we have used bS = 80 and bC = 20. The section geometry is processed from keel to top. When a separation point is found, the geometry above the separation point is replaced by a line that resembles the surface of the separated ow. The formula used to dene the line was used by Zhao et al.,8 and it depends on the local geometry just below the separation point. In Fig. 1, the rst separation point is S1. We continue to process the section along the new separation line until it intersects with the original geometry. If it never intersects, we truncate the separation line at a vertical position that is N times higher than the highest point on the original geometry. In the present calculations we use N = 1.5. For the section in Fig. 1, the rst separation line intersects with the original geometry at R1. At this point, the ow is assumed to reattach, and we continue to process along the original geometry. For the present section, no more separation points are found before we reach the top of the original geometry. The top of the geometry is treated as a sharp corner, hence it is dened as a separation point, S2, and a new separation line is generated. This line is truncated when it reaches a height of 1.5 times the vertical position of the highest point on the original geometry, according to the rule described above. The process is now nished, and the original geometry is replaced

by the modied geometry described by the dotted line in Fig. 1. This modied geometry is used in the slamming pressure calculations. The calculated pressure is set to zero above the highest separation point. Horizontal parts of a section are also modied by introducing a small deadrise angle of about 3. If the section has a nonzero initial roll angle, the section is rotated before the processing starts. In that case, and for asymmetric sections, both sides of the section are processed.

3 Model tests 3.1 Ship characteristics The present method was validated by comparing its results with the available data from model tests of a 290-m cruise ship at low speed in head seas. The experiments were carried out by Fincantieri and MARINTEK16 within the framework of the SAFENVSHIP EUREKA !2835 project, related to methodology to be used for ever-safer passenger ship design, pertaining in particular to hydroelasticity activities. Excerpts of the main particulars and body plan are given in Table 1 and Fig. 2, respectively. The scale ratio was 36.2.

13

164 Table 2. Approximate location of the center of the slamming panels

J Mar Sci Technol (2007) 12:160182

Panel no. 1 2 3 4

Distance from AP (m) 250 246 240 243

Distance from CL (m) 8.3 3.6 4.0 7.5 (port side)

Distance from BL (m) 20.0 15.0 13.0 16.5

AP, aft perpendicular; CL, centre-line; BL, baseline

Table 3. Test program in irregular head seas Condition ID U0H7 U0H9 U6H7 U6H9 U9H7 U9H9 Run numbers 1216 1217 1205, 1211, 1208, 1213, Speed U (knots) 0 0 6 6 9 9 Hs (m) 7 9 7 9 7 9 Tp (s) 11 13 11 13 11 13

1206 1212 1209, 1210 1214, 1215

Hs, signicant wave height; Tp, peak period

Fig. 3. Panels 1, 2, and 3 just prior to a slamming event

3.2 Instrumentation For the present validation, the most important part of the instrumentation is the four circular slamming panels in the bow are and the optical system measuring the ship motions. The panels have a full-scale diameter of 2.35 m, and their position is shown in Fig. 2 and Table 2. Panels 1, 2, and 3, which are located on the starboard side, are depicted in Fig. 3. Each panel is thin and stiff and is mounted on a force transducer. The other end of the transducer is connected to the bottom of a heavy steel capsule, which in turn is rmly connected to the hull. A thin, elastic rubber membrane covers the small gap between the circular panel and the hull surface. The natural period of the panel gauge (panel/transducer system) is not known exactly, but tests with similar panel gauges at MARINTEK indicated that the natural period is almost an order of magnitude lower than the rise-time of the slamming load. We therefore expect the signal from the panel gauge to give a good estimate of the applied load. When the force measured by the panel gauge is divided by the panel area we obtain the average slamming pressure on the panel, and this value will be used later. Compared to local pressures obtained from small pressure cells, this space-averaged panel pressure is more relevant from a structural design point of view. The model was segmented, and shear forces and moments were measured at three longitudinal positions. The natural frequency of the two-node vertical vibration

mode was 0.86 Hz (full-scale) in calm water. All measurements were sampled at 500 Hz and ltered at 200 Hz (model scale values). The model was tested in the extension of the towing tank at MARINTEK. This extension tank is 85 m long, 10.5 m wide, and 10 m deep. During the tests at 6 and 9 knots, the model was self-propelled with two conventional propellers. During the zero-speed tests, the model was kept in place by four relatively exible springs. 3.3 Test program The model was tested in two sea states and at three speeds, U. Both head seas and following seas were tested, but for bow are slamming, we only consider head seas. This leaves us with a matrix of six conditions, as shown in Table 3. All tests were performed in irregular waves with a JONSWAP spectrum with peakness parameter 3.3. The duration of each test was 1 h (full-scale). At 6 knots, two runs along the tank were needed to achieve 1 h of data, while three runs were needed at 9 knots.

4 Results and discussion 4.1 Reproducing the measured irregular waves In this section, calculated ship motions and slamming pressures will be compared with measured data from irregular waves. Because we have experimental data corresponding to only 1 h of full-scale operation, there will be signicant statistical uncertainties relating to the waves. To eliminate this uncertainty, it is desirable to use

13

J Mar Sci Technol (2007) 12:160182

165

the same time series of the wave elevation during the calculations as during the tests. During the tests, the wave elevation was measured by a probe that was xed to the towing tank carriage. The carriage followed the model, but due to surge motions, the distance between the carriage and the model was not constant. The mean distance between the wave probe and the center of gravity of the ship was about 278 m (fullscale). The ship motion simulation program takes the frequency, amplitude, and phase of a number of wave components as input. We could obtain these data by taking the fast Fourier transform (FFT) of the measured wave time series. Then the frequencies were transformed from the encounter frequency to the wave frequency by using the mean measured speed of the carriage. The phase angles were transformed so that they relate to the ships center of gravity and the results were written to a le. The ship motion calculation program transformed the circular frequencies from wave frequencies to encounter frequencies and generated the wave elevation at each station by adding the wave components using the appropriate kx value, where k is the wave number and x is the longitudinal coordinate of the section. In the simulations, the ship speed was specied and kept constant. For accurate results, the specied ship speed should be the same as the mean measured speed of the carriage. 4.2 Required number of wave components During the third run in condition U9H7 (speed = 9 kn; Hs = 7 m; see Table 3), at t = 442 s a very high and steep wave hit the wave probe. The entire time series of the wave elevation during this run is shown in the upper part of Fig. 10, and it is clearly seen that the wave is outstanding. The wave had a crest height of almost 13 m, followed by a nearly 7-m-deep trough. Hence, the crest-to-trough height is almost three times larger than the signicant wave height, which was 7 m. We will now investigate how many wave components are needed to reproduce the wave elevation, the ship motions, and the slamming pressures during this extreme event. Fig. 4 shows the wave and ship motions calculated by including wave components with frequencies up to 0.15 Hz, 0.25 Hz, 0.50 Hz, 0.75 Hz, and 1.0 Hz. The wave at the probe is shown in the top row, with a close-up of the wave crest to the right. It is seen that the wave elevation and vertical velocity converge when components up to 0.75 Hz were used. Quite good results were also obtained when including only frequencies up to 0.5 Hz. Far fewer frequencies were necessary to obtain convergent results for heave, pitch, and bow vertical velocity: a cut-off of 0.15 Hz was sufcient. The present run has a 1200 s duration, and a

