You are on page 1of 9

Numerical Prediction of Impact-Related Wave Loads on Ships

Thomas E. Schellin
e-mail: thomas.schellin@gl-group.com

Ould el Moctar
Germanischer Lloyd AG, Vorsetzen 35, Hamburg 20459, Germany

We present a numerical procedure to predict impact-related wave-induced (slamming) loads on ships. The procedure was applied to predict slamming loads on two ships that feature a ared bow with a pronounced bulb, hull shapes typical of modern offshore supply vessels. The procedure used a chain of seakeeping codes. First, a linear Green function panel code computed ship responses in unit amplitude regular waves. Ship speed, wave frequency, and wave heading were systematically varied to cover all possible combinations likely to cause slamming. Regular design waves were selected on the basis of maximum magnitudes of relative normal velocity between ship critical areas and wave, averaged over the critical areas. Second, a nonlinear strip theory seakeeping code determined ship motions under design wave conditions, thereby accounting for the nonlinear pressure distribution up to the wave contour and the frequency dependence of the radiation forces (memory effect). Third, these nonlinearly computed ship motions constituted part of the input for a Reynolds-averaged NavierStokes equations code that was used to obtain slamming loads. Favorable comparison with available model test data validated the procedure and demonstrated its capability to predict slamming loads suitable for design of ship structures. DOI: 10.1115/1.2429695 Keywords: wave impact, slamming loads, seakeeping codes, RANSE solver, ships

Introduction
Wave-impact related slamming loads can induce high stresses and cause deformation of local structural components. The accurate assessment of such loads is essential for the design of a ships structure. Classication society rules contain formulas for slamming loads, see e.g., Ref. 1 . Generally, these formulas are adequate for conventional ships, as they are based on operational experience. However, for many modern ships it becomes necessary to resort to direct computations of slamming loads. A satisfactory theoretical treatment of slamming has been prevented so far by the complexity of the problem. Most theories and their numerical procedures were applied on two-dimensional bodies; however, slamming is a strongly three-dimensional nonlinear phenomenon that is sensitive to the relative motion between the ship and the water surface 2 . Slamming is characterized by highly peaked local pressures of short duration. Hence, slamming peak pressures cannot be applied on larger areas to estimate structural response to slamming impacts. Moreover, the inuence of hydroelasticity, compressibility of water, and air pockets may have to be accounted for. Mainly because of these phenomena, potential ow methods are not well suited to accurately predict slamming loads 3 . However, recent progress has been made in the development of numerical methods that predict slamming pressures 4 . Kinoshita et al. 5 , for example, predicted ship motions and loads with a NavierStokes solver. Although no extreme load cases were presented, the applied method has the potential to compute slamming loads in extreme wave conditions. A number of research projects during the recent past were initiated to develop numerical techniques that reliably predict slamming loads on ships. Within the framework of the European research project DEXTREMEL 6 , slamming loads on the bow
Contributed by the Ocean Offshore and Arctic Engineering Division of ASME for publication in the JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING. Manuscript received June 26, 2006; nal manuscript received November 8, 2006. Review conducted by Antonio C. Fernandes. Paper presented at the 25th International Conference on Offshore Mechanics and Arctic Engineering OMAE2006 , Hamburg, Germany, June 49, 2006.

