Sie sind auf Seite 1von 26

ABS TECHNICAL PAPERS 2003

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design
Y.S. Shin, Associate Member, American Bureau of Shipping, V.L. Belenky, Member, American Bureau of Shipping, W.M. Lin, Member, Science Applications International Corporation, K.M. Weems, Member, Science Applications International Corporation, A.H. Engle, Associate Member, Carderock Division, Naval Surface Warfare Center. ABSTRACT
This paper describes recent developments and new applications in the field of numerical simulation for nonlinear ship motions using the Large Amplitude Motion Program (LAMP), following up on our 1997 paper (Shin, et al. 1997). The objective of LAMPs development is to develop an analysis tool for highly realistic prediction of wave loads and behavior for a ship in severe seas. This approach is based on physics and does not rely on statistical information gathered from model tests or existing ships, so it is expected to be especially useful for new ship types. The kernel of LAMP is panel-based potential flow solution of the ship-wave hydrodynamic problem with many numerical options, including a recently-added Rankine singularity model with a damping beach and an option to rapidly compute the perturbation potential using pre-computed Impulse Response Function (IRF) potentials. The paper describes important implementation details for these and other computational options added since 1997. The LAMP System is structured so that numerical or empirical models of other systems or effects can be directly incorporated into the time domain ship motions and load calculations using a series of optional features. One such feature is a multi-level green-water-on-deck model including a finite-volume solution of 3-D shallow water flow over the deck. This model can be used for analyzing ship behavior with water on deck and loads caused by green-water-on-deck. Numerical results illustrate the effect of green-water-on-deck on the pitch motion and vertical bending moment of a cruiser in head seas and on the roll behavior of a fishing vessel. Another feature involves anti-roll fin and tank systems, including an integrated anti-roll tank model that solves for fluid motion in a U-tube tank concurrently with the wave-body hydrodynamics. The paper evaluates the use of passive anti-roll tanks to mitigate parametric roll resonance. Recent development has focused on applying LAMP to unconventional ships, multi-hull high-speed displacement ships, and non-ship-like configurations. Results are presented from studies for an advanced Naval hull form, a trimaran, and a semi-submersible platform. INTRODUCTION New ship designs could be characterized by large variation of lines and speed range. New container carriers have not only become larger, but their hull shapes have also experienced a significant transformation as large bow flares and stern overhangs become standard features. Bulk carriers have also become larger and wider, but relatively more shallow (Shin, 2001). Increasing demand for high speed generates more multi- and mono-hull high-speed designs both for commercial and naval applications. Such a variety of new and unusual hull forms poses a new challenge for the maritime industry as design methods based on semi-empirical processes become suspect. Even direct calculation methods with proven track records for accuracy and reliability may lose their applicability, as their assumptions are no longer met. A typical example for such a situation is a recent case of parametric rolling in head seas (France et al., 2003). The physical phenomenon of parametrically excited roll motions in following seas has been known since the 1950s (Paulling and Rosenberg, 1959), but the appearance of parametric roll in head seas conditions was unexpected. The cause was the large bow flare and stern overhang that created strong coupling between longitudinal and transverse ship motions, a phenomena that is not observed for more traditional hull shapes. This resulted in unusually large roll motions for which the cargo securing system was not designed. Another example of new problems generated by modern hull shapes is bow flare slamming in oblique seas (Finn, et al., 2001). Problems associated with the influence of green-water-on-deck on ship motions and loads are also becoming serious issues for new hull shapes where operational experience has not yet been gathered.

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 257

ABS TECHNICAL PAPERS 2003 The natural reply to this challenge is to develop calculation methods suitable for these new designs. Such new methods were thoroughly reviewed by Beck and Reed (2001), who pointed out that time domain simulations of ship motions based on potential flow formulations is becoming a more practical way to achieve engineering results for new ship designs. BRIEF DESCRIPTION OF LAMP The LAMP System is a time-domain simulation model specifically developed for computing the motions and loads of a ship operating in extreme sea conditions. LAMP System development began with a 1988 DARPA project for advanced nonlinear ship motion simulation, and has continued under the sponsorship of the U.S. Navy, U. S. Coast Guard, the American Bureau of Shipping (ABS), and Science Applications International Corporations (SAIC) IR&D program. LAMP uses a time stepping approach in which all forces and moments acting on the ship, including those due to the wave-body interaction, appendages, control systems, and green-water-on-deck, are computed at each time step and the 6-DOF equations of motions are integrated in the time-domain using a 4th-order RungeKutta algorithm. In addition to motions, LAMP also computes main hull-girder loads using a rigid or elastic beam model and includes an interface for developing Finite-Element load data sets from the 3-D pressure distribution (Shin et al., 1997; Weems et al., 1998). The central part of the LAMP System is the 3-D solution of wave-body interaction problem in the timedomain (Lin and Yue, 1990, 1993). A 3-D perturbation velocity potential is computed by solving an initial boundary value problem using a potential flow boundary element or panel method. A combined body boundary condition is imposed that incorporates the effects of forward speed, the ship motion (radiation), and the scattering of the incident wave (diffraction). The potential is computed using a hybrid singularity model that uses both transient Green functions and Rankine sources (Lin et al., 1999). Once the velocity potential is computed, Bernoullis equation is then used to compute the hull pressure distribution including the second-order velocity terms. The perturbation velocity potential can be solved over either the mean wetted surface (the body linear solution) or over the instantaneously wetted portion of the hull surface beneath the incident wave (the body nonlinear approach). In either case, it is assumed that both the radiation and diffraction waves are small compared to the incident wave and the incident wave slope is small so that the free-surface boundary conditions can be linearized with respect to the incident-wave surface. Similarly, the incident wave forcing (FroudeKrylov) and hydrostatic restoring force can also be computed either on the mean wetted surface or on the wetted hull up to the incident wave. The combinations of the body linear and body nonlinear solutions of the perturbation potential and the hydrostatic/Froude-Krylov forces provide multiple solution "levels" for the ship-wave interaction problem. These levels are: LAMP-1 (Body linear solution): Both perturbation potential and hydrostatic/Froude-Krylov forces solved over the mean wetted hull surface LAMP-2 (Approximate body nonlinear solution): The perturbation potential is solved over mean wetted hull surface while the hydrostatic/Froude-Krylov forces are solved over the instantaneous wetted hull surface LAMP-4 (Body nonlinear solution): Both the perturbation potential and the hydrostatic/FroudeKrylov forces are solved over the instantaneous wetted hull surface

For most problems, the most practical level is the approximate body-nonlinear (LAMP-2) solution, which combines the body-linear solution of the perturbation potential with body-nonlinear hydrostatic-restoring and Froude-Krylov wave forces. This latter approach captures a significant portion of nonlinear effects in most shipwave problems at a fraction of the computation effort for the general body-nonlinear formulation. However, bodynonlinear hydrodynamics and nonlinear incident wave effects can be important depending on ship geometry and operating conditions. Mixed Source Formulation In the context of time-domain potential-flow boundary-element methods, the most commonly used approaches fall in two categories: (1) methods that use transient Green functions and (2) methods that use Rankine sources (Weems, et al., 2000). For the methods in the first category (e.g. Lin et al., 1994), the transient Green function satisfies the linearized free surface boundary conditions and radiation conditions in the far field, so that the singularities need to be distributed only on the wetted portion of the body surface. For ships with non-wall-sided geometry, numerical difficulties may arise in the area where the intersection angles between the body surface and the free surface become small, as with wave-piercer ships. This is mainly due to the highly oscillatory nature of the transient Green function adjacent to the free surface. For the methods in the second category (e.g. Nakos et al., 1993), the Rankine source is used as a kernel in the boundary integral equation. Rankine sources are fairly robust for modeling either wall-sided or non-wall-sided geometries. To satisfy the free surface boundary condition, Rankine sources must be distributed not only on the body surface but also on the free surface. In order to limit the size of the computation domain, the free surface region is typically truncated at several ship lengths away from the ship, and a numerical damping zone is employed to absorb wave energy.

258 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003 In view of the pros and cons of these two approaches, a hybrid numerical approach was developed to use both transient Green functions and Rankine sources (Lin et al., 1999). This approach is implemented in the LAMP System as the mixed source formulation. In the mixed source formulation, the fluid domain is split into two regions as shown in Figure 1. The outer domain is solved with transient Green functions distributed over an arbitrarily shaped matching surface, while the inner domain is solved using Rankine sources. The advantage of this formulation is that Rankine sources behave much better than the transient Green function near the body and free surface juncture, and the matching surface can be selected to guarantee good numerical behavior of the transient Green functions. The transient Green functions satisfy both the linearized free surface boundary condition and the radiation condition, allowing the matching surface to be placed fairly close to the body. This numerical scheme has resulted in robust motion and load predictions for hull forms with non-wall-sided geometries. Another advantage of the mixed formulation is that the local free surface elevation is part of the solution, and no additional evaluation is needed as in the case of the transient Green function approach. In addition, a nonlinear free surface boundary condition can be implemented at modest computational cost. hydrodynamic problem by panelizing the bottom or using image sources. IRF-Based Formulation A drawback to time domain hydrodynamics is the computational cost. To mitigate this, an Impulse Response Function (IRF) based hydrodynamic formulation (Liapis 1986, King et al., 1988, Bingham et al. 1993) was integrated into the LAMP System to complement the mixed source formulation. In the IRF formulation, velocity potentials are pre-computed for steady forward speed, impulsive motion in up to six modes, and impulsive incident waves for each speed and heading angle. The hydrodynamic problem is reduced to a convolution of the IRF potentials with the actual ship motions and incident wave elevation, thereby allowing numerical simulations to be performed faster than real time using modest computational resources and without compromising the accuracy of the hydrodynamic calculation. The IRF potentials are convoluted and summed on a panel-by-panel basis, so that the complete potential distribution of the hull can be computed in the time domain. This allows the panel pressure to be computed directly, including the nonlinear terms in Bernoullis equation, in the same fashion as in the mixed source formulation. The only restriction is that the IRF formulation can only be used with body linear (LAMP-1) and approximate body nonlinear (LAMP-2) hydrodynamics. In the IRF-based formulation, the perturbation velocity potential on each body panel is decomposed as
r k ( x, t ) =

k =1

k ( x , t ) + 7 ( x , t ) + 8 ( x , t )