cut-off frequency of 1.0 Hz means that 4750 frequency components were used in the time-domain ship motion simulations. This slows down the simulation, and it is therefore convenient to be able to use a lower cut-off frequency. To ensure convergent results in the present calculations we used a cut-off of 0.25 Hz for the ship motion simulation. Figure 5 shows the slamming pressures during the encounter with the extreme wave. It was seen in Fig. 4 that there is a small time lag between the measured and calculated ship motions, and this caused the calculated slamming pressures to occur 11.5 s before the measured pressures. We observe that a cut-off frequency of 0.5 Hz gives sufciently convergent pressures in this case. To ensure convergent results, we will use a cut-off frequency of 0.75 Hz for all conditions when the relative velocity is calculated. 4.3 Calculated and measured ship motions In Fig. 6 to Fig. 11, the complete time series of the measured and calculated results for the six conditions shown in Table 3 are presented. The top graph in each case shows the wave elevation at the fore perpendicular, based on the FFT analysis described in Sect. 4.1. The subsequent two graphs give the heave and pitch motions, and the last four graphs show the slamming pressure on each of the four panels in the bow are. The standard deviations of the heave and pitch motions are plotted in Fig. 12. For comparison, the gure also contains the results from linear calculations. It is evident from the experimental data that the heave and pitch motions are largest for the higher of the two sea states. The motions also increase as the forward speed of the ship increases from 0 to 9 knots, except for pitch in the lowest sea state, which is almost constant. These trends are also captured by the calculations. Generally, the experimental heave motion is slightly larger than predicted, while the experimental pitch motion is slightly smaller. In general, pitch agrees better with the measured data than does heave. In the calculations, the ship has a constant forward speed; in the experiments, each group of large waves caused a signicant speed reduction. An example is given in Fig. 13, in which the ship encounters a group of large waves during the second run for condition U9H9. In the calculations, the ship proceeds through these waves with no speed reduction, and when the group of large waves has passed, the physical ship model is lagging behind the numerical model. After large waves there will normally be a period with relatively small waves, and due to the difference in position relative to the waves, the agreement between the calculated and measured motions will

13

166

J Mar Sci Technol (2007) 12:160182

Wave elevation at probe [m]

Wave elevation at probe [m]

10

13 12 11 10 9 8

Experiments Regenerated. Regenerated. Regenerated. Regenerated. Regenerated.

Cut-off Cut-off Cut-off Cut-off Cut-off

0.15 Hz 0.25 Hz 0.50 Hz 0.75 Hz 1.00 Hz

-5

430

435

440

445

450

455

460
Wave vertical velocity at bow [m/s]

441.0

441.5

442.0

442.5

443.0

443.5

444.0
Cut-off Cut-off Cut-off Cut-off Cut-off 0.15 Hz 0.25 Hz 0.50 Hz 0.75 Hz 1.00 Hz

Wave elevation at bow [m]

10

6 4 2 0 -2 -4

Regenerated. Regenerated. Regenerated. Regenerated. Regenerated.

-5

445

450

455

460

465

470

475

480

445 6

450

455

460

465

470

475

480

2 1 0 -1 -2 445 6 4 2 0 -2 -4 445 450 455 460 465 470 475 480 450 455 460 465 470 475 480

Pitch [deg]

Heave [m]

2 0 -2 -4 445

Experiments Calculated. Cut-off Calculated. Cut-off Calculated. Cut-off Calculated. Cut-off Calculated. Cut-off

0.15 Hz 0.25 Hz 0.50 Hz 0.75 Hz 1.00 Hz

450

455

460

465

470
Calculated. Calculated. Calculated. Calculated. Calculated.

475
Cut-off Cut-off Cut-off Cut-off Cut-off

480
0.15 Hz 0.25 Hz 0.50 Hz 0.75 Hz 1.00 Hz

Bow-Wave relative velocity [m/s]

Bow vertical velocity [m/s]

10

-5

445

450

455

460

465

470

475

480

Time [s]

Time [s]

Fig. 4. Convergence with respect to the number of wave components used for the condition U9H7. The wave elevation at the probe with a close-up of the wave crest to the right are shown (top row). Also shown are the wave elevation and wave vertical velocity

at the bow (second row), the heave and pitch motion (third row), and the bow vertical velocity and bow-wave relative velocity (bottom row)

now be relatively poor. This effect is increased by the fact that small waves normally are quite short, and this makes the responses more sensitive to the ships longitudinal position relative to the waves. During the period of small waves the physical model will increase its speed and gradually catch up with the numerical model, which travels with the mean speed. By the time the next group of large waves encounters the ship, the physical model has regained its position relative to the waves, and therefore the motions agree quite well during the next group of large waves. This is illustrated in the lower plot in Fig. 13, which shows the relative wave elevation at the bow.

The dotted lines in Fig. 12 are calculations that include nonlinear modications of the hydrostatic and FroudeKrylov forces, while the dashed lines are results from linear calculations. For the lowest sea state [signicant wave height (Hs) = 7 m], the nonlinear forces have little inuence on the heave and pitch standard deviations. However, it will be shown below that nonlinear motions have some inuence on the extreme slamming pressures for Hs = 7 m. For the highest sea state (Hs = 9 m), the nonlinear effects cause a small reduction in the calculated heave and pitch standard deviations.

13

J Mar Sci Technol (2007) 12:160182


400 300

167

Panel 1 [kPa]

200

Panel 2 [kPa]

300

Experiments Calculated. Cut-off Calculated. Cut-off Calculated. Cut-off Calculated. Cut-off Calculated. Cut-off

0.15 Hz 0.25 Hz 0.50 Hz 0.75 Hz 1.00 Hz

200

100

100

0 455.5 456.0 456.5 457.0 457.5 458.0 458.5

0 455.5 456.0 456.5 457.0 457.5 458.0 458.5

250

200

Panel 3 [kPa]

150 100 50 0 455.5 456.0 456.5 457.0 457.5 458.0 458.5

Panel 4 [KPa]

200

150

100

50

0 455.5 456.0 456.5 457.0 457.5 458.0 458.5

Time [s]

Time [s]

Fig. 5. Convergence with respect to the number of wave components used for condition U9H7. Slamming pressures at the four panels are shown during the encounter with the extreme wave at t = 455 s seen in Fig. 4