door of a generic RoRo Ferry design were investigated. To predict bow door loads on this ferry, Sames et al. 7 applied two methods that both start with linear seakeeping predictions of ship motions. Their rst procedure, using a nite volume code, relies on computed impact pressure coefcients obtained from two-dimensional water entry simulations of vertical bow sections. Their second procedure, using a boundary element code, comprises twodimensional water entry simulations of tilted bow sections. Comparison with model test measurements showed that neither method is capable of accurate predictions although the two methods are able to dene upper and lower bounds. Methods that directly solve the Reynolds-averaged Navier Stokes equations RANSE , possibly including the two-phase ow of water and air, are better able to describe the physics associated with slamming. However, the computational effort for a threedimensional RANSE method to simulate motions and loads on a ship at small, successive instances of time over a long time period appears beyond current computational capabilities. This paper presents a recently developed numerical procedure 8 to predict slamming loads. The procedure consists of rst making use of the two potential ow seakeeping codes GLPANEL 9,10 and GLSIMBEL 11 . GLPANEL selects the design waves, and GLSIMBEL determines the corresponding ship motions under these large amplitude design wave conditions. The resulting ship motions then serve as part of the input for the RANSE solver COMET 12 to yield slamming loads. The procedure was applied to predict slamming loads for two reference ships, designated Hull 1 and Hull 2. Both ships feature a ared bow with a pronounced bulb and a at overhanging stern, hull shapes typical of modern offshore supply vessels. Table 1 lists principal particulars. Slamming loads were examined at the ships forefoot. Specically, areas under the ared bow were considered. Model test measurements for one of the ships were available for comparison with computations. The favorable agreement validated this method. FEBRUARY 2007, Vol. 129 / 39

Journal of Offshore Mechanics and Arctic Engineering Copyright 2007 by ASME

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Principal particulars of the reference ships Hull 1 Length between perpendiculars Molded breadth Draft Displacement Service speed Froude number 70 m 15 m 5m 1550 t 16 kn 0.31 Hull 2 130 m 20 m 6m 5470 t 18 kn 0.26

patches panels that discretize the wetted hull. Pressures are obtained according to the linearized Bernoulli equation. Integrating pressures over the hull surface results in hydrodynamic forces and moments. Solving the motion equations yields ship motions.

The Nonlinear Strip Method


The nonlinear strip theory seakeeping code GLSIMBEL simulates large-amplitude rigid-body motions of a ship in six degrees of freedom by time-domain integration of the motion equations. Considered are forces and moments caused by gravity, Froude Krylov pressure, radiation and diffraction pressure, speed effects, and rudder and propeller actions. Forces and moments caused by radiation and diffraction are deduced from the two-dimensional potential ow at each of the transverse ship sections strips . The FroudeKrylov forces, resulting from the water pressure induced by the incident waves, are integrated up to the wave contour. In this way, these pressures also cause changes in the righting arm curve due to the changing position of the wave crest relative to the ship. The waves themselves, however, are treated linearly according to Airy theory, which means that nonlinear effects in steep seaways, such as higher and steeper wave crests than wave troughs, are neglected. Hydrodynamic radiation forces added mass and damping are generated by ship motions, whereas diffraction forces arise from the difference between the forces acting on the stationary ship in waves and the FroudeKrylov forces. This difference is caused by the change of the wave pressure eld due to the presence of the stationary ship. For instance, if the ship partly emerges from the water, the diffraction component of the emerged part must be zero and thus cannot be computed for the nonmoving ship. Therefore, the computation of the sum of radiation and diffraction forces is based on the relative motion hypothesis, that is, by assuming that the orbital velocity of the wave motion averaged over a ship cross section determines the uid force at that cross section. Summing the contributions from all ship sections strips yields the total force on the ship. Longitudinal interactions due to forward speed effects are treated in the same way as for linear strip methods. The hydrodynamic pressure distribution depends not only on the ships momentary position, velocity, and acceleration, but also on the past history of the ships motion, which is reected in the wave pattern. In frequency domain computations, this so-called memory effect is expressed in the frequency dependence of added masses and damping. In time domain simulations, one solution is to use impulse-response functions by formulating the hydrodynamic memory forces and moments , F, as convolution integrals that account for the dependency of these forces on the accelerations at different time steps
t