Figure 1. Mixed source formulation In the LAMP System, a 2nd-order free surface boundary condition can be applied on the local portion of the free surface; see more details in Weems et al. (2000). However, in the case of nonlinear free-surface boundary condition in the local portion of the free surface, the matching surface has to be placed further away from the body to minimize errors caused by a mismatch of the free surface condition. Rankine Singularity with Damping Beach While the mixed-source singularity model works very well for low to modest speeds (Fr0.5), it can be difficult to obtain a stable solution at higher speeds. For this reason, an alternative singularity model has been implemented that replaces the external domain and the matching surface with a numerical damping region on the outer edge of the inner regions free surface. The body and free surface boundary conditions are otherwise identical to the mixed-source singularity model. While this singularity model typically requires a considerably larger free surface grid then the mixedsource model, it has been successfully applied at Fr=0.85. The model also allows shallow water to be modeled in the

where the k, k=1..6 are the radiation potentials for the six rigid-body motions, 7 is the diffraction potential related to the incident wave potential 0, and 8 is the steady state potential related to the constant forward speed U. To solve for the six radiation potentials k, six corresponding impulse response functions k are introduced via the convolution integral
r r & k ( x , t ) = k ( x , t ) X ()d

where Xk is the ship motion in mode k and the dot signifies the derivative with respect to time. Each of the radiation IRF potentials is the solution for an impulsive velocity in that mode and is obtained by solving the initial boundary value problem: 2k = 0 in V k = n k (t ) + m k h (t ) on S b n 2k + g k =0 on S f 2 z t k = k = 0 at t = 0 on S f t

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 259

ABS TECHNICAL PAPERS 2003

where (t) and h(t) are the Dirac delta and Heavyside step functions with respect to time t, respectively, and k runs from 1 to 6. To solve for the diffraction potential 7, the diffraction IRF 7 is introduced via the convolution integral

7 =

7 0 ( ) d

where 0 is the incident wave elevation at the origin of the ship-fixed frame. The diffraction IRF 7 is the diffraction potential due to an impulsive unit incident wave and is obtained by solving the initial boundary value problem:

2 7 = 0

in

7 on S b = 0 n n 2 7 + g k 7 = 0 on S f 2 z t 7 = 7 = 0 at t = 0 on S f t where 0 is the potential due to an impulsive unit incident wave at the origin of the ship-fixed frame:
0 = Re g 2

kz ik ( x cos + y sin ) + it

d e

where is the wave heading angle relative to the ship, is the wave frequency, and e is the frequency of encounter. For the head seas (/2 3/2) and zero speed cases, there is single valued relationship between and e so there is a single impulsive wave, and hence a single diffraction IRF, for each wave heading angle. However, for the following seas cases, there are up to three wave frequencies for each frequency of encounter, so there are three impulsive incident waves, and hence three diffraction IRFs, for each wave heading angle. The final component of the decomposed wave potential, the steady forward potential, can be determined by simply solving a forced steady forward speed problem with no incident waves until a converged solution is reached. All of the IRF potentials can be computed using the mixed-source formulations numerical scheme and saved in a series of files. Each of the IRFs depends only upon the ship geometry, speed, and incident wave heading angle. In principle, the convolution of the radiation IRFs must start at the beginning of the calculation and the convolution of the diffraction IRFs must run from =- to =. In practice, however, the radiation convolutions need only include the most recent 400 or so time steps (for a typical time step of 0.15 seconds) while the diffraction convolutions need only include 400 steps for head seas and 800 steps for following seas. As a result, the IRFs

are not excessively expensive to compute a full set of IRFs for a long crested head sea case requires the same calculation effort as about 200 seconds of simulation using the mixed-source formulation. The computational cost of IRFs for short-crested and/or following sea cases will be higher as additional diffraction IRFs will be required. Once computed, however, the resulting IRF-based LAMP calculation can be run 40-100 times faster than the mixed source formulation. This allows faster-than-realtime calculations for most configurations using very modest computational resources. Furthermore, the IRFs are independent of the incident waves frequency and amplitude, so an entire series of regular or irregular wave conditions can be run with the same IRFs (Weems et al., 2000). As mentioned earlier, a restriction of the IRF-based formulation is that it can only be used with the bodylinear solution for the perturbation potential where the perturbation potential is solved over the mean wetted surface of the hull. However, this does not imply that the IRF formulation is restricted to a classical linear seakeeping solution. Classical linear seakeeping linearizes Bernoullis equation, hydrostatic restoring forces, and the equations of motion and assumes a direct linear relation between the incident wave and the response. In contrast, the IRF-based velocity potentials due to ship motion (radiation) are linear in the instantaneous ship velocity as computed by the general solution of the equations of motion. Also, the IRFs are used to compute a distribution of velocity potential, so that the second order terms in velocity in Bernoullis equation are preserved. In the approximate body-nonlinear calculation (LAMP-2), the incident wave forcing and hydrostatic restoring forces are computed using a nonlinear approach. Additional forces such as those due to viscosity, control surfaces, green-water-on-deck, etc. can also be treated using this nonlinear approach. As a result, LAMPs approximate body-nonlinear solution is capable of capturing significant large amplitude wave phenomena that are missed in classical linear seakeeping, such as parametric rolling and the dependence of motions and loads on wave slope. Using the IRF-formulation with the approximate body-nonlinear calculation can provide faster-than-real-time calculation of large amplitude responses, thereby facilitating the large number of relatively lengthy simulations required to properly characterize a ships response in large amplitude seas.

LARGE AMPLITUDE WAVE LOADS


Comparisons Between the Mixed and Rankine Source Formulations In an effort to assess possible differences in results between the Mixed and Rankine Source formulations in LAMP, several comparisons were made against available model test data. The seakeeping and loads model data used for these comparisons were compiled during an advanced hull form development study at Carderock Division, Naval Surface Warfare Center (NSWC/CD).

260 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

To support this analysis, a free running 1/20-scale segmented model was built, with a continuous longitudinal backspline beam and instrumented to provide structural loads data (see Figure 2). The test program consisted of calm water, regular wave, and irregular wave measurements over a range of ship speeds and wave headings.

20 Wave At CG, ft 10 0 -10 -20 0

Model test LAMP4 10 20 30 40 50

Time, s 60 70

10 5 Pitch, deg. 0 -5 -10 0

Model test LAMP: Rankine LAMP: mixed 10 20 30 40

Heave, ft

Figure 2. Advanced monohull structural loads model Subsequent to testing, the LAMP System was used to compare motions and loads predictions with model test results. The visualization of the LAMP simulation for an extreme wave case is shown in Figure 3.

Time, s 50 60 70

20 10 0 -10 -20 0

Model test LAMP: Rankine LAMP: mixed 10 20 30 40

Time, s 50 60 70

Bow accelerations. g's

1.0 0.5 0.0 -0.5 -1.0

Model test LAMP: Rankine LAMP: mixed

Figure 3. LAMP simulation for advanced Monohull form A sample of the body nonlinear formulation (LAMP4) results obtained using both the Rankine and the mixed formulations is shown in Figures 4 and 5. The first series of figures compare the two formulations for a regular wave run with a wave steepness of 1/30 and a ship speed of 15 knots. Although there are some difference in phasing between the model test results and the predictions, both formulations show fairly good agreement. The next set of comparisons shows the model in a Hurricane Camille seaway. As can be seen from Figure 5, predicted motions and loads using either formulation are quite good, the main differences between the predicted and measured loads being that slam-induced whipping predictions were not performed. Hence the high frequency component to the load time history is not accounted for.

0
e

10

20

30

40

50

60

70

0 VBM, ft lb 10-6 -100 -200 -300 -400 0 Model test LAMP: Rankine LAMP: mixed 10 20 30 40

Time, s 50 60 70

Figure 4. Run 727: /L=1.0, H/=1/30, 15kts. (a) Wave at CG. (b) pitch, (c) heave, (d) bow acceleration, and (e) vertical bending moment

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 261

ABS TECHNICAL PAPERS 2003 Application of LAMP-FEM Interface for Computation of Wave Loads for Large Container Carrier Shin et al. (1997) contains detailed information on the interface between LAMP and finite element method (FEM) analysis software. Another sample application is shown in Figure 6. This time, a large container ship was the subject of the case study. Since no experience existed in designing a container ship of such a size, the designer recognized the simulation results as an important source of information for making technical decisions.

a
Heave, ft (full-scale) 2 0 1 0 0

Model test LAMP: mixed LAMP: Rankine

-10 -20 0 25 50 75 100 125 150 175 Time, s (full-scale)

10 Pitch (deg) 5 0 -5

Model test LAMP: mixed LAMP: Rankine

25

50

75

100

125

150

175

Time (sec, full-scale)

d
(full-scale)

c
0 -100

Model test LAMP: mixed LAMP: Rankine

VBM, ft-lb 10

-200 -300 0 25 50 75 100 125 150 175

Figure 6. Structure responses of large container ship for different loading cases (a) hogging; (b) sagging; (c-e) extreme torsion for different location of wave crest along the ship

Time (s) (full-scale)

Figure 5. Run 127: (a) heave, (b) pitch, (c) vertical bending moment

262 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

LARGE MOTION PHENOMENA AND ASSOCIATED LOADS


Parametric Roll and its Stabilization Parametric excitation in head seas is known to be a factor in the very high amplitude roll motion that caused lost cargo (France et al., 2003). One possible way to mitigate this kind of motion is a roll stabilization system. A method for approximating a U-tube type passive anti-roll tank system has been implemented as an optional plug-in module in LAMP. The system calculates fluid motion in the U-tube tank and determines the coupled nonlinear forces acting on a floating body in full six degrees of freedom. These forces are added to LAMPs calculation of motion and loads as an external (nonpressure) force. A schematic of the anti-roll tank is shown in Figure 7.

An initial investigation into the implementation and effectiveness of the anti-roll tank module was performed for a regular synchronous roll case. In this example, the sample container ship shown in Figure 8 was simulated at 15 knots in regular beam sea waves.

Figure 8. Panel representation of a sample large modern container ship The wave period was equal to the ships natural roll period of 25 seconds, and the wave height was 12 meters. Figure 9 compares roll motions for a case with no anti-roll tanks and a case with two identical anti-roll tanks. The case with no anti-roll tanks has a steady-state roll angle of approximately 22 degrees while the case with the anti-roll tanks has a maximum roll angle of 3.5 degrees. This investigation clearly shows the ability of an appropriately tuned anti-roll tank system to reduce resonant roll motions.

Figure 7. Schematic of anti-roll tank system The LAMP anti-roll tank formulation is based on the improved passive anti-roll tank model developed by Youssef, et al. (2002). It has been successfully applied in the nonlinear motion range (including parametric roll suppression) as shown in the above reference. Several assumptions are used in the anti-roll tank formulation including: Vertical tank columns and connecting tubes are assumed to be circular. Shear stress at the wall is related to a friction factor that is dependant on Reynolds number and based on either the Colebrook formula or the Swamee and Jain (1976) formula. Energy loss in the elbows connecting the vertical columns to the cross pipe is expressed in terms of a loss coefficient. Tank free surface correction does not account for sloshing effects and assumes a flat calm free surface that remains parallel to the global horizontal plane.