4.4 Calculated and measured slamming pressures The general trends observed in Fig. 6 to Fig. 11 are that slamming pressures increase in frequency and magnitude as the ship speed increases and when the severity of the sea state increases. It is also seen that slamming is less frequent for panel 1 and most frequent for panel 3. This is reasonable, since panel 1 and panel 3 are located furthest away and closest to the water surface, respectively. To a large extent these trends are also captured by the calculations. It is evident that the magnitude of the slamming pressures varies signicantly between slamming events. In general, a few events dominate in each condition and give pressures that are signicantly larger than the other pressures. This is particularly pronounced for panel 1, where the frequency of slamming is low. There is a clear correlation between large incident waves and large slamming pressures. The largest calculated slamming pressures are comparable to the largest experimental pressures. However, the experiments display many moderate slams that are not captured by the calculations. This phenomenon is most pronounced for the lower sea state and it increases with the ship speed. Slamming pressures are very sensitive to roll motions. It was found by Hermundstad and Moan9 that slamming pressures in the bow are of a 120-m Ro-Ro vessel varied signicantly in regular head waves due to roll. Calculations demonstrated that this effect is stronger for lower waves, and a roll angle of 2 could increase the slamming pressures on one side by 80% in moderate

waves. This effect depends on the bow are angle and the phasing between the roll motion and the relative vertical motion between the bow and water surface. An investigation of the measured roll motion of the present cruise ship revealed that for condition U0H7 the maximum roll amplitudes were of the order of 0.75. For the highest speed (U9H7), the maximum roll amplitudes increased to about 1.5. In the highest sea state, the maximum roll amplitudes increased from about 1.25 to 2.5 as the speed increased from 0 to 9 knots. A closer examination of the measured time series show that large roll amplitudes generally coincided with large slamming pressures. An extreme event happened at t = 150 s during the rst run in condition U9H9 (see Fig. 11). Roll amplitudes of 4 to both sides were measured during this event, and this coincided with very large slamming pressures, particularly on panels 3 and 4. Because the roll angle was zero in the calculations, the predicted slamming pressures will be too low in situations were the roll motion reduced the impact angle. The roll amplitudes increased with ship speed, and one would therefore expect an increasing discrepancy between the calculated and measured slamming pressures for the higher speeds in the lowest sea state. In the higher sea state, the roll motions were larger, but since the inuence of roll was smaller for larger waves, the agreement between measurements and calculations would be better in this sea state. This is in line with the data shown in Fig. 6 to Fig. 11. The heave and in particular the pitch motion is very important for the bow are slamming pressure. We have

13

168 Fig. 6. Measured and calculated results for the heave, pitch, and slamming pressure on the four panels for condition U0H7 [ship speed (U) = 0 knots, Hs = 7 m, Tp = 11 s]. Horizontal axes show the time in seconds, and all values are full-scale values. FP, fore perpendicular
8

J Mar Sci Technol (2007) 12:160182


Incident wave at FP

Wave elevation [m]

6 4 2 0 -2 -4 -6 0 500 1000 1500 2000 2500 3000 3500

1.5

Experiments Calculations

Heave [m]

1.0 0.5 0.0 -0.5 -1.0

500

1000

1500

2000

2500

3000

3500

2 1 0 -1 -2

Experiments Calculations

Pitch [deg]

500

1000

1500

2000

2500

3000

3500

80

Panel 1 [kPa]

60 40 20 0 -20 0 500 1000 1500 2000 2500 3000

Experiments Calculations

3500

120

Panel 2 [kPa]

100 80 60 40 20 0 0 500 1000 1500 2000 2500 3000

Experiments Calculations

3500

120

Panel 3 [kPa]

100 80 60 40 20 0 0 500 1000 1500 2000 2500 3000

Experiments Calculations

3500

160 140

Panel 4 [kPa]

120 100 80 60 40 20 0 0 500 1000 1500 2000 2500 3000

Experiments Calculations

3500

seen that there are some discrepancies between measured and calculated heave and pitch motions, and therefore we cannot expect a perfect agreement for the slamming pressures even for events where the roll motion is insignicant. For a better evaluation of the slamming calculation method, we can identify time windows in which the measured and calculated ship motions agree well and compare the slamming pressures. One example is given in Fig. 4 and Fig. 5, where the ship encountered an extremely high wave. Considering that there still were some discrepancies in the measured and calculated mo-

tions in this case, the agreement between measured and calculated pressures was reasonable. Two other examples are presented in Fig. 14, and the results are discussed in the following paragraphs. A group of three slamming events from the rst run in condition U6H9 are presented on the left side of Fig. 14. The measured and calculated pitch motions are in quite good agreement and the agreement is reasonable also for the heave motion. It is seen that the measured and calculated slamming pressures agree fairly well in this case. The agreement is best for the third slam, where

13

J Mar Sci Technol (2007) 12:160182 Fig. 7. Measured and calculated results for the heave, pitch, and slamming pressure on the four panels for condition U0H9 (U = 0 knots, Hs = 9 m, Tp = 13 s). Horizontal axes show the time in seconds, and all values are full-scale values

169
Incident wave at FP
5

Wave elevation [m]

-5

500

1000

1500

2000

2500

3000

3500

Experiments Calculations

Heave [m]

1 0 -1

500

1000

1500

2000

2500

3000

3500

4 2 0 -2 -4 0 500 1000 1500 2000 2500 3000

Experiments Calculations

Pitch [deg]

3500

120

Panel 1 [kPa]

100 80 60 40 20 0 -20 0 500 1000 1500 2000 2500 3000

Experiments Calculations

3500

150

Panel 2 [kPa]

Experiments Calculations

100

50

500

1000

1500

2000

2500

3000

3500

140 120

Experiments Calculations

Panel 3 [kPa]

100 80 60 40 20 0

500

1000

1500

2000

2500

3000

3500

140 120

Experiments Calculations

Panel 4 [kPa]

100 80 60 40 20 0 0 500 1000 1500 2000 2500 3000 3500

the calculated heave and pitch are very close to the empirical data. Another group features four slamming events and is taken from the second run for condition U9H9; this group is presented on the right side of Fig. 14. In this case, the calculated pitch motion is slightly out of phase with the measurements, but the measured and calculated pitch amplitudes agree quite well. It is seen that the phase difference observed for the pitch motion also occurs for the slamming pressures. This indicates, as expected, that the pitch motion is much more important

than the heave motion. Again, the agreement between measured and calculated slamming pressures is fairly good, except for the last slam in the group. The group of slams shown on the left side of Fig. 14 also contained a fourth slam, which was not captured by the calculations. The reason why these last slams are sometimes not captured is probably the speed reduction of the physical model, as discussed in Sect. 4.3. Fig. 13 shows the wave elevation at the bow, the surge motion, and the carriage speed during the group of slams shown on the right side of Fig. 14. Note that the two lower plots contain a longer

13

170 Fig. 8. Measured and calculated results for the heave, pitch, and slamming pressure on the four panels for condition U6H7 (U = 6 knots, Hs = 7 m, Tp = 11 s). Horizontal axes show the time in seconds, and all values are full-scale values
6