Computational Procedure
The objective was to obtain spatial mean slamming pressures that can be applied as equivalent static design loads to determine scantlings of hull structural elements for the forebody of the two reference ships. This was accomplished by integrating computed local slamming pressures over selected critical areas of the hull. For these ships with ared bows, the area under the bow are even when there is no keel emergence is generally subject to slamming. The following steps comprised the computational method: 1. The linear, frequency-domain Green-function panel code GLPANEL computed ship responses in unit amplitude regular waves. Wave frequency and wave heading were systematically varied to cover all possible combinations that are likely to cause slamming. Results were then linearly extrapolated to obtain responses in wave heights that represent severe conditions, here characterized by steep waves close to breaking. Under such conditions, the added resistance tends to result in voluntary or involuntary speed loss. Thus, the ships were assumed advancing at reduced speeds. 2. Regular design waves were selected on the basis of maximum magnitudes of relative normal velocity between ship critical areas and wave averaged over these critical areas. This velocity is dened as follows An i
n= i i

1 An
i i

where An i is the area vector normal to a surface patch i on the ships hull dened by the computational grid generated for the GLPANEL computations; i is the relative velocity at this patch i; and i An i is the examined critical area made up of i surface patches. 3. The nonlinear strip method GLSIMBEL determined motions of the ships under regular design wave conditions, thereby accounting for the nonlinear pressure distribution up to the wave contour and the frequency dependence of the radiation forces memory effect . 4. The nonlinearly computed ship motions constituted part of the input for the RANSE code. To compute slamming loads, numerical volume grids surrounding the ships were generated, appropriate boundary conditions were applied, and convergence criteria were dened.

Ft =

The Linear Panel Method


The linear, frequency domain seakeeping panel code GLPANEL uses a zero-speed Green function and a forward speed correction based on the encounter frequency approach. A velocity potential is found by distributing singularities over the mean wetted surface. Their strengths are determined by satisfying appropriate boundary conditions, leading to integral equations for the singularities. The integral equations are solved numerically by replacing them with a set of linear equations, yielding the strengths of the singularities situated on a nite number of surface 40 / Vol. 129, FEBRUARY 2007

where the matrix K is determined from potential ow computations of the individual strips at different immersion drafts and inclination angles of the waterline. The velocity u contains not only ship motions, but also the incident wave particle velocities. The waves are generated by the ship at time t . As they are still present at time t, they will continue to exert forces and moments on the ship. Although time domain simulations of ship response can be carried out with memory effects being updated at each time step by recalculating all convolution integrals, it is numerically far more efcient to use a nite state space approximation as advocated by Schmiechen 13 . The code makes use of this alternative. Instead of relying on the proportionality between force and acceleration, a relation is established between acceleration and a few of its time derivatives on the one hand and of the force and a few of its time derivatives on the other hand Transactions of the ASME

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

A 0u + A 1 u + A 2 2 u + = B 0 + B 1 + B 2 + f f t t

where f is the force per unit length; and u is the acceleration. Both f and u are three-component column vectors. Matrices A0, A1, A2 , and B0, B1, B2 , . . . are 3 3 complex matrices that do not depend on frequency, but they still depend on section shape, submergence depth, and heel, and thus implicitly on time. They are computed from the frequency dependent added mass and damping coefcients by regression analysis. Three degrees of freedom have to be considered for each strip, namely, the transverse and vertical translations and the roll motion. Thus, to enhance the numerical efciency of simulations, these matrices are determined for a sufcient number of encounter frequencies, immersion drafts, and inclination angles before starting the simulations. During a simulation, values for the actual immersion and waterline inclination are obtained by interpolation. Terms account for effects of viscous hull damping in roll, using Froude number dependent coefcients. The inuence of bilge keels is approximated semi-empirically. Propeller thrust is considered, whereby a motion equation for the propulsion system determines the propeller rate of rotation. Rudder forces and moments are determined by considering the orbital motion of water particles and the immersion of the rudder in the seaway.

Fig. 1 Part of numerical grid domains surrounding Hull 1

tion of the distribution of the volume fraction and the local Courant number. The free surface is smeared over two to three control volumes. Fluid structure interaction effects are presently not accounted for, i.e., the body is assumed to be rigid. The uid is assumed to be viscous and incompressible.