Figure 9. Roll motion in regular beam seas (wave period = ships natural roll period) Several calculations were performed on a sample container ship to determine the effectiveness of the passive anti-roll tank system in reducing the ships susceptibility to parametric roll. Figure 10 shows the predicted maximum roll angles for this container ship in regular head seas as a function of encounter frequency. A range of encounter frequencies in which this particular ship might be susceptible to parametric roll can be clearly seen in the curve with no anti-roll tanks. The roll natural frequency for this ship is 0.251 rad/sec, which corresponds to a 25 second period, so parametric rolling would be expected near an encounter frequency of 0.502 rad/sec, which is clearly shown in the results. The nonlinear dependence of parametric roll on the incident wave amplitude is also clearly illustrated. This ship had a

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 263

ABS TECHNICAL PAPERS 2003

large range of encounter frequencies where parametric roll is predicted, which indicates that the ship is very susceptible to parametric roll.

subsequently updated, after which the green-water-ondeck pressure and forces are passed back to the ship motions calculation to be used in solving the equations of motions and computing sectional loads. The structure of this computational approach is illustrated in Figure 12.

Figure 10. LAMP computed parametric roll domain for sample container ship The other cases plotted in Figure 10 show the same container ship with a single passive anti-roll tank where the tank mass is equivalent to 0.27, 0.35, 0.71, and 1.4 percent of the ships displacement, with a tank system natural period equal to the ships natural roll period of 25 seconds. The 0.27 percent case shows a reduction in the bandwidth where parametric roll occurs, but there is still a region of large roll angles. Both the 0.35 percent and 0.71 percent cases show virtually no roll past the initial 5degree roll angle at the start of the simulation. The 1.4 percent case shows a region of slightly elevated roll motion (<10 degrees) for low encounter frequencies. The increased roll motion is due to the large mass in the tank system, which is not in equilibrium at the beginning of the simulation. The tank fluid continues to oscillate for a short period of time until it eventually damps out as the simulation progresses. Analogous principles and procedures have been used to couple a numerical solution of sloshing in cargo and anti-rolling tanks with time-domain ship motions (Kim, 2002).
Influence of Green-Water-on-Deck on Motions and Loads

Figure 11. Decomposition of the computation domain

Figure 12. Structure of the LAMP System with greenwater-on-deck The green-water-on-deck calculation in the LAMP System has been integrated in a modular fashion so that different levels of calculations are available. Hydrostatic and Froude-Krylov pressure: Deck pressure is computed from the submergence of each deck element and the linear incident wave pressure (/t). This is a straightforward modeling of the deck as a bodynonlinear hydrostatic/Froude-Krylov surface and represents the basic deck-in-water calculation. Semi-empirical model: Water-on-deck is calculated directly from deck edge exceedance using an empirical expression derived from Zhou et al. (1999). The resulting deck pressure, including the effects of deck motion, can then be evaluated following the approach proposed by Buchner (1995). This model is intended to provide a quick estimate of the effect of foredeck green-water-on-deck on ship motions and main girder loads. Shallow water flow calculation: Computes the longitudinal and transverse flow over the deck with the assumption of shallow water using a finite-volume technique. This is the principal green-water-on-deck

Recent efforts have developed a sophisticated greenwater-on-deck model that can be integrated directly into LAMP, allowing water-on-deck effects to be included in time domain calculations of nonlinear ship motions and loads. In the approach that was selected, the ship motions and green-water-on-deck calculations run concurrently in the time domain. As illustrated in Figure 11, this approach subdivides the computational domain into an outer problem where the ship-wave interaction problem is solved and an inner problem where the green-water-ondeck problem is solved (Belenky, et al., 2002). At each time step of the LAMP calculation, the ship motion and wave definition are used to compute the relative motion at the deck edge (often called the deck edge exceedance), the deck tilt, and the deck acceleration. These values are in turn passed to the green-water-ondeck calculation. The green-water-on-deck calculation is

264 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

calculation used in the present study and is described in detail below. Fully 3-D flow calculation: Computes the flow over the deck, including vertical gradients, using a 3-D volume grid and a finite-volume technique. This new technique is under development and is a likely future path for greenwater-on-deck flow calculation methods. The most sophisticated of LAMPs current greenwater-on-deck calculation methods is the one based on the solution of the shallow water flow equations using a novel finite volume strategy. In this method, the equations for conservation of mass and momentum are solved in the time domain, shallow-water flow is assumed, and viscous effects are ignored except for a relatively simple deck friction term. The shallow water assumptions are that the fluid acceleration normal to the deck can be ignored while the tangential fluid velocity and pressure are constant across the depth of water-on-deck (Stoker, 1957). These assumptions reduce the 3-D fluid domain to a 2-D computational domain. This method can handle a variety of boundary and initial conditions, and it is capable of supporting arbitrary motions and general ship-deck geometries, including partial height walls (e.g. bulwarks), infinite walls (e.g. deckhouses), and stepped or raised sections (e.g. hatch covers). This approach has been validated with available experimental data and has been successfully integrated with the LAMP System (Liut, et al., 2002). For this approach to be useful as an integrated part of the LAMP simulation, the green-water-on-deck calculation must be reasonably fast, robust, and capable of calculating the flow on a deck that is moving with largeamplitude six-degrees-of-freedom motions and also exchanging water with the environment. Because of these requirements, a considerable portion of the effort in developing and implementing the numerical method has been spent in ensuring a stable and reasonable solution even when the assumptions of shallow water flow are stretched to their limits. In order to test and validate the numerical solution of the shallow-water calculation, computational studies have been made for several configurations for which detailed theoretical and experimental data are available. One such study examined the dam-breaking problem, studying the level of water behind a dam after the dam is suddenly removed. A second study examined the outflow from a box whose end is suddenly removed. Both studies show good agreement between the present method and published data (Liut et al., 2002). In order to test the green-water-on-deck implementation in the LAMP System, a series of calculations were made for a US Navy CG-47 Class cruiser, a U.S. Coast Guard Cutter, and a large modern container ship. The qualitative validation continues with the fishing boat study described below. The fishing vessel analyzed in the present study is the 21-meter stern dragger Italian Gold, which sank in heavy

seas off Massachusetts in 1994. All calculations were made at a heavily loaded condition with a displacement of 194 long tons and the center of gravity 0.08 meters above the mean waterline. The resulting transverse metacentric height (GMt) is 0.22 meters. The basic geometry model used in the hydrodynamic and hydrostatic calculations is shown in Figure 13.

Figure 13. Fishing boat geometry In the simulations, the skeg and rudder were modeled using non-pressure force models, so their effect was included in the equations of motions but not in the wavebody interaction potential. Green-water-on-deck was calculated over the afterdeck, with inflow/outflow allowed over both sides and the stern of the ship. The first and, to this point, most extensive portion of the study involved a series of roll decay calculations using different computational models and options. For clarity, these calculations were made in calm water at zero speed with the ship free only to roll. The initial roll angle was 30 degrees, at which the deck edge and bulwark are reasonably deeply submerged. For the case involving the shallow water green-water-on-deck calculation, the deck elevations were initialized with stationary water-on-deck up to the calm waterplane Z = 0. An initial calculation was made for the boat with no water-on-deck calculation of any kind on the afterdeck. A time history of the roll angle, in degrees, is shown in red in Figure 14. The blue square-tooth at the bottom of the graph indicates the relative motion of the minimum freeboard point of the deck edge; a high value indicates that the deck edge is submerged while a low value indicates that it is above the wave surface. This point, which is located just aft of the deckhouse, submerges at about 10 roll angle in calm water. To some extent, this indicates the beginning of the transition from water-ondeck to deck-in-water.
30 , deg 20 10 0 20 40

Deck enters water

10 20 30

t, s

Deck enters water

Figure 14. Roll decay angle and deck edge submergence without effect of water-on-deck The initial roll period is 5.17 seconds and increases slightly to 5.47 seconds as the roll angle decreases. This change in roll period is due primarily to the change in

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 265

ABS TECHNICAL PAPERS 2003

hydrostatic restoring moment of the body-nonlinear formulation. A second calculation computed water-on-deck pressure from hydrostatics. The results for this calculation are shown in Figure 15. In addition to the roll angle and deck edge submergence shown in the bottom graph, the top graph shows a red line indicating the volume of wateron-deck in m3 and a blue line shows the resulting heeling moment in Newton-meters. Since the model contains no lag for water entering or leaving the deck, the volume of water and the heeling moment are exactly in77 phase with the roll angle. In other words, only the deck-in-water situation is modeled. The initial roll period increases to 5.96 seconds, an effect caused by the water-on-deck heeling moment, which in this case effectively decreases the restoring moment when the deck is submerged. As the roll angle decreases, the roll period decreases to 5.47 seconds, matching the period from the previous calculation. A small but noticable decrease in the roll damping can be observed over the early portion of the run, where the longer roll period has decreased the roll velocity and reduced the damping due to skin friction and appendages.
40 VGW, m3 20

as the water flows off the deck, although there are small periods of inflow during subsequent deck edge submergences. The initial roll period is similar to the hydrostatic pressure case, which is reasonable as the water-on-deck heeling moment is similar, while subsequent roll periods match both of the previous calculations at similar roll angles. The influence of wateron-deck and deck-in-water is realized mainly through a significant increase of roll damping (see Figure 16). Since the problem is otherwise nearly identical to that of the previous calculation, the damping influence must be credited to the water-on-deck. The most likely explanation is that the lag of the water flowing off the deck or to the other side of the boat when the boat rights itself results in a significant volume of water on the rising side of the deck, where the resulting deck pressure produces a significant moment in phase with but opposed to the roll velocity. Evidence for this interpretation is offered in Figure 17, which shows a snapshot of the boat, which was initially healed to starboard, with the computed water-ondeck surface as it rolls through 0-degree heel in the first cycle of the calculation.
3 60 V GW, m

MGW, 10 , Nm VGW

-4

MGW, 10-4, Nm

40

MGW
20

VGW
0 20 40

0 20 40 40 , deg 20

20

40

t, s

t, s

20 , deg 40 20

20

40

t, s

0 20 40

20

40

t, s

20

Figure 16. Roll decay results including finite-volume water-on-deck effect without bulwarks