J Mar Sci Technol (2007) 12:160182

Wave elevation [m]

4 2 0 -2 -4 -6 -8 0 500 1000 1500

-5

Incident wave at FP
0 500 1000 1500

2 1 0 -1 -2

Experiments Calculations

Heave [m]

1 0 -1 -2

500

1000

1500

500

1000

1500

3 2

3 2 1 0 -1 -2 -3
0 500 1000 1500

Experiments Calculations

Pitch [deg]

1 0 -1 -2 -3

500

1000

1500

120 100

120 100 80 60 40 20 0

Panel 1 [kPa]

80 60 40 20 0

Experiments Calculations

500

1000

1500

500

1000

1500

250 200 150 100 50 0 0 500 1000 1500

250 200 150 100 50 0 0 500 1000 1500

Panel 2 [kPa]

Experiments Calculations

250

250 200 150 100 50 0


0 500 1000 1500

Panel 3 [kPa]

200 150 100 50 0

Experiments Calculations

500

1000

1500

150

Panel 4 [kPa]

150

Experiments Calculations

100

100

50

50

0 0 500 1000 1500

0 0 500 1000 1500

time window. We see that during the group of slams between t = 740 and t = 780 s, the ship model slows down. As discussed in Sect. 4.3, the numerical model, which maintains a constant speed, will then be located ahead of the physical model, and there will be a difference between the waves encountered by the two models. In the upper part of Fig. 13, the measured and calculated relative wave elevation at the bow is shown. Based on the measured relative wave and the heave and pitch motions, the incident wave elevation at the bow can be derived. This is plotted as black dots for positive wave ampli-

tudes. The dashed line is the incident wave used in the calculations. We see that measured and calculated results agree well for the rst three waves, but for the last event, there is a large discrepancy in the incident wave. 4.5 Nonlinear motions, steady wave elevation, and hydrostatic pressure In the calculations presented in Fig. 4 to Fig. 13, we included nonlinear modications of the hydrostatic and

13

J Mar Sci Technol (2007) 12:160182 Fig. 9. Measured and calculated results for the heave, pitch, and slamming pressure on the four panels for condition U6H9 (U = 6 knots, Hs = 9 m, Tp = 13 s). Horizontal axes show the time in seconds, and all values are full-scale values
10

171
10

Incident wave at FP

Wave elevation [m]

-5

-5

500

1000

1500

500

1000

1500

3 2

3 2 1 0 -1 -2 -3
0 500 1000 1500

Experiments Calculations

Heave [m]

1 0 -1 -2 -3

500

1000

1500

4 2 0 -2 -4

Experiments Calculations

Pitch [deg]

2 0 -2 -4

500

1000

1500

500

1000

1500

300

300 250 200 150 100 50 0


0 500 1000 1500

Panel 1 [kPa]

250 200 150 100 50 0

Experiments Calculations

500

1000

1500

300

300 250 200 150 100 50 0

Panel 2 [kPa]

250 200 150 100 50 0 0 500 1000 1500

Experiments Calculations

500

1000

1500

300

300 250 200 150 100 50 0

Panel 3 [kPa]

250 200 150 100 50 0 0 500 1000 1500

Experiments Calculations

500

1000

1500

350 300

350 300 250 200 150 100 50 0


0 500 1000 1500

Panel 4 [kPa]

250 200 150 100 50 0

Experiments Calculations

500

1000

1500

FroudeKrylov forces in the ship motion predictions. When calculating the relative velocity for slamming analysis, we have included the effect of the speedinduced steady wave elevation, as described in Sect. 2.2. Finally, after calculating the slamming pressure, we added the hydrostatic pressure. One may question whether these renements are really necessary, and we will investigate this more closely in this section. We will compare results from four different calculation methods. These are:

Method 1: Including all the renements mentioned above. Denoted Calculations (complete) on the respective gures. Method 2: As Method 1, but using linear ship motions. Denoted Linear ship motions. Method 3: As Method 1, but excluding the hydrostatic pressure. Denoted No hydrostatic pressure. Method 4: As Method 1, but excluding the correction of the relative velocity due to the steady wave elevation. Denoted No steady wave elevation.

13

172
10 10 10

J Mar Sci Technol (2007) 12:160182

Wave elevation [m]

Incident wave at FP

-5

-5

-5

200

400

600

800

1000

1200

200

400

600

800

1000

1200

200

400

600

800

1000

1200

2 1 0 -1 -2

2 1 0 -1 -2

Heave [m]

1 0 -1 -2 0 200 400 600 800 1000 1200

Experiments Calculations
0 200 400 600 800 1000 1200

200

400

600

800

1000

1200

Pitch [deg]

Experiments Calculations

-2

-2

-2

-4 0 200 400 600 800 1000 1200

-4 0 350 300 250 200 150 100 50 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 200 400 600 800 1000 1200

-4 0 350 300 250 200 150 100 50 0 0 200 400 600 800 1000 1200 200 400 600 800 1000 1200

350 300

Panel 1 [kPa]

250 200 150 100 50 0

Experiments Calculations

300

300

300

Panel 2 [kPa]

Experiments Calculations

200

200

200

100

100

100

0 0 300 250 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

300 250 200 150 100 50 0

300 250 200 150 100 50 0

Panel 3 [kPa]

200 150 100 50 0 0 200 400 600 800 1000 1200

Experiments Calculations

200

400

600

800

1000

1200

200

400

600

800

1000

1200

150

150

150

Panel 4 [kPa]

100

100

Experiments Calculations

100

50

50

50

0 0 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

Fig. 10. Measured and calculated results for the heave, pitch, and slamming pressure on the four panels for condition U9H7 (U = 9 knots, Hs = 7 m, Tp = 11 s). Horizontal axes show the time in seconds, and all values are full-scale values

13

Incident wave at FP

Wave elevation [m]

-5

-5

-5

0
4 3

200

400

600

800

1000

1200
4 3 2 1 0 -1 -2 -3

200

400

600

800

1000

1200

200

400

600

800

1000

1200

4 3 2 1 0 -1 -2 -3

Heave [m]

2 1 0 -1 -2 -3 0 200 400 600 800 1000 1200

Experiments Calculations

0
6 4 2 0 -2 -4

200

400

600

800

1000

1200

200

400

600

800

1000

1200

6 4

6 4 2 0 -2 -4

Experiments Calculations

Pitch [deg]

2 0 -2 -4

200

400

600

800

1000

1200

200

400

600

800

1000

1200

200

400

600

800

1000

1200

Panel 1 [kPa]

300

300

300

Experiments Calculations

200

200

200

100

100

100

0 0 400 200 400 600 800 1000 1200

0 0 400 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

400

Panel 2 [kPa]

300

300

300

Experiments Calculations

200

200

200

100

100

100

0 0 200 400 600 800 1000 1200

0 0 500 400 300 200 100 0 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

500 400

500 400 300 200 100 0

Panel 3 [kPa]