Flow Simulation
The numerical volume grids surrounding the ships comprised about 1 1 / 2 million hexahedral control volumes. To avoid ow disturbances at outer grid boundaries, these boundaries were located at a distance of one ship length ahead of the bow, two ship lengths aft of the stern, one ship length beneath the keel, and one ship length above the deck. The large domain of the mesh, especially below the keel and above the deck, was chosen to allow large pitch motions in head waves. Near the ship hull and ahead of the ship, grid density was high to resolve the wave, whereas aft of the ship the grid became course to dampen the waves. The innermost dimensionless cell thickness was chosen such that n+ = 100 on average 15 . Figure 1 shows part of the numerical grid domains surrounding Hull 1. Front, side, bottom, and top ow boundaries were specied as inlets of known velocities and known void fraction distributions dening water and air regions. On the hull surface a no-slip condition was enforced on uid velocities and on the turbulent kinetic energy. The wake ow boundary was specied as a zero-gradient pressure boundary hydrostatic pressure . All computations were performed using the RNG-k- turbulence model with wall functions 16 . The time step size was chosen such that the Courant number was smaller than unity on average. The momentum equations were discretized using 85% central differences and 15% upwind differences. Ship motions were realized by moving the entire grid at each time step. Thus, all boundary conditions were newly computed at each time step. Volume fractions and velocities that initialized the ow eld arose from superposition of ship speed and orbital particle velocities of the design waves. The inuence of numerical damping on the wave height was taken into account. Numerical diffusion caused by the course grid aft of the ships dampened the incident wave to such an extent that no signicant wave reection occurred at the outlet boundary. For runs in regular waves, simulation of the ow eld continued until a periodic solution was reached. After a simulation time of 25 encounter periods, depending on ship motions, ship speed, wave height, and position of the investigated plate elds, periodically converging solutions were obtained. For each time step up to 15 outer iterations were needed. An earlier study 8 systematically investigated errors associated with the use of different schemes to discretize the momentum equations. Slamming pressures as well as vertical forces on the hull were examined. It was found that the use of higher order approximation schemes increases the accuracy of pressure predictions; however, at the cost of encountering numerical instabilities. We chose an 85% CDS approximation as this was a reasonable compromise between accuracy and numerical stability. This earlier study also found that the inuence of the approximation scheme on vertical force predictions is almost insignicant. FEBRUARY 2007, Vol. 129 / 41

The RANSE Solver


The RANSE solver of COMET, a code that employs interfacecapturing techniques of the volume-of-uid VOF type, was an obvious choice for computing complex freesurface shapes with breaking waves, sprays, and air trapping, hydrodynamic phenomena that should be considered to predict slamming pressures. In this code, the conservation equations for mass and momentum in their integral form serve as the starting point. The solution domain is subdivided into a nite number of control volumes that may be of arbitrary shape. The integrals are numerically approximated using the midpoint rule. The mass ux through each cell face is taken from the previous iteration, following a simple Picard iteration approach. The remaining unknown variables at the center of the cell face are determined by combining a central differencing scheme CDS with an upwind differencing scheme UDS for the convective terms. The diffusive terms are discretized using CDS. The CDS employs a correction to ensure second-order accuracy for an arbitrary cell. A second-order CDS can lead to unrealistic oscillations if the Pclet number exceeds two and large gradients are involved. On the other hand, an UDS is unconditionally stable, but leads to higher numerical diffusion. To obtain a good compromise between accuracy and stability, the schemes are blended. Near the ship hull, the blending factor is chosen between 0.8 and 0.9. The three time level methods are used to integrate in time. Pressure and velocity are coupled by a variant of the SIMPLE algorithm 14 . All equations except the pressure correction equations are under-relaxed using a relaxation factor of 0.8. The pressure correction equation is under-relaxed using a relaxation factor between 0.2 and 0.4 for unsteady simulations, nding in each case a suitable compromise between stability and convergence speed. The two-uid system is modeled by a two-phase formulation of the governing equations. No explicit free surface is dened during the computations, and overturning breaking waves as well as buoyancy effects of trapped air are accounted for. The spatial distribution of each of the two uids is obtained by solving an additional transport equation for the volume fraction of one of the uids. To accurately simulate the convective transport of the two immiscible uids, the discretization must be nearly free of numerical diffusion and must not violate the boundedness criterion 14 . For this purpose, the high resolution interface capturing HRIC scheme is used 15 . This scheme is a nonlinear blend of upwind and downwind discretization, and the blending is a funcJournal of Offshore Mechanics and Arctic Engineering

