Hydrodynamics Prediction of High Speed Sea Lift (HSSL) Ships Joseph Gorski and Ronald Miller US Naval Surface Warfare Center, Carderock Division (NSWCCD), West Bethesda, MD {joseph.j.gorski, ronald.w.miller}@navy.mil Pablo Carrica, Mani Kandasamy, Tao Xing, and Fred Stern The University of Iowa, IIHR Hydroscience & Engineering, Iowa City, IA {pablo-carrica, tau-xing, frederickstern}@uiowa.edu and mkandasa@engineering.uiowa.edu Abstract High Speed Sea Lift (HSSL) is an important area of interest for the US Navy. Computational tools are needed to predict the hydrodynamics of these multihull configurations for their proper design and analysis in many areas including resistance and powering, motions and habitability, loads in service and maneuverability. To address these issues requires a hierarchy of computational and experimental capabilities. Described here are efforts underway to use Unsteady Reynolds Averaged Navier-Stokes (URANS) solvers to answer some of these needs. Specifically, computations are demonstrated with CFDSHIP-Iowa to predict relevant HSSL catamaran and trimaran configurations under a variety of operating conditions in both calm water and with waves. Many of these predictions, particularly those representing behavior in various sea states, are very computationally intensive and thus require high performance computing resources. The predictions demonstrated here are paving the way for a computational capability to aid in the design for a new generation of naval ships. 1. Introduction The US Navy is investigating high-speed sealift (HSSL) ships that will allow rapid deployment of forces from CONUS to foreign ports and needs computational tools to predict the hydrodynamics of these configurations. The computational hydrodynamic tools need to be able to predict the resistance of these multihull vessels, propulsive performance, seakeeping and structural loads through survival sea states, maneuvering characteristics, and the effects of shallow water on their performance. Computational capability is needed for early stage design analysis through to detailed evaluation of final configurations. Consequently, a suite of computational tools is being developed and evaluated for predicting the performance of these vessels from simple inviscid methods to high end viscous techniques requiring High Performance Computing Modernization Program (HPCMP) resources and is the subject of this paper. To predict viscous flow physics of HSSL concepts correctly the Unsteady Reynolds-Averaged Navier-Stokes (URANS) code CFDSHIP-Iowa is being used. Computations are demonstrated for both catamaran and trimaran vessels as these multihull configurations are expected to be the types of hulls needed to reach HSSL requirements as opposed to conventional monohulls. Additionally, it is expected that waterjet propulsion systems will be needed due to the speeds of interest. Waterjet propulsion is a new area for the US Navy and predictions are needed to provide hull/propulsor interaction as well as provide propulsion performance. Pertinent information includes calm water resistance as well as ship response in waves over a range of conditions. Because tools are needed to impact the design of these vessels it is necessary to achieve fast turnaround. There is also the need for detailed flow predictions at the main transit speeds of interest, which for the designs investigated here is currently 43 knots. Resistance predictions, waterjet propulsion, and behavior in regular waves are all demonstrated. These calculations for different configurations and operating conditions demonstrate the ability to use URANS codes in an efficient manner to support the Navy s need for design and analysis tools for these concepts. 2. CFDSHIP-Iowa CFDSHIP-Iowa is a general-purpose research URANS CFD code developed at the University of Iowa (UI) over the past ten years for support of student thesis and project research at UI as well as transition to Navy
laboratories, industry, and other universities. During this time the simulations have consistently demonstrated excellent results at ship hydrodynamics workshops and conferences. CFDSHIP-Iowa includes state-of-the-art modeling for: free surface, turbulence, inertial and noninertial coordinate systems, forced and predicted sixdegrees-of-freedom (6DOF) motions, propulsors, incident waves; numerical methods for higher-order conservative discretization, embedded overset grids, advanced iterative solvers, implicit coupling flow field and predicted motions; and HPC for MPI-based domain decomposition. Previous applications include surface combatant breaking bow waves (Wilson et al., 2004), roll decay (Wilson and Stern, 2002), forward speed diffraction (Carrica et al., 2005), and static drift propeller-hull-rudder interaction (Simonsen and Stern, 2003). Most applications follow rigorous verification and validation procedures. Recent applications include high-speed and shallow water (Wilson et al., 2005), wave-induced separation (Kandasamy et al., 2005) and optimization (Campana et al., 2004). For HSSL ships, extensions and further developments include: single-phase level-set free surface modeling for high speed flow; static overset grids for complex geometries and local refinement (Carrica et al., 2006); blended k-ω and detached-eddy-simulation (DES) turbulence models; higher-order finite-difference discretization, advanced iterative solvers (PETSC toolkit); high performance computing using an MPI-based domain decomposition approach; incident waves; and prescribed 6DOF motions. High speed (Fr=0.62) unsteady breaking bow and transom waves have been demonstrated for the Athena research vessel. It was shown that the use of overset refinement blocks was required to accurately resolve the wide range of physical scales associated with the free surface from the overturning bow wave sheet (~10-4 L) to the scale of the Kelvin wave pattern (~2L). Also, high-speed