300 200 100 0 0 200 400 600 800 1000 1200

Experiments Calculations

0 800

200

400

600

800

1000

1200

200

400

600

800

1000

1200

800

800

Panel 4 [kPa]

600

600

600

400

400

Experiments Calculations

400

200

200

200

0 0 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

0 0 200 400 600 800 1000 1200

Fig. 11. Measured and calculated results for the heave, pitch, and slamming pressure on the four panels for condition U9H9 (U = 9 knots, Hs = 9 m, Tp = 13 s). Horizontal axes show the time in seconds, and all values are full-scale values

13

174

J Mar Sci Technol (2007) 12:160182

1.0

Heave st.dev. [m]

0.8 0.6

Hs=9m Hs=7m

Wave elevation at bow [m]

Experiments Calculations Linear

10

0.4 0.2 0.0 0


1.8 1.6

-10

-20

9
740 750 760

Rel. wave exp. Rel. wave calc. Wave elevation exp. Wave elevation calc.
770 780
6

Surge (relative carriage) [m]

Pitch st.dev. [deg]

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0

Hs=9m

-10 5 -20 4 -30 3 -40 -50 -60 750 800 850 900

Carriage speed [m/s]

Hs=7m

Surge (Exp) Carriage speed

2 1 0

9
15

Speed [knots]
Rel. wave at bow [m]

Experiments Calculations

Fig. 12. The standard deviations of the measured and calculated values for heave (top) and pitch (bottom) for three ship speeds and two sea states: Hs = 7 m, Tp = 11 s and Hs = 9 m, Tp = 13 s. All values are full-scale values

10 5 0 -5 -10 -15 750 800 850 900

The simulations in all six conditions were repeated with the three simplied calculation methods and we will compare the largest slamming pressures obtained in a 1-h period. Instead of focusing on the single most extreme pressure, we will take the average of the ve largest pressures. The results are presented in Fig. 15, where the pressures are plotted as a function of ship speed. The left side of the gure shows results for Hs = 7 m, while the right side contains results for Hs = 9 m. The experimental results are also included in the plots. One should note that the ship encounters a different set of waves at different ship speeds, even if Hs and peak period (Tp) remain the same. Hence, the extreme pressures for the different speeds may vary partly due to speed effects and partly due to variations in the incident wave. However, there is a quite clear trend showing that the extreme measured slamming pressures increase with increasing forward speed. We also see, as pointed out in the previous sections, that the pressures increase as the signicant wave height increases. The calculations generally follow these trends, but it appears that they give ex-

Time [s]

Fig. 13. Wave elevation at bow (top), experimental surge and carriage speed (middle), and relative wave elevation at bow (bottom) during the second run for condition U9H9. Because of the limited length of the wave probe, the measured relative wave elevation is truncated when the wave moves more than 5 m below the mean waterline. Note the different time scales for the top graph and the other two graphs

treme slamming pressures somewhat too low for the moderate sea state and somewhat too high for the higher sea state. An exception occurs for panels 3 and 4 in the highest sea state; the calculated extreme pressures seem to decrease when the speed increases from 6 to 9 knots. This however, could be attributed to differences in the incident waves, rather than forward-speed effects. For the highest speed, the measured extreme pressures on panels 3 and 4 were much higher than the calculated values, particularly for panel 4. From Fig. 11, we see that there are two single pressure peaks on panel 4 that are

Fig. 14. Heave, pitch, and slamming pressures during an encounter with two groups of large waves (the group shown on the right here is the same as shown in Fig. 13). Case 1 (left): between t =

200 s and t = 233 s during the rst run for condition U6H9 (see Fig. 9). Case 2 (right): between t = 740 s and t = 780 s during the second run for condition U9H9 (see Fig. 11)

13

J Mar Sci Technol (2007) 12:160182


3 2

175

3 2

Heave [m]

Heave [m]
200 205 210 215 220 225 230

1 0 -1 -2 -3

1 0 -1 -2 740 6 4 750 760 770 780

4 2

Pitch [deg]

Pitch [deg]
200 205 210 215 220 225 230

2 0 -2 -4

0 -2 -4

740

750

760

770

780

140 120

300

Panel 1 [kPa]

100 80 60 40 20 0 200
200

Panel 1 [kPa]

200

Experiments Calculations

100

0
205 210 215 220 225 230

740 400

750

760

770

780

Panel 2 [kPa]

Panel 2 [kPa]

150

300

100

200

50

100

0 200 205 210 215 220 225 230

0 740 400 750 760 770 780

140 120

Panel 3 [kPa]

100 80 60 40 20 0 200 205 210 215 220 225 230

Panel 3 [kPa]

300

200

100

0 740 300 750 760 770 780

150

250

Panel 4 [kPa]

Panel 4 [kPa]

200 150 100 50

100

50

0 200 205 210 215 220 225 230

0 740 750 760 770 780

13

176

J Mar Sci Technol (2007) 12:160182

Hs=7m Tp=11s
400

Hs=9m Tp=13s
400 300 200 100 0

Panel 1 [kPa]

300 200 100 0 0 1

Experiments Calculations (complete) Linear ship motions No hydrostatic pressure No steady wave elevation

400

400 300 200 100 0

Panel 2 [kPa]

300 200 100 0 0 1 2 3 4 5 6 7 8 9

400

400 300 200 100 0

Panel 3 [kPa]

300 200 100 0 0 1 2 3 4 5 6 7 8 9

400

400 300 200 100 0


0 1 2 3 4 5 6 7 8 9

Panel 4 [kPa]

300 200 100 0

Speed [knots]

Speed [knots]

Fig. 15. Average of the ve largest slamming pressure peaks plotted as function of ship speed for each of the four panels using the full calculation and three simplied methods. Two sea states and

a duration of 1 h were employed: Hs = 7 m, Tp = 11 s (left) and Hs = 9 m, Tp = 13 s (right)

signicantly higher than all others. A relatively large roll angle contributed to the rst of these peaks, while the second peak has not been studied in detail. The sensitivity to small roll angles and the statistical uncertainties due to variations in the incident waves make it difcult

to draw rm conclusions from the comparisons with the measured data in Fig. 15. The calculated results in Fig. 15 can more easily be compared with each other, since within each condition they are all obtained with the same wave train. The