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 2 Computed and measured amplitudes of a pitch deg and b vertical acceleration m / s2 of Hull 1 in regular head waves

Model Tests
The Hamburg Ship Model Basin HSVA conducted systematic seakeeping model tests of Hull 1 to experimentally validate the computations. The self-propelled model, constructed at a scale of 1:10, was tted with xed stabilizing ns, twin shafts and propellers, and a set of steerable twin rudders. To obtain model test measurements of hydrodynamic loads acting on the bow, a forebody hull segment was separated from the wooden model. Located above the stillwater line, this separated segment was connected to the ship model by a special force balance that enabled measuring six-degree-of-freedom forces and moments acting on this segment. The vertical separation of this hull segment was located 4.0 m aft of the forward perpendicular at building frame 90 ; the lower and upper horizontal separations were situated 3.7 m and 7.2 m above the base line, respectively. The segments surface area amounted to 53 m2. Global loads as well as local pressures acting on the separated hull segment were measured. Local pressures were recorded by two pressure sensors located at 42 / Vol. 129, FEBRUARY 2007

a height of 5.2 m above the base line on each side of this hull segment, with the port one positioned 2.0 m aft and the starboard on 0.5 m aft of the forward perpendicular. All the given data refer to full-scale values. Measured loads on the separated bow section were corrected for inertial effects. During seakeeping tests, a gyro unit recorded pitch motions. Three accelerometers located at the aft perpendicular, at the position of the separated bow segment, and at the forward perpendicular recorded vertical accelerations. Three wave probes located at the aft perpendicular, at the longitudinal position of the pressure transducers, and at the ships bow recorded ship motions relative to the wave elevation. The free-running model was hand operated by the helmsman accommodated on the carriage of HSVAs large towing tank. A total of eight tests were run in regular head waves: Tests 1, 2, 3, 4, 7, and 8 in waves of 2.0 m height and Test 5 and 6 in waves of 3.5 m height. The wave length to ship length ratios ranged from 0.8 to 1.4, and the propeller turning rates corresponded to calm Transactions of the ASME

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 3 Locations of separated bow section dark shading and critical plate elds light shading for Hull 1

water speeds of 14 kn for Tests 1 to 6 and to 16 kn for Tests 7 and 8. As head waves were investigated, only pitch and heave motions were considered. Computed GLPANEL results generally compared favorably with measurements, which was expected because the wave heights were relatively small during the model tests. Figure 2 shows amplitudes of pitch motion and vertical acceleration at the forward perpendicular for all eight test cases. Here the dark bars identify computed results; the light bars, measured results. Positive as well as negative values are depicted. They relate to a righthanded coordinate system, xed with respect to the mean position of the ship and translating in the positive x direction. The y axis is positive to port; the z axis, positive upward. Measured values represent steady state conditions. The location of the separated bow section and the critical plate elds 1 and 2 are shown in Fig. 3. Under the HSVA test conditions, the RANSE solver performed computations of waveinduced slamming loads and local pressures acting on the separated bow section of the ship. Simulations of the ow eld were performed over two consecutive encounter periods, while measurements lasted over several periods. Over the rst two periods, the computed vertical force on the separated bow compared favorably with measurements Fig. 4 . For the corresponding slamming pressures, averaged over the plate eld areas, the functional relationship of computed values compared favorably with experimental data, but peak values differed. This deviation was largely attributed to the relatively strong variation of the measured peaks, most likely caused by the inability of the model to attain steady state conditions during tests in regular waves. At both plate elds, the time histories of measured and computed pressures were similar. In Fig. 5 these histories are shown for plate eld 1. The force and pressure histories in Figs. 4 and 5 are presented as nondimensional values. Force FV was normalized by a maxi-