Figure 15. Roll decay results including hydrostatic forces on deck: water-on-deck volume (VGW, top) and roll moment (MGW, top), roll angle (, bottom), and deck edge submergence (bottom) The next calculation uses the finite-volume solution of the shallow water flow over a deck with bare deck edge (i.e. no bulwark). The volume of water-on-deck starts out large (as noted above, the finite-volume model is initially full to mean Z = 0) and generally decreases in volume

Figure 17. Computed water-on-deck surface for the fishing boat as it rolls though 0 heel

266 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

The next calculation, whose results are shown in Figure 18, introduces a solid (no freeing ports) bulwark with height 0.853 meters to the finite-volume calculation. The bulwark, which traps the water-on-deck and raises the angle at which water can enter the deck in calm water to over 20 degrees, changes the whole picture of motions. Damping is such that the second peak of roll angle is too small to immerse the bulwark top or to allow significant inflow. The amount of green-water-on-deck stays constant and the ship quickly assumes a static heel angle of about 7 degrees.
MGW, 10-4, Nm 3 40 V GW, m
20

40 20

t, s
520 20 40 540 560 580 600

Figure 19. Regular wave roll without the effect of wateron-deck MGW, 10-4, Nm VGW, m3 VGW

VGW

100
40 60

0 20 40 20

20

t, s

MGW
520 540 560 580

t, s
600

, deg
100
0 20 40 60

MGW , deg

t, s

80 40

20

Figure 18. Roll decay results including finite-volume water-on-deck calculation with 0.853 m bulwarks The second portion of the fishing boat study involved calculations for the boat at zero speed in regular (single frequency) beam waves. As in the roll decay calculation, the boat was only free to roll. While not a realistic situation, it simplifies the comparison of the different calculations and reduces the risk of capsize, especially during the initial transient period. The wave height was 3.05 meters with the period of 6.28 sec, which makes the wave forcing close to the roll resonance regime. We consider only steady state rolling in the present discussion, so the results shown in Figures 19-22 are plotted beginning 520 seconds into the simulations. The initial transient behavior, which is both interesting and very important, is simply beyond the scope of the present discussion. Figure 19 shows the results of the regular wave calculation for the fishing boat with no water-on-deck model at all. As for the roll decay results, the blue square-tooth indicates the submergence of the decks minimum point, which now is a function of the roll angle and the incident wave elevation.

t, s
520 40 80 540 560 580 600

Figure 20. Regular wave results including hydrostatic forces on deck: water-on-deck (vgw, top, solid) and roll moment (mgw, top, dashed), roll angle ( bottom), and deck edge submergence (bottom) Figure 20 shows the results of the regular wave calculation with the deck pressure calculated as the hydrostatic and linear incident wave (Froude-Krylov) pressure only. The response is very regular and primarily periodic at the forcing frequency. An increase in roll amplitude, attributed mostly to the decrease in restoring moment and increase in wave forcing when the deck is submerged, leads to an increasing amount of time deckin-water.Figure 21. Regular wave results including finite-volume water-on-deck calculation but with no bulwark The next regular wave calculation used the finitevolume calculation of the shallow water flow over the deck with bare deck edges (no bulwark). The volume of water-on-deck, plotted in red on the top graph, changes by approximately 80 percent over the wave cycle. The roll response is smaller than the hydrostatic and linear incident wave pressure calculation, probably because of the wateron-deck damping seen in the roll decay study, and shows

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 267

ABS TECHNICAL PAPERS 2003

a bias of about 5 toward the up-wave direction. A weak subharmonic roll response can be observed. MGW, 10-4, Nm VGW, m3 VGW 50

t, s
520 50 540 560 580 600

100 40 20

MGW , deg

t, s
520 20 40 540 560 580 600

Figure 21. Regular wave results including finite-volume water-on-deck calculation but with no bulwark
MGW, 10-4, Nm 100 VGW, m3 VGW
50

ship dynamics: the subharmonic character of the response might be attributed to water-on-deck influence. Garkavy (1991) observed subharmonic rolling of a ship with small freeboard (and as a result, one with water-on-deck) during model tests. Observations of subharmonic roll could be a result of period-doubling bifurcation, which is known to happen for nonlinear rolling in beam seas; see, for example, Nayfeh and Sanchez (1990). Accumulation of water-ondeck decreases the instantaneous restoring moment; as a result, ship roll could be shifted close to the nonlinear region in the vicinity of the GZ curves peak, which might facilitate the bifurcation of the periodic solution. However, in order to state that the observed behavior is the manifestation of period-doubling bifurcation, it is necessary to analyze the stability of steady state motion, which is beyond of the scope of this study. At the same time, the period-doubling bifurcation could be seen as a possible physical explanation for the observed results and as an indicator of validity for the numerical model of water-on-deck. To check influence of the green-water-on-deck on longitudinal motions and global loads another set of calculations have been made for the US Navy CG-47 Class cruiser in two sea conditions: Head seas, steep regular waves, = 810 feet and H = 48 feet (ship length is 528 feet) Head seas, storm sea condition, H1/3 = 37 feet The latter case matches an experimental recreation of the waves measured during a hurricane and would be classified as Sea State 8. For the green-water-on-deck calculations, a CG-47 input geometry was prepared with an additional component representing the deck forward of the superstructure. This input geometry is presented in Figure 23.

520 50

540

560

580

600

t, s

MGW
100 80 40

, deg

t, s
520 40 80 540 560 580 600

Figure 23. LAMP geometry for CG-47 including forward deck The first of the two test cases is somewhat unrealistic but was intended to test the robustness of the calculation and exaggerate the green-water-on-deck effects for evaluation purposes. The heave and pitch motion histories are shown in Figure 24. Note that the heave is normalized by ship length, the pitch is in radians, and a positive pitch angle is bow down in this and all subsequent motion plots. The horizontal axis is nondimensional time, in which unit time would correspond to about 4.0 seconds for the fullscale ship. As seen in these plots, water-on-deck does not significantly change the motion response.

Figure 22. Regular wave results including finitevolume water-on-deck calculation with 0.853 m bulwarks Figure 22 shows the regular wave results with a 0.843 m bulwark added to the finite-volume calculation. The maximum volume of water-on-deck is slightly higher than in the no-bulwark calculation, but the mean value of water-on-deck has increased substantially. Subharmonic roll response and increased roll amplitudes are observed. Comparing Figures 21 and 22, it can be concluded that accumulation of water-on-deck has a major impact on

268 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

0.04 0.02 Heave, (-) 0 0.02 0.04 No greenwater effect 0 2 4 Hydrostatics / FK Semi-empirical Finite volume model 0.2
0.1

8 10 Time, (-)

12

14

16

The second of the two test cases duplicates a model test that recreated the sea spectrum derived from wave probe data taken during a major hurricane. This case is thought to represent physically realistic extreme sea conditions that a ship might encounter. As for the regular wave case, the various green-water-on-deck models do not have a significant apparent effect on the motions, shown in Figure 27, although the bow emerges from the water somewhat more slowly when green-water-on-deck effects are present.
6 10 4 10 VBM, (-)
5 5 5

Pitch, (-)

2 10

0 0.1 0.2

0 2 10
5 5

8 10 Time, (-)

12

14

16
___

4 10 No greenwater effect 0 2 4
1 10 5 10
Hydrostatics / FK Semi-empirical 4 Finite volume model

Figure 24. Water-on-deck effects on heave (top) and pitch (bottom) motion for CG-47 in large regular waves Figure 25 shows a detail of the pitch response for two cycles of the calculation. All four of the calculations agree at the maximum bow down position (+ pitch) but three calculations that include water-on-deck effects all predict a lag in the bows recovery and some differences in the bow up ( pitch) peak.
0.15 0.1 Pitch, (-) 0.05 0 0.05 0.1 0.15No greenwater effect 10 11 12
Hydrostatics / FK Semi-empirical Finite volume model

8 10 Time, (-)

12

14

16

VBM, (-)

0 5 10 1 10
5

8 10 Time, (-)

12

14

16

Figure 26. Water-on-deck effects on vertical bending moment at X/L=0.26 (top) and X/L=0.49 (bottom) in large regular waves (VBM is normalized by gL4) Figure 28 shows the total green-water-on-deck force and moment for the semi-empirical and 3-D shallow water models. The results for the two models are similar, although the forces from the 3-D model appear to be larger and slightly delayed. The force results also show how the 3-D model allows water on deck to linger somewhat after the free surface falls below the deck edge

13 14 Time, (-)

15

16

Figure 25. Detail of water-on-deck effects on pitch motion of CG-47 in large regular waves The green-water-on-deck effect is far more noticeable in the load calculations shown in Figure 26. In all of the calculations, the water-on-deck reduces the peak sagging (+ VBM) moment. The magnitude of this reduction is similar for the three methods but the details of the peak differ. As might be expected, the water-on-deck effect is greatest for the loads near the bow and smallest for those near the stern.

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 269

ABS TECHNICAL PAPERS 2003

0.06 0.04 Heave, (-) 0.02 0 0.02 0.04 0 5 No greenwater effect Hydrostatics / FK Semi-empirical Finite volume model 10 Time, (-) 15 20

regular wave case, the green-water-on-deck models all predict a reduction in the peak sagging bending moment.
No greenwater effect Hydrostatics / FK Semi-empirical Finite volume model

1 10 4 5 10 VBM, (-)
5

0 5 10 1 10
5

0.2 0.1 Pitch, (-) 0 0.1 0.2

10 Time, (-)

15

20

Figure. 29 Green-water-on-deck effects on vertical bending moment at X/L=0.49 for CG-47 in head storm seas
0 5 10 Time, (-) 15 20

Figure 27. Water-on-deck effects on heave (top) and pitch (bottom) motions for CG-47 in irregular head seas, H1/3 = 37 feet.

For the storm sea run, pressure output files were written that include the deck geometry and computed pressure for any green-water-on-deck calculations. LAMPs pressure data post-processor can use the computed deck pressure, along with the computed hull pressure distribution, to calculate nodal pressure and FE load data sets. Figure 30 shows a plot of the deck and hull pressure distributions at T=10.04 of the storm sea run.

1 10

Pitch moment

0 1 10 2 10 3 10
4

Time

Surge force

4
Vertical force

10

15
Pitch moment Time

20

1 10

Figure 30. Plot of hull and deck pressure (normalized by gL) for CG-47 in storm seas at T=10.05
Multi-Hull Applications

0 1 10 2 10 3 10
4 4

Surge force Vertical force

10

15

20

Figure 28. Water-on-deck force (normalized by gL3) and moment (normalized by gL4) for CG-47 in irregular head seas by semi-empirical (upper) and 3-D shallow water (lower) models for water-on-deck. Figure 29 shows the predicted vertical bending moment near midships for the storm sea case. As with the

With its 3-D formulation and general geometry model, the LAMP System is well suited for analyzing multi-hull ships such as catamarans and trimarans. Figure 31 shows the geometry of a notional trimaran design for which some LAMP validation has been performed.