13

J Mar Sci Technol (2007) 12:160182

177

importance of including nonlinear effects in the ship motion calculations was discussed in Sect. 4.3. It was shown that these effects have only a small inuence on the heave and pitch motions in the investigated conditions. However, it is seen from Fig. 15 that even a small difference in the ship motions can have a quite signicant effect on the extreme slamming pressures. Linear ship motion calculations yield higher slamming pressures, particularly for the most severe condition and for the panel located furthest away from the mean water surface (panel 1). The effect of local hydrostatic pressure is most pronounced for the lowest panel (panel 3). This is reasonable, since the hydrostatic pressure is proportional to the submergence of the panel during the slamming event. In general however, the hydrostatic pressure seems to contribute very little to the extreme pressures. As expected, the speed-induced steady wave elevation has no effect on the slamming pressures at zero ship speed. The effect is also very small at 6 knots, but it is noticeable at 9 knots for panels 1 and 2. Hermundstad and Moan9 demonstrated that the steady wave elevation had a large effect on a car carrier at 20 knots. For the present cruise ship at 9 knots, it appears that the effect is moderate, and for the lowest panel there is no effect at all on the most extreme pressures. A close-up of one single slamming event for condition U6H9 is presented in Fig. 16. This is the third event in the group that was presented in Fig. 14. It is seen that the slightly larger linear ship motions lead to higher slamming pressures. The effect of the steady wave elevation is small in this case, since the ship speed is only 6 knots. The hydrostatic pressure does not contribute signicantly to the peak pressure, but it contributes toward the end of the slamming event, particularly for the lowest panel (panel 3). Slamming calculations were performed only until the relative vertical velocity of the section changed sign from positive to negative. Hence, the gradual decrease in the hydrostatic pressure during the water exit phase was not included in the calculations. This is the reason for the sudden drop in the pressure at the end of the event for calculations with hydrostatic pressure. 4.6 Three dimensional ow and shipwave interaction

3 2

Experiments Calculations (nonlinear) Linear ship motions

Heave [m]

1 0 -1 -2 -3 200 205 210 215 220 225 230

Pitch [deg]

2 0 -2 -4

200

205

210

215

220

225

230

150

Panel 1 [kPa]

100

Experiments Calculations (complete) Linear ship motions No hydrostatic pressure No steady wave elevation

50

0 227 200 228 229 230 231 232

150

Panel 2 [kPa]

100

50

0 227 200 228 229 230 231 232

Panel 3 [kPa]

150

100

50

0 227 228 229 230 231 232

200

Panel 4 [kPa]

150 100 50 0

As pointed out by Hermundstad and Moan,9 one would expect a two-dimensional slamming calculation method to produce slamming pressures that were too high, proFig. 16. Ship motions and slamming pressures during the rst run for condition U6H9

227

228

229

230

231

232

Time [s]

13

178

J Mar Sci Technol (2007) 12:160182

vided that the input relative roll angle and relative velocity were correct. Hermundstad and Moan9 used results from Scolan and Korobkin17 to estimate that 3D ow effects could reduce the pressures by up to 30% for panels in the are of a 120-m Ro-Ro ship. They found that the 2D calculations, which excluded the effect of the steady wave, underestimated the slamming pressures in small and moderate waves. Since the Ro-Ro ship traveled at almost 20 knots, the steady wave elevation was important, and when it was accounted for, the 2D calculations produced pressures that were too high. For the present cruise ship, the effect of the steady wave is small, and when it is accounted for, the 2D calculations still underpredict the slamming pressures in moderate conditions. It may seem strange that the slamming pressures in moderate conditions were overpredicted for a 120-m Ro-Ro ship, whereas they were underpredicted by the same theory for a 243-m cruise ship. One possible reason is that the simplied way of accounting for the steady wave elevation overestimates the effect, so that the calculated pressures for the Ro-Ro ship were too high. Another possible reason is that the present ship has a more pronounced are, and the local deadrise angle at the position of the panels is therefore smaller than for the Ro-Ro ship. Hence, the slamming pressures on the cruise ship will be even more sensitive to small roll motions. As discussed in Sect. 4.4, it is believed that roll motions increased the measured slamming pressures in the most moderate sea state and for the higher speeds. This increase is probably larger than for the Ro-Ro ship, and it may be larger than the decrease due to threedimensional ow. This may explain why the same theoretical method seems to overestimate pressures in moderate conditions for the Ro-Ro ship and to underestimate them for the cruise ship. There may also be other physical phenomena that lead to increased pressures on the cruise ship. There may be nonlinearities in the incident waves, and the presence of the ship will cause diffraction and radiation waves. This may increase the height and steepness of the waves as they impact the bow, and thereby increase the slamming pressures. It was shown by Hellan et al.18 that the measured wave at the bow of an FPSO model was signicantly higher and steeper than the linear incident wave. The bow of an FSPO is much blunter than the present cruise ship, but the same mechanisms are probably present, but to a smaller extent. These effects will increase the slamming pressures and counteract the expected reduction due to three-dimensional ow. The effect of diffracted waves is probably largest for the small and the moderate slams, where the bow does not emerge completely. This may be another reason why there are generally more small and moderate slams in the experi-

ments as compared to the present calculations. In Hermundstad and Moan,9 the ship displayed large resonant pitch motions in regular waves, and the bow emerged completely for each wave encounter. In such conditions, and when the ship has a relatively high forward speed, the errors involved in using undisturbed linear incident waves are probably quite small. 4.7 Dynamic response and hydroelastic effects The panel gauges on the model were designed with a natural period that is much smaller than the typical rise time of the slamming load. This was done to avoid dynamic amplication of the response in the gauge. The paneltransducer system is expected to respond quasistatically, and the force measured by the transducer will then equal the slamming force on the panel. By inspection of the time signals from the panel gauges, it seems that the response is quasi-static for the large majority of slams. However, for a few slams, the panel gauge seemed to respond dynamically. An example is shown in Fig. 17, which is taken from the rst run for condition U9H9. Here, the response of panel 3 seems to be dynamic with oscillations at about 15 Hz (90 Hz model scale). The oscillations are less pronounced for the other panels. It is also seen in Fig. 16 that there are some dynamic effects in the response of panel 3. Of the 120 extreme pressure peaks used to produce the experimental results in Fig. 15, there were 15 peaks where the response displayed some dynamic behavior. Most of the dynamic response recordings occurred in the most severe condition and primarily for panels 2 and 3. Based on inspection of these recordings, the applied slamming loads have been estimated. This was done by comparing the recordings with similar measurements analyzed and reported by
800

600

Pressure [kPa]

400

Panel 1 Panel 2 Panel 3 Panel 4

200

0 157.2 157.4 157.6 Time [s] 157.8 158.0 158.2

Fig. 17. The measured slamming pressures for the four panels during the rst run for condition U9H9. Panel 3 shows some pronounced dynamic effects