Fig. 5 Time histories of measured and computed pressures on plate eld 1 of Hull 1

mum value of F0 = 2350 kN; pressure psl , by a maximum value of p0 = 265 kPa; and time Time , by the wave encounter period of T0 = 6.0 s.

Design Wave Conditions


For both ships, only head wave conditions were investigated because these conditions were rated most critical from the standpoint of slamming loads on the forebody. Based on the systematic GLPANEL computations, performed for ve different ship speeds and 30 different wave frequencies, equivalent regular design waves were selected according to the following three limiting criteria: The vertical acceleration on the bridge may not exceed 1.0g g = acceleration of gravity ; The propeller tips may not emerge from the sea surface avoid propeller racing ; and The wave height may not exceed one-tenth of the wave length consider only nonbreaking waves .

We examined all waves that met these criteria and selected design wave conditions on the basis of the largest relative normal velocity that was experienced at the center of each critical plate eld at the time of slamming. In principle, this process yielded different design wave conditions for each plate eld. However, the critical plate elds under the ared bow of Hull 1 as well as under the ared bow of Hull 2 were located closely together, and the relative normal velocities at these plate elds were nearly the same. Therefore, we considered only one design wave condition for each hull. Table 2 summarizes the two design wave conditions.

Results for Hull 1


The nonlinear GLSIMBEL computations provided ship motions at design wave conditions. Since the response had an almost harmonic behavior in the frequency of the incident wave, we used
Table 2 Design wave conditions Hull 1 Wave height Wave length Wave length / ship length Wave length / wave height Wave encounter period Ship speed 7.0 m 112.8 m 1.61 16.1 5.8 s 12.1 kn Hull 2 11.0 m 120.6 m 0.93 10.9 6.0 s 12.0 kn

Fig. 4 Time histories of measured and computed vertical force on bow section of Hull 1

Journal of Offshore Mechanics and Arctic Engineering

FEBRUARY 2007, Vol. 129 / 43

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 8 Time histories of slamming pressures at plate elds 14 for Hull 1

Fig. 6 Locations of a critical plate eld 1 and b critical plate eld 2 for Hull 1

Fig. 7 Locations of a critical plate eld 3 and b critical plate eld 4 for Hull 1

time harmonic functions to describe all motion characteristics. The corresponding heave and pitch amplitudes were 4.6 m and 11.1 deg, respectively. Four areas under the ared bow were considered critical for the structural design of Hull 1, and the computed slamming pressures were averaged over these areas. The locations of these areas are shown in Figs. 6 and 7, where they are identied as plate elds numbered 14 having areas of 0.2 m2, 4.5 m2, 5.1 m2, and 4.5 m2, respectively. The three-dimensional ow eld surrounding the hull under the inuence of design wave conditions was computed as a transient process. Computed pressures hardly changed after the rst period. Over the third period, computed time histories of these averaged slamming pressures are shown in Fig. 8. Short duration peak pressures are seen to be highest at plate elds 1 and 2, reaching peak values of about 180 kPa. A typical pressure distribution over the ships hull, corresponding to the time of maximum pressure peak occurrence, is shown in Fig. 9. The corresponding design value according to classication society rules 1 is 105 kPa, which is signicantly less than the computed value of 180 kPa. These two values are not directly comparable, because, amongs others, a safety factor between the design rule value and the computed extreme value must be considered to account for the difference between the allowable stress and the yield stress of the material.