270 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

used both the linear and 2nd-order free surface boundary conditions. Figure 34 shows the predicted local wave field at one time instant of a calculation for a regular wave propagating down the long axis of the platform.

Figure 31. Geometry of sample trimaran This trimaran was designed by a team of graduate students and tested at the U.S. Naval Academy (Harris, 1999). The tests analyzed several longitudinal and transverse locations of the outer hulls and included seakeeping tests in regular and irregular head seas. Figure 31 compares the predicted and measured heave and pitch motions for the regular wave runs at a Froude number of 0.15. Like the experiments, the LAMP predictions were computed as a series of regular wave runs that used the approximate body-nonlinear (LAMP-2) model and matched the wave slope of the experimental runs.

Figure 34. Predicted local wave elevations for semi-submersible platform


LARGE AMPLITUDE MOTIONS IN FREQUENCY DOMAIN One of the main advantages of the time-domain simulation approach is its more precise description of the forces on a ship, compared with frequency domain calculations. Once the amplitude of response grows large, its nonlinear nature becomes evident. Figure 35 illustrates this through motion and load predictions for a 195 m container ship. The heave, pitch and VBM response is shown in dimensionless form in Figure 35, so it is very clear how the responses lose their sinusoidal shape with increasing wave amplitude. It is noticeable that the VBM reacts more to increasing wave amplitude than do heave and pitch. However, the heave and pitch responses are also quite nonlinear for the 8.12 m wave amplitude. Increased fidelity of ship motion and load predictions comes with a price. Results are nonlinear and they no longer can be presented in a form of an RAO, because the amplitude of response (and shape, too) now depends on the amplitude of wave excitation. Moreover, the amplitude of a nonlinear response may depend on initial conditions, so several responses may be possible at the same frequency. At the same time, amplitude vs. frequency is a conventional way to present a designs ship motion. This makes it important to find the appropriate method of frequency domain presentation for nonlinear simulation results. The natural way to do it is to apply methods of nonlinear dynamics. However, these methods were developed for objects, behavior of which can be described by a system of ordinary differential equations. This is not the case for a ship in waves. Radiated and diffracted waves interfere with incident waves, which causes excitation to depend on motion history beyond the previous time step. As a result, initial conditions (including the excitation phase) are no longer enough to determine the status of a dynamical system. There are additional issues that make the application of nonlinear dynamics non-trivial.

Figure 32. RAO for heave (m/m) and pitch (deg/m) motion in head seas for trimaran at Fr=0.15 Figure 33 offers a similar comparison at a Froude number of 0.30. The comparison for both cases is good. LAMP calculations have been made for this configuration for speeds up to a Froude number of 0.74, but no experimental data is available at that higher speed.

Figure 33. RAO for heave (m/m) and pitch (deg/m) motion in head seas for trimaran at Fr=0.15 The LAMP System has also been used to analyze large offshore structures, including a detailed evaluation of the disturbed wave field beneath a large semisubmersible platform (Weems et al., 1999). The study, which was part of an effort to examine the risk of wave impact on the underside of the main platform structure,

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 271

ABS TECHNICAL PAPERS 2003


Heave / Wave Amplitude
0.5 0.25 0 30 -0.25 -0.5 time (sec) 35 40 45 50 55 60

Pitch / Wave Amplitude


2 1.5 1 0.5 0 -0.5 30 -1 -1.5 -2 time (sec) 35 40 45 50 55

A=1m A=3m A=5m A=8.12m

far from resonance, so that roll response was almost linear. This calculation was used to determine initial conditions corresponding to steady state roll motions. However, these initial conditions could only be considered as an approximation, since they correspond to certain state of wave field that might not be reproduced in the next run. Nevertheless, the transition process initiated with approximately steady state conditions was much faster than calculations initiated from rest, and the resulting response could be considered as almost steady state. This can be seen in Figure 37, where process (a) started from approximate steady state initial conditions and process (b) from zero initial conditions.
5

60

Midship VBM / wave Amplitude


6.00E+08 3.00E+08 0.00E+00 30 -3.00E+08 -6.00E+08 time (sec) 35 40 45 50 55 60 A=1m A=3m A=5m A=8.12m

20

40

60

5 10 5 0 5 10

Figure 35. Dimensionless responses of 195 m container ship One of the standard ways to present nonlinear responses in a frequency domain is the response curve. The response curve shows dependence of amplitude on frequency. Its main difference from an RAO is that each response curve corresponds to specific excitation amplitude. Sample calculations to illustrate nonlinear dynamics were performed for the rolling fishing vessel used in the green-water-on-deck simulation (shown in Figure 13). The only difference was that for these calculations, the forecastle was removed to smooth the GZ curve and to simplify the corresponding nonlinear term. The righting arm (GZ) is plotted in Figure 36.

20

40

60

Figure 37. Transition in roll response started with approximate steady state initial conditions (a) and zero initial condition (b). Wave height 0.4 m, circular frequency 0.9 1/s Once approximate steady state initial conditions were found, then the response curve could be calculated from two series of LAMP runs. Each series contains runs with small changes of wave frequency. Approximate steady state initial conditions were determined using the previous run. As a result, the frequency step had to be small enough for the initial conditions to be close. The first series of predictions was with the wave frequency increasing, the second series with the wave frequency decreasing. The results are presented in Figure 38. As would be expected, there is an area where two different responses exist for the same wave frequency (points A and B). LAMP calculations were supplemented with direct numerical integration of the approximate differential equation for rolling using the actual GZ curve and FroudeKrylov wave excitation. && + 2 + 2 f () = 2 sin t
W

GZ, m
0.

0.

Heel,

Figure 36. Righting arm (GZ) curve in meters of the fishing vessel for response curve calculation Response curves were calculated in the following manner. The calculation started from a wave frequency that was

with natural roll frequency, W amplitude of wave slope, and

272 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003


W GZ () A44 + I x

f ( ) =

Here W is weight displacement and Ix is the transverse moment of inertia of ship mass. Damping was assumed to be linear and the damping coefficient was determined by means of a special roll decay run using LAMP. Figure 38 also shows the theoretical response curve developed using equivalent linearization. In the equivalent linearization, the nonlinear differential equation for roll was replaced with a linear one, providing equality of roll natural periods for large amplitude roll using both the nonlinear and linearized equations. It is known from nonlinear dynamics (see, for example Thompson and Stewart, 1986), that the structure in Figure 38 with two steady state responses might constitute fold bifurcation, existence of which has been confirmed experimentally (Francescutto et al., 1994). The conventional way to confirm a bifurcation is by a motion stability analysis. The behavior of the eigenvalues of a Jacobean identifies the type of bifurcation. A technique similar to those employed by Virgin (1987) and Belenky (2000) was used here. Three short LAMP runs were used for the motion stability analysis. The first run was carried out using approximate steady state conditions to produce values of the state variables (roll angle and roll velocity) after exactly one wave period. For the second run, the initial roll value was perturbed (0.51 percent perturbation typically gives reasonable results) and the run was repeated. The third run used approximate steady state initial roll angle and perturbed the roll velocity. The eigenvalues of the Jacobean matrix and further analysis indicated no difference from conventional analysis (Virgin, 1987). LAMP
0.
Roll amplitude

Figure 39(a) shows the behavior of the eigenvalues in the circular frequency range 0.91.18 rad/sec calculated as described above, part (b) of this figure shows the eigenvalues calculated using the equivalent linearized differential equation for roll over the same frequency range, and part (c) shows the theoretical behavior before and during fold bifurcation. The similarity is quite evident; moreover, the numerical values of the eigenvalues are close, taking into account the approximate method used to determine the steady state conditions with LAMP. However, loss of accuracy was observed for the same frequency range for large amplitude and for the circular frequency range 1.181.55 rad/sec. The above study was the first attempt to apply nonlinear dynamics methods to the results obtained with LAMP simulations. Despite the difficulties in defining steady state initial conditions, the outcome looks promising. Accuracy problems with steady state initial conditions still must be addressed but, in general, nonlinear dynamics gives a tool to interpret results of time domain simulation and helps to present them in the frequency domain. The next problem will be to determine how to use the response curves for further design analysis, rather than an RAO. It will also be important to apply the prescribed methodology to other ship responses in order to ascertain the influence that nonlinearities have on motions and loads.

1 1

Positive peak Positive peak


1 Imaginary 1 1

Integration of roll equation

c 1
-1 Real 1

0.

B
Backbone

0.

Equivalent linearizatio n:

2
-1

0.

Equivalent linearization:
0. 1 1. 1. 1. Circular frequency of wave,

Figure 39. Eigenvalues of Jacobean matrix for (a) LAMP calculation, (b) nonlinear roll equation (c) theoretical prediction; circular frequency range 09-1.18 rad/sec: low response, wave amplitude 0.4 m
CONCLUDING REMARKS This paper is a sequel to Shin, et al. (1997). As such, we have described the latest developments in nonlinear simulation technology and its applications for new ship designs that took place since 1997. Therefore, this paper is a review rather than a focused study.

Figure 38. Response curve of roll based on LAMP calculation, numerical integration of nonlinear roll equation and equivalent linearisation.