13

J Mar Sci Technol (2007) 12:160182

179

Lehn.19 He modeled the paneltransducer system as a single-degree-of-freedom system and simulated the response for different applied slamming loads until the simulated response matched the measured response. Based on his results, we found a reduction factor for each of the slams featuring a dynamic response. By multiplying the recorded peak by the reduction factor we arrive at an estimate for the applied slamming load. For the slams in question, the factor varied between 0.62 and 0.9. These estimates are not very accurate, but since they apply to only a few slams, their inuence on the plots in Fig. 15 is small. 4.8 Simplied slamming analysis In the present calculations, slamming events have been identied using the simplied slamming calculations described in Sect. 2.3. A possible slamming event, where the bow moves into the water, was only analyzed if the pressure estimated by the simplied method exceeded a critical value at some time during the event for at least one section. In the present calculations the critical limit was set as low as 2 kPa to also capture smaller slams. The local pressure from the simplied method was averaged over a segment of length 1.5 m. This was done to avoid very large local pressure peaks. The time series that result from the simplied method will resemble those obtained from a 1.5-m pressure panel (sensor) that moves along the side of the ship, following the body water intersection zone. In Fig. 18, the experimental and calculated slamming pressures during the rst run for condition U6H9 are shown together with the pressures obtained with the simplied calculations. Generally, there is a clear correlation between the pressure peaks produced by the simplied method and those obtained from the rened calculations (BEM). However, the simplied calculations also give some additional pressure peaks, and these peaks are particularly pronounced for the section where panel 3 is located. Because the simplied method resembles a moving pressure panel, it produces a peak for all areas on the section where the slamming pressures are large. For the sections containing panels 1, 2, and 4, there is only one such area, and that is in the are where the panels are located. The section containing panel 3 contains a bulb (see Fig. 1), and the rounded bottom gives rise to very large pressures. From the close-up shown on the right side of Fig. 18, we see that large peaks occur prior to the second peaks, which correspond to panel 3. Inspection of time series for other conditions shows that the simplied method is well suited for determining which wave encounters that will lead to signicant slamming.

The simplied method also seems to give quite good quantitative estimates of the peak pressures, at least in the are. If only a specic location on the section is of interest, one can use the simplied calculations for that panel only. This is reminiscent of the conventional procedure when using k-factor methods (e.g., Stavovy and Chuang20). One of the advantages of the present simplied method is that it scans the entire section and returns a time series of the instantaneous peak pressure during each slamming event. It is also straightforward to return the time series of the corresponding position on the section where the peak is located. This may be useful in the design process. Hermundstad et al.12 applied a similar method to a 120-m car carrier in head regular waves. Instead of using Eqs. 2 and 3, they used the k-factor for each panel, which was calculated using the 2D BEM method. They showed that the simplied method gave pressures that were too low for panels in the are, at least for moderate slams and for panels far above the design waterline. The reason is that the simplied method does not account for the pile up of water that is generated during the slamming event. Hence, the present simplied method needs more testing to assess its applicability for various cases. The present simplied method can also be used directly in oblique seas, since the k-factors are calculated together with the pressure calculation, using the instantaneous value of the relative angle. It does not rely on a precalculated k-factor, which would need to be recalculated for each possible relative angle and then tabulated. One should note that the simplied method cannot be used to predict the total slamming force on the section, because it only gives the local peak pressure and not the pressure distribution. The 2D BEM on the other hand, predicts the pressure distribution for each time step and can be used to calculate the total slamming force. 4.9 Accuracy and efciency of precalculations The present 2D BEM calculations have been performed using a two-step method with precalculated velocity potentials, as described in Sect. 2.4. In order to show the accuracy and the applicability of this method, some calculations were also performed in which the velocity potentials were calculated for every slamming event. In Fig. 19 we have plotted the slamming pressure on panel 3 for the second run for condition U9H9. The results obtained by using precalculated velocity potentials are compared with the results obtained by calculating the velocity potentials for every slamming event. A close-up of a group of events is shown in the lower part of the gure. There is practically no difference between the results from the

13

180
300 250

J Mar Sci Technol (2007) 12:160182

Panel 1 [kPa]

200 150 100 50 0 0

Panel 1 [kPa]

Experiments Calculations (BEM) Simplified

200

150

100

50

0
500 1000 1500

200 200

205

210

215

220

225

230

300

Panel 2 [kPa]

200 150 100 50 0 0 800 500 1000 1500

Panel 2 [kPa]

250

150

100

50

0 200 800 205 210 215 220 225 230

Panel 3 [kPa]

Panel 3 [kPa]

600

600

400

400

200

200

0 0 500 1000 1500

0 200 205 210 215 220 225 230

200 300

Panel 4 [kPa]

200

Panel 4 [kPa]
0 500 1000 1500

150

100

100

50

0 200 205 210 215 220 225 230

Time [s]

Time [s]

Fig. 18. Slamming pressures during the rst run for condition U6H9 giving a comparison of the experimental results with the current calculations [two-dimentional boundary element method (2D BEM)] and simplied calculations. The right column shows a selected group of slams

two calculation methods. For slamming events where a section emerges completely, the two methods are expected to give identical results. Some minor differences may occur when a section does not emerge completely, because we then use precalculated results for an initial draft, which is generally not identical to the real initial

draft. The differences will depend on the resolution of the precalculated drafts, i.e., for how many drafts we store precalculated velocity potentials. In the present case, the sections where panels 1, 2, and 4 are located did all emerge completely for all events, and only one precalculation was therefore needed for each of these

13

J Mar Sci Technol (2007) 12:160182

181

350 300

Panel 3 [kPa]

Experiments Calculations (two-step) Direct calculations

250 200 150 100 50 0 0 200 400 600 800 1000 1200

about 7 min. This gives a total computer time of approximately 9 min. The conventional method, which does not use precalculations, took about 10 h in this case. Hence, the precalculation method is almost 70 times faster.

5 Conclusions and perspectives An efcient method for the calculation of slamming pressures on ship hulls in irregular waves is presented. Ship motions were calculated by a time-domain strip theory, and slamming pressures were calculated by a 2D BEM method based on the generalized Wagners theory presented by Zhao et al.7 The method has been modied to account for nite initial section draft and it has been divided into two separate steps. In the rst step, the velocity potentials are calculated for unit relative velocities between the section and the water, and stored in a le. In the next step, these precalculated velocity potentials are used together with the real relative velocities experienced in a seaway to calculate the slamming pressure and total slamming force on the section. This saves considerable computer time for slamming calculations in irregular waves. A simplied method for calculation of the instantaneous peak pressure on each ship section in irregular waves is also presented. The method was used to identify the slamming events to be analyzed with the more rened 2D BEM method. The methods were validated by comparison with slamming measurements on four panels in the bow are of a scale model of a 290-m cruise ship. The ship was tested at 0, 6, and 9 knots in head seas and in two sea states: Hs = 7 m and Hs = 9 m. The incident wave elevation from the experiments was used as input to the ship motion calculations. The following observations were made: The heave and pitch motions increased, and the frequency and intensity of the slamming pressures increased as Hs increased from 7 to 9 m, and as the forward speed of the ship increased from 0 to 9 knots. These trends were generally also captured by the calculations, and the calculated pressures were comparable in magnitude to the measured pressures. The measured data generally displayed more frequent slamming than predicted by the calculations. For some cases, there were also some measured slams that were much higher in pressure than the largest calculated slamming peaks. Even though the tests were done in head seas, the physical model displayed some roll motions. Bow are slamming is very sensitive to roll motions, and it is believed that small roll