Fig. 9 Typical predicted pressure distribution during slamming for Hull 1

44 / Vol. 129, FEBRUARY 2007

Transactions of the ASME

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 10 Locations of critical plate elds for Hull 2

Results for Hull 2


Again, GLSIMBEL computations provided ship motions at design wave conditions, and the response also had an almost har-

monic behavior in the frequency of the incident wave. Using time harmonic functions to describe the motion characteristics yielded heave and pitch amplitudes of 2.3 m and 6.8 deg, respectively. For Hull 2, three areas under the ared bow were considered critical for structural design and, as for Hull 1, the computed slamming pressures were averaged over these areas. Their locations, shown in Fig. 10, are indentied as plate elds numbered 1 to 3 having areas of 7.7, 0.7, and 5.8 m2, respectively. The three-dimensional ow eld surrounding the hull under the inuence of design wave conditions was computed as a transient process. As seen in Fig. 11, the ships bow emerged and the deck immersed. The re-entry of the bow after emergence not only resulted in slamming under the ships bow are, but it also caused water to ow onto the deck. Figure 12 shows the corresponding pressure distribution, demonstrating that impactrelated slamming pressures occurred under the ared bow directly above the bulb. Computed pressure histories acting at plate elds 1 and 3 are shown in Fig. 13. The nondimensional pressure psl is normalized

Fig. 11 Simulation of Hull 2 under design wave conditions: a deck immergence and b bow emergence

Journal of Offshore Mechanics and Arctic Engineering

FEBRUARY 2007, Vol. 129 / 45

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 12 Pressure distribution on Hull 2

by the maximum value of p0 = 265 kPa; the nondimensional time Time , by the wave encounter period of T0 = 6.0 s. Pressures acting at plate eld 2 were almost identical to pressures acting at plate eld 1 and, therefore, are not plotted. Because of its location ahead of plate elds 2 and 3, plate eld 1 experienced the highest pressure peak of about 275 kPa. The peak pressure acting at plate eld 3 was only about 65% as high as the peak pressure acting at plate eld 1. The corresponding design value for Hull 2 according to classication society rules 1 is 163 kPa. As for Hull 1, this value is signicantly less than the computed value of 275 kPa. Here again, the same reasoning applies also for Hull 2 in that these two values are not directly comparable, because, amongst others, a safety factor between the design rule value and the computed extreme value must be considered to account for the difference between the allowable stress and the yield stress of the material.

Discussion
Accurate prediction of slamming loads continues to be difcult, mainly because of the many parameters involved. The procedure presented here attempted to combine the physics of a ship in waves with the use of advanced numerical techniques.

The generally favorable agreement between predicted slamming loads and comparable data from model test measurements was mainly due to averaging the pressures over selected areas. It was this approach that made it possible to use the resulting slamming load predictions for design purposes. The functional relationships of computed pressures corresponded to classical measured slamming pressures. Slamming occurred when the incident wave caused the plate elds to immerse. Pressures then suddenly increased to their peak values and decreased rapidly afterwards to about on fth of their peaks. Pressures then slowly decreased further until they reached atmospheric pressure. We modeled the seaway by dening equivalent regular design waves although such design waves only roughly approximate reality. The inuence of structural deformation was not accounted for although it does affect impact-related loads. Including this effect would most likely have led to somewhat smaller local slamming loads.

Conclusions
Computed results of wave-induced slamming loads demonstrated that the procedure presented here was capable of predicting slamming suitable for design of a ships structure. The generally favorable comparison of computations with model test measurements validated the procedure. Before using the RANSE solver, it was necessary to identify wave conditions likely to cause severe slamming and to obtain reliable predictions of ship motions under such conditions. Therefore, motions could not be predicted from linear methods alone. A subsequent nonlinear analysis was necessary to yield sufciently accurate motion predictions that were then used as part of the input for the RANSE solver. The RANSE code COMET is currently being modied to also solve the ship motion equations, so that in future it will not be required to supply predetermined ship motions as part of the input.