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 273

ABS TECHNICAL PAPERS 2003

We do not consider state-of-the-art nonlinear simulation technology as a ready replacement for linear theory (in both 2-D and 3-D formulations); it is a complement to rather than a replacement for linear theory. However, the progress achieved during the last five years allows us to state that the next steps toward a more sophisticated seakeeping analysis tool, indeed, have been made. In particular: New numerical procedures based on impulse response functions (IRF) allow calculation efficiency comparable with linear theory. Limitations of this procedure do not prevent adequate modeling of new hull shapes, since the hydrostatic and Froude-Krylov forces are calculated on the instantaneous wetted surface, radiation forces are computed with the instantaneous ship velocity, and there is no need to linearize the pressure evaluation or the equations of motion. The approximate body-nonlinear approach (LAMP-2), using the direct or IRF-based approach, is capable of reproducing known nonlinear phenomena like fold bifurcation. Since it has already been shown that LAMP-2 could predict the correct amplitude of parametric rolling (France, et al., 2003), we can conclude that it provides an adequate description of the nonlinear dynamics of a ship. Comparisons between model tests and simulation results have shown that reasonable agreement between both measured and predicted motions and loads in extreme sea conditions is possible. However, a rigorous validation effort is still required. The time-domain solution of the equations of motion and the modular structure of the code makes it possible to incorporate numerical or empirical models of greenwater-on-deck, roll control, maneuvering systems, tanks, and other effects directly into the ship motion simulation. This provides an opportunity to improve both the completeness and fidelity of the ship motion simulation, and provids a simulation platform for designing and evaluating such systems. As an illustration of this, it was shown how LAMPs anti-roll tank module could be used to evaluate different engineering solutions for suppressing parametric roll. The green-water-on-deck module allows not only observations of changes in ship dynamics due to the influence of water-on-deck, but also the evaluation of additional loads caused by green-water-on-deck. Despite the progress reported above, many issues remain for making time domain simulation more practical. Issues to be addressed include: Current versions of the simulation codes need reliable information on roll damping for accurate prediction of extreme roll motion. Reliable methods for the a priori prediction of viscous and vortex components of roll damping will be needed in the future. Ship motion simulations in large, oblique seas can be sensitive to rudder and hull side forces, so sophisticated

models of the autopilot and of maneuvering forces will be required to accurately model them. Like roll damping, reliable methods for developing control and maneuvering force models will be needed. Consideration of nonlinear effects greatly expands the amount and complexity of seakeeping data required to characterize a ship, both in terms of analyzing simulations and developing long-term extreme statistics. The methods of nonlinear dynamics could be a great help in interpreting simulation results. Finally, consideration of nonlinear incident waves seems to be a logical continuation of the present effort: extreme motions and loads have to be modeled in large waves that may not be well predicted by linear wave theory. The construction of nonlinear regular and irregular wave fields, and the modeling of these fields in LAMP, is part of an ongoing development effort.

ACKNOWLEDGEMENT
The authors wish to express their appreciation and gratitude to the management of the American Bureau of Shipping, Science Application International Corporation, and Carderock Division, Naval Surface Warfare Center. The authors also wish to thank W. F. Belknap (NSWC/CD), Y. Kim (currently MIT), D. Liut (SAIC), T. Treakle (SAIC), G. Wang (ABS), and S. Zhang (SAIC) for their help in developing the LAMP System and numerical results presented in this paper. The development of the LAMP System has been supported by the U.S. Navy, the Defense Advanced Research Projects Agency (DARPA), the U.S. Coast Guard, ABS, and SAIC. The green-water-on-deck development has been supported by the Office of Naval Research (ONR) under Dr. Patrick Purtell, by the Ship Structure Committee under Mr. Peter Minnick, and by ABS.

REFERENCES
Beck, R.F. and Reed, A.M. (2001). Modern Computational Methods for Ships in Seaway, Trans. SNAME, 109:148. Belenky, V. L. (2000). Piecewise linear approach to nonlinear ship dynamics, in Contemporary Ideas of Ship Stability, D. Vassalos, M. Hamamoto, A. Papanikolaou and D. Moulyneux (Editors), Elsevier Science, pp. 149 60. Belenky, V.L., Liut, D., Weems, K. and Shin Y.S., (2002). Nonlinear Ship Roll Simulation With Water-OnDeck, Proc. 6th International Ship Stability Workshop, Webb Institute, New York. Bingham, B., Korsmeyer, F., Newman, J., and Osborne, G. (1993), The Simulation of Ship Motions, Proc. 6th International Conference on Numerical Ship Hydrodynamics, Iowa City, Iowa, U.S.A., pp. 2747. Buchner, B. (1995). On the Impact of Green Water Loading on Ship and Offshore Unit Design, Proc.6th

274 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

Symposium on Practical Design of Ships and Mobile Units, pp.1.4301.443. Finn, P. Beck, R.F., Troesch, A.W. and Shin, Y.S. (2001). Nonlinear Impact Loading in an Oblique Seaway, Proc. of 20th International Conference on Offshore Mechanics and Arctic Engineering, Rio de Janeiro, 11 p. France, W.G., Levandou, M. Treakle, T.W., Paulling, J.R., Michel, R.K. and Moore, C. (2003). An investigation of head seas parametric rolling and its influence on Container Lashing Systems, Marine Technology, 40(1):119. Francescutto, A., Contento, G. and Penna, R. (1994). Experimental Evidence of Strong Nonlinear Effects in the Rolling Motion of the Destroyer in Beam Seas, Proc. of STAB94: 5th International Conference on Stability of Ships and Ocean Vehicles, vol. 1, Melbourne, Florida. Garkavy, V.V. (1991). Deterministic chaos in the task of the roll motion ship with a small freeboard, Proc. of HADMAR91: International Symposium On HydroAnd Aerodynamics In Marine Engineering, Paper 49, Vol. 2, Bulgarian Ship Hydrodynamics Centre, Varna. Harris, N. (1999). Effective horsepower and seakeeping tests on a trimaran model, Report EW-01-99, United States Naval Academy. Kim, Y. (2002). A Numerical Study on Sloshing Flows Coupled with the Ship Motions, J. Ship Research, 46(1):111. King, B., Beck, R., and Magee, A. (1988). Seakeeping Calculations with Forward Speed Using Time Domain Analysis, Proc. 17th Symposium on Naval Hydrodynamics, The Hague, The Netherlands. Liapis, S. (1986). Time-Domain Analysis of Ship Motions, Report 302, Department of Naval Architecture and Marine Engineering, University of Michigan, U.S.A. Lin, W.M., and Yue, D.K.P. (1990). Numerical Solutions for Large-Amplitude Ship Motions in the TimeDomain, Proc. 18thSymposium on Naval Hydrodynamics, The University of Michigan, U.S.A., pp. 4166. Lin, W.M., and Yue, D.K.P. (1993). Time-Domain Analysis for Floating Bodies in Mild-Slope Waves of Large Amplitude, Proc. 8th International Workshop on Water Waves and Floating Bodies, Newfoundland, Canada. Lin, W.M., Meinhold, M., Salvesen, N., and Yue, D.K.P. (1994). Large-Amplitude Ship Motions and Wave Loads for Ship Design, Proc. 20th Symposium on Naval Hydrodynamics, The University of California, U.S.A Lin, W.M., Zhang, S., Weems, K., and Yue, D.K.P. (1999). A Mixed Source Formulation for Nonlinear Ship-Motion and Wave-Load Simulations, Proc. 7th

International Conference on Numerical Hydrodynamics, Nantes, France, pp. 1.3.112.

Ship

Liut, D., Weems, K., and Lin, W.M. (2002). Nonlinear Green Water Effects On Ship Motions and Structural Loads, Proc. of the 24th Symposium on Naval Hydrodynamics, Fukuoka, Japan. Nakos, D., Kring, D., Sclavounos, P. (1993), Rankine Panel Methods for Transient Free-Surface Flows, Proc. 16th Symposium on Naval Hydrodynamics, Iowa, U.S.A Nayfeh, A.H., and Sanchez, N.E. (1990). Stability and complicated rolling response of ships in regular beam seas, International Shipbuilding Progress, 37(412):331 52 Paulling J.R. and Rosenberg R.M. (1959). On Unstable Ship Motions Resulting from Nonlinear Coupling, J. Ship Research, 3(1):3646. Shin, Y.S., Chung, J.S., Lin, W.M., Zhang, S. and Engle, A. (1997). Dynamic Loadings for Structural Analysis of Fine Form Ships Based on Non-linear Large Amplitude Motions and Loads Method, Trans. SNAME, 105:12754. Shin, Y.S. (2001). Official Discussion to Report of Special Tack Committee on Extreme Hull Girder, Proc. of 14th International Ship and Offshore Structure Congress, Elsevier, 3:297303. Stoker, J.J. (1957). Water Waves The Mathematical Theory with Applications, Interscience Publishers, Inc., New York, U.S.A., xxviii+567 p. Swamee, P.K. and Jain, A.K. (1976). Explicit equations for pipe-flow problems, Journal of Hydraulic Division Proceedings, ASCE, pp. 65764. Thompson, J.M.T. and Stewart, H.B. (1986). Nonlinear dynamics and chaos, John Wiley, Chichester. Virgin, L.N. (1987). The nonlinear rolling response of a vessel including chaotic motions leading to capsize in regular seas, Applied Ocean Research, 9(2):8995. Weems, K., Zhang, S., Lin, W.M., Shin, Y.S, and Bennett, J. (1998). Structural Dynamic Loadings Due to Impact and Whipping, Proc. 7th International Symposium on Practical Design of Ships and Mobile Units, The Hague, The Netherlands, 7 p. Weems, K., Lin, W.M, Zhang, S., and Treakle, T. (1999). "Application of the Large Amplitude Motion Program (LAMP) for Design and Operation of A Mobile Offshore Base (MOB)," Proc. of 3rd International Workshop on Very Large Floating Structures (VLFS 99), Honolulu, Hawaii. Weems, K., Lin, W.M Zhang, S, and Treakle, T. (2000). Time Domain Prediction for Motions and Loads of Ships and Marine Structures in Large Seas Using a Mixed-Singularity Formulation, Proc. OC2000, Osaka, Japan.

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 275

ABS TECHNICAL PAPERS 2003

Youssef, K.S., Mook, D.T., Nayfeh, A.H., and Ragab, S. A. (2002). Roll Stabilization by Passive Anti-Roll Tanks Using an Improved Model of the Tank-Liquid Motion, submitted for publication, J. Vibration and Control. Zhou, Z.Q., De Kat, J.O., and Buchner, B. (1999), A Nonlinear 3-D Approach to Simulate Green Water Dynamics on Deck, Proc. 7th International Conference on Numerical Ship Hydrodynamics, Nantes, France.