Time [s]
200

Panel 3 [kPa]

150

100

50

0 730 740 750 760 770

Time [s]

Fig. 19. Slamming pressures during the second run for condition U9H9. Comparison of calculations using precalculated velocity potentials (two-step method) and calculations in which the velocity potentials were calculated for each slamming event (direct calculations). There are no visible differences between the two calculation methods in this case

sections. The section where panel 3 is located does not always emerge completely prior to a slamming event, and for the present condition we used precalculations with two different initial drafts for this section. The difference in computation time for the two methods is signicant, particularly for long simulations that contain a large number of slams. The calculated results plotted in Fig. 19 involve 68 slamming events during a 20-min full-scale simulation. The precalculation of velocity potentials for the four sections takes about 1 min on an ordinary laptop computer (2003 model, 2.2 GHz, P4 M, 1 GB RAM), while the pressure calculation for the 68 events, using the precalculated velocity potentials, takes another minute. Hence, the total computer time is about 2 minutes. When the velocity potentials are calculated for each event, the total computer time is about 60 min. The method with precalculation is therefore about 30 times faster in this case. To compare the performance of the new method for longer simulations, we performed 3-h simulations with the same ship at 9 knots in a sea state with Hs = 9 m and Tp = 13 s, represented by a JONSWAP spectrum. In this condition, there were 645 slamming events. Precalculation of velocity potentials took about 2 min in this case, while the pressure calculations for the 645 events took

13

182

J Mar Sci Technol (2007) 12:160182

motions contributed signicantly to the measured slams. Including nonlinear hydrostatic and FroudeKrylov forces in the ship motion calculations tended to reduce the heave and pitch slightly. The effect on the slamming pressures was more pronounced, at least in the most severe condition, for which nonlinear ship motion calculations gave lower pressures. Hydrostatic pressure had little inuence on the pressures in the are, except for areas in the lowest part of the are. Including the steady wave elevation in a simplied manner increased the calculated slamming pressures when the ship speed was 9 knots. For 0 and 6 knots the effect was insignicant. The simplied method for calculation of the instantaneous peak pressure on each ship section, based on the shipwave relative motions, agreed quite well with the measured panel pressures. It appears to be useful also for a quantitative assessment of peak pressures, and not only for selecting events for more rened analysis.

To learn more about the accuracy and applicability of the slamming calculation methods, they need to be validated for other ships and other conditions. A 120-m Ro-Ro vessel was analyzed in regular waves by Hermundstad and Moan.9,11 The Ro-Ro vessel has also been tested in head irregular waves and bow quartering irregular waves, and validation of the method in these conditions is currently in progress. The two-step method, which uses precalculated velocity potentials, can be used to calculate the total slamming force on each section in the bow of a ship in irregular waves. This can be used to improve the calculation of hull girder whipping responses. Currently, a 2D momentum method is normally used to estimate the sectional slamming forces. The new two-step BEM method will enable more accurate slamming calculations to be carried out with little increase in computer time. Implementation of this method in whipping calculations is another future task.
Acknowledgments. This work was funded by the Research Council of Norway and MARINTEK. The experiments were carried out by Fincantieri and MARINTEK under the EUREKA !2835 SAFENVSHIP project, which was coordinated by Fincantieri. The experiments were supported by the Research Council of Norway, the Norwegian Shipowners Association, and the Italian government.

References
1. von Karman T (1929) The impact of seaplane oats during landing. NACA Technical note 321, Washington, DC

2. Wagner H (1932) Uber stoss- und Gleitvergange an der Oberache von Flussigkeiten. Zeitschr Angew Math Mech 12(4):193235 3. Chezhian M (2003) Three-dimensional analysis of slamming. Ph.D. thesis 2003:85, Norwegian University of Science and Technology, Trondheim 4. Schumann C (1998) Volume-of-uid computations of water entry of bow sections. In: Proceedings of EUROMECH 374. Poitiers 5. Sames PC, Schellin TE, Muzaferija S, et al (1998) Application of a two-uid nite volume method to ship slamming. In: Proceedings of the 17th international conference on offshore mechanics and arctic engineering (OMAE98). Lisbon 6. Muzaferija S, Peric M, Sames P, et al (1998) A two-uid NavierStokes solver to simulate water entry. In: Proceeding of the 22nd symposium of naval hydrodynamics. Washington, DC, pp 277289 7. Zhao R, Faltinsen OM, Aarsnes JV (1996) Water entry of arbitrary two-dimensional sections with and without ow separation. In: Proceedings of the 21st symposium of naval hydrodynamics. Trondheim, pp 408423 8. Zhao R, Faltinsen OM, Haslum HA (1997) A simplied nonlinear analysis of a high-speed planing craft in calm water. In: Proceedings of FAST97. Sydney, pp 431438 9. Hermundstad OA, Moan T (2005) Numerical and experimental analysis of bow are slamming on a Ro-Ro vessel in regular oblique waves. J Mar Sci Technol 10:105122 10. El Moctar O, Brehm A, Schellin TE (2004) Prediction of slamming loads for ship structural design using potential ow and RANSE codes. In: Proceedings of the 25th symposium on naval hydrodynamics. St. Johns, Canada 11. Hermundstad OA, Moan T (2004) Numerical and experimental analysis of bow are slamming on a Ro-Ro vessel in oblique waves. In: Proceedings of the 25th symposium on naval hydrodynamics, St. Johns, Canada 12. Hermundstad OA, Moan T, Mrch HJ (2004) Motions and slamming loads on a Ro-Ro ship. In: Proceedings of the 9th international symposium on practical design of ships and other oating structuresPRADS-2004. LuebeckTravemuende, Germany, pp 487495 13. Wu M-K, Moan T (1996) Linear and nonlinear hydroelastic analysis of high-speed vessels. J Ship Res 40(2):149163 14. Wu M-K, Hermundstad OA (2002) Time-domain simulation of wave-induced nonlinear motions and loads and its applications in ship design. Mar Struct 15:561597 15. Zhao R, Faltinsen OM (2003) A nonlinear method for predicting wave resistance of ships with low block coefcients. In: Proceedings of the 8th international conference on numerical ship hydrodynamics. Busan, Korea 16. MARINTEK (2003) Comfort and whipping test of M2618A in towing tank. Report 602025.00.01 MT82 F03263 (restricted) 17. Scolan Y-M, Korobkin AA (2001) Three-dimensional theory of water impact. Part 1. Inverse Wagner problem. J Fluid Mech 440:293326 18. Hellan , Hermundstad OA, Stansberg CT (2001) Design tool for green sea, wave impact and structural response on bow and deck structures. In: Proceedings of OTC 2001. Houston 19. Lehn E (2003) On water impact loads. MARINTEK Publication P09.03 20. Stavovy AB, Chuang S-L (1976) Analytical determination of slamming pressures for high-speed vehicles in waves. J Ship Res 20(4):190198

13

Das könnte Ihnen auch gefallen