References
1 Germanischer, L., 2005, Rules for the Classication and Construction, I Ship Technology, 1 Seagoing Ships, 1 Hull Structures, Hamburg. 2 Tanizawa, K., and Bertram, V., 1998, Slamming, Handbuch der Werften Chap. XXIV, Hansa-Verlag, Hamburg, Germany, pp. 191210 in German . 3 Mansour, A. E., and Ertikin, R. C., eds., 2003, ISSC: Technical Committee I.2 LOADS, Proceedings 15th International Ship and Offshore Structures Congress, Vol. 1, San Diego, CA, August 1115, pp. 8790. 4 Ohtsubo, H., and Sumi, Y., eds., 2000, ISSC: Technical Committee I.2

Fig. 13 Time histories of slamming pressure for Hull 2 under design wave conditions

46 / Vol. 129, FEBRUARY 2007

Transactions of the ASME

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

LOADS, Proceeding 14th International Ship and Offshore Structures Congress, Vol. 1, Nagasaki, Japan, pp. 102107. Kinoshita, T., Kagemoto, H., and Fujino, M., 1999, A CFD Application to Wave-Induced Floating-Body Dynamics, Proceedings International Conference on Numerical Ship Hydrodynamics, Nantes, France. DEXTREMEL, 2001, Design for Structural Safety Under Extreme Loads, Brite Euram III Research Project, Contract No. BRPR-CT97-0513, Final Technical Report at http://research.germanlloyd.org/Projects/DEXTREMEL/ DEXTR. Sames, P. C., Kapsenberg, G. K., and Corrignan, P., 2001, Prediction of Bow Door Loads in Extreme Wave Conditions, Proceedings of the International Conference Design and Operation for Abnormal Conditions II, RINA, London, November 67. El Moctar, O., Brehm, A., and Schellin, T. E., 2004, Prediction of Slamming Loads for Ship Structural Design Using Potential Flow and RANSE Codes, Proceeding 25th Symposium on Naval Hydrodynamics, St. Johns, August 813. stergaard, C., and Schellin, T. E., 1995, Development of an Hydrodynamic Panel Method for Practical Analysis of Ships in a Seaway, Trans. Schiffbau-

technische Gesellschaft, 89, pp. 561576 in German . 10 Papanikolaou, A. D., and Schellin, T. E., 1992, A Three-Dimensional Panel Method for Motions and Loads of Ships with Forward Speed, J. Ship Techn. Research 39, pp. 147156. 11 Pereira, R., 1988, Simulation of Nonlinear Sea Loads, J. Ship Techn. Research, 35, pp. 173193. 12 ICCM, 19982000, User Manuel COMET Version 2.0, Institute of Computational Continuum Mechanics GmbH, Hamburg, Germany. 13 Schmiechen, M., 1973, On State Space Models and Their Application to Hydrodynamic Systems, Univ. of Tokyo, Dept. of Naval Arch., NAUT Report No. 5002. 14 Ferziger, J., and Peric, M., 1996, Computational Methods for Fluid Dynamics, Springer, Berlin. 15 Muzaferija, S., and Peric, M., 1998, Computation of Free-Surface Flows Using Interface-Tracking and Interface-Capturing-Methods, Nonlinear Water Wave Interaction, Computational Mechanics Publ., Southampton, pp. 59100. 16 Speziale, C. G., and Thangam, S., 1992, Analysis of an RNG Based Turbulence Model for Separated Flows, Int. J. Eng. Sci. 30, pp. 13791388.

Journal of Offshore Mechanics and Arctic Engineering

FEBRUARY 2007, Vol. 129 / 47

Downloaded 09 Nov 2011 to 117.211.86.75. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

You might also like