276 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

DISCUSSION Kevin McTaggart, Member


I had the pleasure of seeing the related 1997 paper on LAMP presented in Ottawa. Although I must pass on my regrets for the current meeting, I feel privileged to be a discusser. The authors note that this paper is a sequel to the 1997 paper. I saw a television program in which a learned film critic explained the difference between a sequel and a serial. A sequel tends to be more of the same, with no significant character development from the original. In contrast, a serial contains significant character development in progressive installments. The Harry Potter books are a good example of a serial. Im of the opinion that the current paper is an installment of a serial rather than a sequel. The body nonlinear solution of LAMP-4 is certainly among the most complex approaches available for prediction of seakeeping and sea loads. The approximate body nonlinear solution of LAMP-2 is more widely used. While potentially more accurate, the body nonlinear solution of LAMP-4 has high costs in terms of code development, code maintenance, and computational time. Based on comparisons with model tests and sea trials, are there cases for which LAMP-4 gives significantly better accuracy than LAMP-2. Furthermore, could you give representative ratios of CPU time to real time using LAMP-2 and LAMP-4. At DRDC Atlantic we continue to devote significant effort to comparison of ship motion and sea load predictions with experimental data from model tests and sea trials. We have observed significant differences among roll predictions from various codes due to varying treatments of viscous and lift forces. Could you comment on the treatment of these forces within LAMP? We all look forward to the next installment of the serial. In your conclusions you note the requirement for rigorous validation, which I think would be of particular interest to readers.

domain simulation computer program which includes memory functions. Previously, Professor Francescutto in Italy was able to verify this nonlinear type of behavior experimentally as referenced in the paper. However, I would like to ask the authors if they found that the solution to the approximate differential equation always gave such good comparisons and if so why? An alternative approach to study this problem is described in a paper which we at UNO prepared in 1995 (Falzarano, Taz Ul Mulk and Esparza, 1995). In this paper we look at how magnification curves change with increasing amplitude and how initial conditions and basins of attraction determined which of the multiple steady state solution solutions were eventually achieved. Over the last several years much progress has been made in applications of nonlinear dynamical systems theory to analyze large amplitude ship rolling motions. Unfortunately, to date much of this analysis has been done using simplified and approximate analysis using single degree of freedom (un-coupled) and periodic excitation. The future lies with more sophisticated dynamical systems analysis considering multiple degrees of freedom (Falzarano and Zhang, 1993) and random excitation (Vishnubhotla, Falzarano and Vakakis, 1998), and more accurate hydrodynamic models (Falzarano, Cheng and Rodrigues, 2002). Finally I would like to thank the authors for the opportunity to comment on their paper.
Additional References

Falzarano, J. M., Esparza, I. and Taz Ul Mulk, M. (1995). A combined steady-state and transient approach to study large amplitude ship rolling motion and capsizing, Journal of Ship Research, 39 (3). Falzarano, J and Zhang, F, (1993) Multiple Degree of Freedom Global Analysis of Transient Ship Rolling Motion, ASME Winter Annual Meeting, Symposium on Nonlinear Dynamics of Marine Vehicles, November 1993. Vishnubhotla, S., Falzarano, J., and Vakakis, A. (1998). Nonlinear Ship Rolling Motion in a Random Seaway, Conference on Computational Stochastic Mechanics, Santorini, Greece. Falzarano J., Cheng, J., and Rodrigues, W. (2002). Transit Draft Heave and Pitch Motion Analysis of The Mobile Offshore Base (MOB) Using Reverse MI/SO Techniques, Proceedings of the 21st Offshore Mechanics and Arctic Engineering Conference, June 2002, Oslo, Norway.

Jeffrey M. Falzarano, Member


The authors are to be congratulated on a comprehensive and complete paper. I am mainly interested in the section large amplitude motions in the frequency domain. The frequency domain is a wellknown and useful tool to analyze small amplitude linear ship motions motion but its use for large amplitude nonlinear motions is generally not as widely understood The authors Figure 38 is particularly interesting in showing the close comparison between the simple equivalent linearization results and the time domain simulations. From this figure one can see the existence of multiple solutions over a narrow frequency range which is common in nonlinear systems. It is satisfying to see this nonlinear behavior verified using a complicated time

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 277

ABS TECHNICAL PAPERS 2003

Bruce L. Hutchison, Life Fellow


The authors are to be congratulated both on this paper and on its underlying accomplishments. In this, the 50year anniversary of the legendary St. Denis-Pierson paper, the authors present us with very practical advances beyond the frontier of the linear seakeeping world. It would seem, at last, that practical nonlinear seakeeping studies may be carried out using time domain simulation as well as by physical model testing. This rich paper provides important extensions to those previously reported to the Society in 1997. There are too many advances to address within the limitations of a formal discussion, so I will limit my comments to the Impulse Response Functions, the modeling of anti-roll tanks and nonlinear rolling. The origins of this papers Impulse Response Functions (IRFs) may be traced to papers by Tick (1959) and Cummins (1962), but the implementation described by the authors has original elements that extend the usefulness of the method. Earlier papers applied the convolution integral to the simulation of motions and thus retained a restriction to the linear domain similar to that of conventional strip theory. In this paper the convolution integrals are applied to the components of the perturbation velocity potential. The perturbation velocity potentials for the radiation processes are linear in the (convoluted) ship velocity. Furthermore, the methods described permit retention of the quadratic terms in Bernoullis equation. The instantaneous body wetted surface may be estimated consistent with the assumptions that the authors identify as approximate body nonlinear, and the Froude-Krylov forces can be computed on that instantaneous wetted surface. Because the IRFs are available for each body panel, there is an immediate connection to loads suitable for dynamic loads analysis. A valuable observation is made in the references cited by the authors for the IRF method. That observation is that the diffraction potential is derived as the solution to a canonical diffraction problem forced by the incident wave potential, given by the authors in the present paper, where the incident wave potential contains all frequencies with equal weight and describes a Dirac function at time zero. The reference for the U-tube anti-roll tank model implemented by the authors has only recently been published (Youssef et al., 2003). As noted in the present paper the U-tube anti-roll tank was implemented as a complete 6-DOF system suitable for large angle rotational motions. It is the experience of this commenter that the 3-DOF local transverse plane motions (sway, heave and roll) are particularly important to the modeling of U-tube anti-roll tank performance. Within the context of ship coordinates, local sway and heave depend also on pitch and yaw, resulting in five degrees of freedom being important. Youssef et al. provide a comparison between a 1-DOF (roll only) model and a full 6-DOF tank model, where the resulting roll motions differ by a factor of two. Using the 6-DOF model they also demonstrate that longitudinal location of U-tube anti-roll tanks within the

ship can influence performance, a result that cannot be revealed by a 1-DOF model. As regards the application of LAMP time domain simulations to nonlinear rolling, and associated frequency domain and phase plane representations of those nonlinear roll responses, the authors present a righting arm curve as their Figure 36. A comparison is shown in the authors Figure 38 between LAMP simulations and direct integrations of an approximate one degree-of-freedom nonlinear differential equation for roll. As a point of clarification, it is this discussers understanding that the righting arm shown in Figure 38 was used in the approximate one DOF roll equation but that all of the restoring forces in LAMP are implicit consequences of instantaneous pressure integrations over the body wetted surface inclusive of hydrostatic pressure terms. Furthermore, it is understood that the LAMP simulations were for 6-DOF motions with the possibility of water over the deck edge (but presumably without any bulwarks). If this understanding is correct, then the agreement shown in Figure 38 between the LAMP simulations and the direct integration of the approximate one degree-of-freedom nonlinear differential equation for roll is quite remarkable and the small differences shown between the two solutions are entirely understandable.
Additional References

Cummins, W.E., (1962). The Impulse Response Function and Ship Motions, Symposium on Ship Theory, Institute for Schiffbau, Hamburg, January 1962, Schiffstechnik, 9, (47). Tick, L.J., (1959). Differential Equations with Frequency-Dependent Coefficients, Journal of Ship Research, 3(2), pp. 45-46. Youssef, K.S., Mook, D.T., Nayfeh, A.H., and Ragab, S.A. (2003). Roll Stabilization by Passive Anti-Roll Tanks Using and Improved Model of the Tank-Liquid Motion, Journal of Vibration and Control, 9(7).

Miroslaw Gerigk, Visitor


This paper addresses the latest developments in nonlinear simulation technology and its applications for the new ship designs. It has been proved in the paper that it is possible to develop more sophisticated seakeeping analysis tools. The general comments may concern as follows: 1. The new numerical nonlinear procedures for the seakeeping calculations based on the impulse response function (IRF) enable the quantitative, efficient and risk-based analysis of hydrodynamic characteristics of a ship with the higher accuracy than in the case of linear procedures. 2. The LAMP-2 system using the approximate bodynonlinear approach is able to analyse the nonlinear phenomena including the large motion phenomena and associated loads for all the

278 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

modern designs including the multi-hulls. The time domain simulation enables one to analyse the green-water-on-deck, roll control, tanks and other effects. It has been shown in the paper that the LAMP-2 system enables one to identify the fold bifurcation which is a distinguished achievement. 3. The LAMP-2 system guarantees a very high accuracy of results using both well-known and novel methods and techniques. The accuracy has been confirmed by validation using the experimental data.

A great achievement showed in the paper that applying the nonlinear dynamics methods to the results obtained with the LAMP-2 simulations it was possible to prove that the results of time domain simulation may be presented in the frequency domain. A few more comments regarding the problems presented in the paper are as follows. Parametric roll and its stabilization - A big advantage of the LAMP system is to have a possibility to consider the parametric roll and its stabilization using a Utube type passive anti-roll tank system. It has been shown that the passive anti-rolling tanks may substantially reduce the resonant roll motions. In the paper it is presented how the parametric rolling may be limited using the single passive anti-roll tank with the different values of the tank mass. It could be very interesting to see how the relative longitudinal position of two passive tanks may affect the coupled nonlinear forces acting on a floating body in full six degrees of freedom and how it may finally affect the roll motions and parametric rolling. Influence of green-water-on-deck on motions and loads It certainly is a great achievement to develop the proposed green-water-on-deck model which allows consideration of the water-on-deck effects during the time domain calculations of the nonlinear ship motions and loads. Having the different levels of calculations of the green-water-on-deck, the LAMP modular system enables one to consider the water-on-deck effects for different designs including the partial height walls, infinite walls, and stepped or raised sections. This is a big advantage to have a possibility to consider the water-on-deck problem for the different types of ships including navy ships and fishing boats as the example. The LAMP system has become more powerful in this respect after applying the sophisticated method based on the solution of the shallow water flow equations using the finite volume strategy. This is a big step forward from the hydrodynamic development point of view. From the scientific, research and design point of view, the value of the LAMP system is very high as it enables one to consider the water-on-deck effects using the hydrostatic and Froude-Krylov pressure, semi empirical model, shallow water flow calculation, fully 3D flow calculation and shallow water flow using a novel finite volume strategy models which can be used independently. It enables one to validate the results obtained applying the different codes.

4.

Using the different levels of calculations, the LAMP system enables one to take into account the stages from water-on-deck to deck-in-water including the precise calculation of the volume of water-on-deck and heeling moment. Using the system the dumping process may be controlled at the design stage and the roll amplitudes may be predicted for different designs using the bulwarks or with no bulwarks. The subharmonic rolling, resonance conditions and how much the restoring moment is decreased due to water-on deck may be identified using the LAMP system as well. A big step forward from the design analysis point of view is to implement the LAMP system to consider the influence of the green-water-on-deck on the longitudinal motions and global loads. It is interesting that there is no influence of the water-on-deck on the heave and pitch values for the ship considered, although there is such an influence regarding the loads. It may that such a phenomena is typical for the naval ship under consideration. Possibly the new investigations are necessary. Multi-hull applications The 3-D formulation and general geometry model enables the LAMP system easily to be applied for analysing the multi-hull ships including High Speed Craft and offshore structures. There is necessity to perform more investigations regarding the mentioned designs. Large amplitude motions in frequency domain The time domain simulation approach applied within the LAMP system enables one to find how the nonlinear nature becomes evident for large amplitudes of response. It has been presented in the paper that for the nonlinear responses the results cannot be presented in the form of an RAO. It follows from the fact that the amplitude of response depends on the amplitude of wave excitation and may depend on initial conditions. This is a good example of progress in modern ship hydrodynamics. A big achievement presented in the paper is the proposal how the nonlinear responses can be presented in a frequency domain. This is distinguished work to show the way how after finding the steady state initial conditions two different responses for the same frequency have been obtained. Thus, two steady state responses might constitute the fold bifurcation confirmed experimentally. It brings the new knowledge how to identify the ship safety in the case of large amplitude motions in frequency domain at the design stage. By applying the nonlinear dynamics methods to the results obtained with the LAMP simulations it was possible to prove that the results of time domain simulation may be presented in the frequency domain. This is a great idea and naval architects look forward to seeing it developed.

Stefan Grochowalski, member


The authors are to be commended for continuing the development of LAMP program. The time-domain simulation approach seems to be, at present, the only way to represent realistically ship behavior and wave loads in

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 279

ABS TECHNICAL PAPERS 2003

severe wave conditions. It provides tools for analysis and prediction of ship dynamics in the range of amplitudes of motions where the strong nonlinearities preclude the use of traditional frequency domain methods. In order to be a reliable tool, the time-domain numerical model has to represent the hydrodynamic forces for the full range of motion amplitudes, including ship positions at which the deck get immersed in water. The deck/green-water interaction may dramatically change ship behavior and load distribution in extreme waves. It consists of two different physical phenomena: Flow of water trapped on deck, called in the paper water-on-deck, and Additional nonlinear hydrodynamic forces on the deeply immersed part of the deck, induced if there is a significant transverse relative motion of the immersed deck with respect to the surrounding water.

simulations in severe, oblique waves when the deck is frequently, deeply immersed, against appropriate model test results, including capsize events. Such an experimental data for ship behavior in extreme, oblique waves exist and is well documented (Grochowalski, 1989). It could provide an invaluable basis for the ultimate validation of the LAMP code. The authors indicate that they are aware of the problem of simulating the maneuvering and course control in large, oblique seas. From my own experience, I can say that this is one of the most challenging tasks, and the relevant simulation of course keeping defines the accuracy of the final prediction of ship behavior in extreme waves. Once again, I wish to congratulate the authors for the excellent paper, and am pleased to learn that the development of the LAMP simulation program will be continued.
Additional References

The latter was presented by Grochowalski (1989 and 1993) and is called deck-in-water effects. It has to be emphasized that water-on-deck and deck-in-water symbolize two different physical phenomena and require two different theoretical modeling (Grochowalski et al., 1998). The paper presents a significant improvement of LAMP program by incorporating calculation of greenwater flow on deck. Of the free approaches to water-ondeck problem presented, the shallow water flow model seems to be the best one. The flow in restricted space in the form of waves induced by deck motion can be well modeled, and the resulting loads and exciting forces subsequently calculated (Huang and Hsiung, 1996, and Grochowalski et al., 1998). The essential element in the water-on-deck problem is calculation of the instantaneous volume of water in the deck well. The paper indicates that in the calculation of water flow on deck, the water inflow and outflow is computed in order to find the real volume of sloshing water. However, the methods used are not given. Could the authors provide some details how the mass of water shipping on deck and escaping off the deck is calculated? The examples presented in the paper show the advantage of the modular structure of the simulation program. This is very welcome and very important feature. It provides possibility of further development of the program by incorporation of various numerical or empirical models of individual physical phenomena, that otherwise could not be considered. In order to achieve completeness in the modeling of deck-water interaction, the missing element, the deck-inwater forces (Grochowalski 1989, 1993) could be incorporated as a separate module. A preliminary mathematical model of the forces has been developed and was used in the simulations presented in Grochowalski et al. (1998). This model has yet to be validated experimentally, but the qualitative results are good. The other element which is missing in the development of LAMP, so far, is validation of numerical

Grochowalski, S. (1989). Investigation into the Physics of Ship Capsizing by Combined Captive and Free Running Model Tests. Trans. SNAME 97, 1989. Grochowalski, S. (1993). Effect of Bulwark and Deck Edge Submergence in Dynamics of Ship Capsizing. Proc. US Coast Guard Vessel Stability Symposium 93, New London, CT. Huang, Z. J. and Hsiung C.C., (1996). Nonlinear Shallow-Water Flow on Deck. J. of Ship Research 40 (4). Grochowalski, S., Hsiung, C.C., Huang, Z.J., Cong, L.Z. (1998). Theoretical Modeling of Ship Motions and Capsizing in Large and Steep Waves. Trans. SNAME 106, 1998.

Authors Closure
The authors appreciate the opportunity provided by SNAME to present the paper on this important topic. We also would like to thank each of the discussers for their interest in the paper, questions, and comments. As brought up by Dr. McTaggart, the general bodynonlinear (LAMP-4) approach is potentially more accurate than the approximate body-nonlinear (LAMP-2) approach but at a not-inconsiderable cost: the LAMP-4 calculation will typically take 5 to 10 times more computational time than the corresponding LAMP-2 run using the conventional, direct calculation technique. In the validation studies done to date, significant differences between these two motions and load calculations have primarily been seen only in very large, very steep regular wave cases, which is why we feel comfortable in using LAMP-2 for general engineering calculations. On the other hand, LAMP-4 does provide a more consistent hull pressure distribution for detailed pressure post-processing such as generating load data sets for Finite-Element structural analysis. In terms of computational time vs. real time, the comparison will depend greatly on the size and type of

280 Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design

ABS TECHNICAL PAPERS 2003

problem being computed, but a typical LAMP-2 ship calculation may run at a computational to real-time ratio of between 1:1 and 4:1 on a modern workstation or highend PC, while LAMP-4 will typically run a ratio of 10:1 to 40:1. However, a major advantage of LAMP-2 is that it can use the IRF-based approach, which pre-calculates portions of the hydrodynamic problem. With the IRFbased technique, LAMP-2 response calculations can run significantly faster than real time, with ratios from 1:5 up to 1:100. Another question raised by Dr. McTaggart concerned viscous and lifting forces. In the present calculation, the potential flow computation of the wave-body hydrodynamic forces is complemented by a series of models for the forces due to viscous roll damping, bilge keels, rudders, internal tanks, etc., some of which include active control systems such as the course keeping autopilot. Most of these models are based on empirical formulae, simple analytic expressions, or even externally specified derived coefficients, but we are also working on more sophisticated models including a fully integrated vortex lattice calculation for unsteady appendage forces. Dr. Grochowalski raises the question of how the inflow and outflow are computed in order to determine the mass of water on deck. At each time step, the amount of water shipped out (or escaping) through each deck edge element is calculated from the current deck edge water height and velocity of the internal (water-on-deck) domain and from the relative motion of the deck to the incident wave surface in the external (wave-body) domain. The details of this calculation can be found in Liut, et al.. (2002) and Belenky, et al.. (2002). Dr. Gerigk comments on the lack of influence that the green-water-on-deck has on the heave and pitch of the CG-47 cruiser in head seas. In these test cases, the mass of green-water is relatively small in comparison with the ships mass and is on the deck for a relatively short time. During that time, however, the green-water on foredeck has significant influence on loads, especially on local loads on foredeck structure. The result is not dissimilar to the comparison calculations for this ship that was presented in Weems, et al.. (1996): the large bow flare has small effect on the motions in extreme waves and a very large effect on the sectional loads. Mr. Hutchison is correct in his comment on the frequency domain roll analysis that the restoring forces and moments are calculated by integrating hydrostatic pressure over the instantaneous submerged hull surface. Mr. Hutchison and Prof. Falzarano discuss a number of issues related to nonlinear dynamics part of the paper. The focus of this part was to check how numerical simulations (LAMP, in this case) are able to reproduce nonlinear phenomena, known both from nonlinear dynamics and model tests by Francescutto et al. (1994) as referenced in the paper. To make such comparison comprehendible and vivid, only roll calculations were performed for this analysis. Nonlinear dynamics in general is a very powerful tool to interpret the results of numerical simulation. What we seek first is a qualitative similarity between approximate

ordinary differential equation and simulation model that includes a more accurate description of hydrodynamics. In other words, it is not necessary to get good quantitative comparison between an approximate model and the simulation (numerical simulation would be close to reality anyway); it is much more important to get good qualitative comparison, which would enable us to distinguish between physical phenomena and numerical artifact. As noted in the very beginning, this paper summarizes developments, experience, and results gathered since the previous LAMP presentation at an annual SNAME meeting in 1997. As a result, the paper was comprised from many different researches, which sometimes aimed at different objectives and applications. These sources, however, are united in their approach to the problems they pursue: time-domain numerical simulation. Prof. Yeung in an oral comment addressed his earlier work in the application of Rankin source approach in time domain as described in the additional references below. The authors are especially grateful to the principal reviewer of this paper, Dr. Arthur Reed. He spotted and corrected many hard-to-find inconstancies, small technical errors, and typos, and his comments on the papers contents and presentation made the paper better and significantly improved its readability.
Additional References

Yeung, R. W. (1973) "A Singularity-Distribution Method for Free-Surface Flow Problems with an Oscillating Body", Report NA73-6, College of Engineering, University of California, Berkeley. Bai, K. J. and Yeung, R. W., (1974) "Numerical Solutions to Free-Surface Flow Problems", with K. J. Bai, Proceedings, 10th Symp. on Naval Hydrodynamics, Cambridge, MA, pp. 609-646 Yeung, R. W., (1985) "A Comparative Evaluation of Numerical Methods in Free-Surface Flows", Invited review, Proceedings, IUTAM Symp. on Hydrodynamics of Utilization of Ocean-Wave Energy, Lisbon, Portugal, pp. 325-356, July, 1985. Yeung, R. W. and Cermelli, C., (1993) "Shell Functions: A Global Method for Computing TimeDependent Free-surface Flows", Proceedings, 8th Int. Workshop on Water Waves & Floating Bodies, St. Johns, Newfoundland, Canada, May, 1993.

Nonlinear Time Domain Simulation Technology for Seakeeping and Wave-Load Analysis for Modern Ship Design 281

Das könnte Ihnen auch gefallen