flow around the Wigley hull (up to Fr=0.99) was simulated with deep and shallow water. For the HSSL effort CFDSHIP-Iowa is applied to high-speed multihull cases for a variety of operating conditions including: ship motions, arbitrary heading, regular and irregular, unidirectional and multidirectional waves. To allow for the computation of large-amplitude motions a dynamic overset grid technology is used. This is accomplished using the interpolation tool SUGGAR (Noack, 2005). The implementation was validated for DTMB 5512 in regular head seas free to pitch and heave. For sinkage and trim calculations, artificial damping coefficients were used and validated for DTMB 5512. For fully appended ships overset grids over solid surfaces are used, which requires the evaluation of the weights of the different active cells that overlap over the solid surfaces. This was implemented as preprocessing steps using the code USURP (Boger, 2006). The new capability was tested for a fully appended Athena R/V with stabilizers, rudders, skeg, shafts, and struts. For massively separated flows, a DES model was implemented. This was tested for Athena R/V, both barehull and fully appended cases. In addition, the code robustness was increased significantly. Currently, waterjet propellers and screw propeller models, and multiple independent/dependent objects (ship-ship interaction, active control surfaces, at sea loads transfer) capabilities are being implemented. Initial simulations have produced results for HSSL demihull with water jet, and Athena with screw propeller using body force model. Multiple shipship interaction simulations have also been carried out for two independent DTMB 5512 hulls, one following the other, in SS6 irregular head waves. In addition, a new numerical towing tank for predicting the full-resistance curve with sinkage and trim by very slow acceleration, tested for Athena, is being implemented. 3. Naval Architect Needs During a ship design phase, naval architects need a large amount of information as the ship concept needs to be evaluated. This includes resistance and power, motions and habitability, loads in service and maneuverability. For resistance and power both effective and delivered power are required as well as added resistance in waves. Motions and habitability typically involves rigid body 6DOF motions and the computation of Response Amplitude Operators (RAOs). From the computations of the hydrodynamics of the ship operating in waves comes the ability to provide load estimates. Beyond the computation of RAOs and loads is the very important aspect of evaluating operability of the ship. Such operability evaluations might entail determining limiting wave heights for various operational scenarios such as transits, underway replenishments, and aircraft operations. Finally, maneuverability needs to be assessed, both in terms of stability and also the ability to predict the maneuvering behavior of the ship. To evaluate all of these areas requires large numbers of computations and/or experiments of varying fidelity and speed. Because of the high cost of experiments, an accurate and economical computational capability is clearly desired, particularly as a design tool to evaluate a number of candidate designs. Computationally, a hierarchy of methods is required of different fidelities, and trade-offs must be made between the speed with which information is needed and the physics, which must be properly accounted for. At the higher end of the spectrum are URANS simulations, which require high performance computing resources. Some areas where URANS contributes to the process are discussed below.
3.1. Resistance Resistance is often the first thing evaluated during a hull design and thus is very important as it is a deciding factor early on in the design stage. For resistance evaluations calculations are performed with calm water over a range of speeds. The current area of interest for HSSL is in the Froude number range of approximately 0.25 to 0.75. Enough different speeds need to be computed to properly identify sensitive areas where resistance changes rapidly with Froude number. CFDSHIP-Iowa has capability to predict resistance directly, including the prediction of sinkage and trim. For HSSL concepts sinkage and trim changes can be significant as it is not uncommon for ships to become semi-planning at these high speeds. Since this may have a significant impact on resistance it should be part of the calculation for reliable design and analysis. In addition to resistance, the boundary layer flow and any vortical flow created by the hull or hull interactions can be evaluated with the URANS code. The design speed of most interest currently is about 43 knots at full scale, equivalent to a Froude number of 0.55, so one would commonly do more detailed analysis at this speed to study the flow and interactions between the hulls. An example computation for Model 5594, a very slender high-speed trimaran hull form with small sidehulls (Figure 1), is shown here. While the hull form is different from those envisioned for HSSL, there is available experimental data at high-speed for evaluating computational capability. Additionally, it has many features that are representative of HSSL concepts including: hull slenderness, large transoms, hull bow bulb, raked-back stem, and multi-hull interactions. The resistance curve, Figure 2, shows a hump of higher drag at about Fr=0.5, which is near the Froude number of most interest. A similar increase in drag resistance has also been seen for other catamaran and trimaran hull forms investigated. Since this is near the transit speed of primary interest an important aspect of HSSL design is attempting to minimize resistance at these speeds. This typically involves trying to minimize hull interaction or even use the hull interaction to cancel wave drag. The predicted resistance gives the same trend as the measured, but under-predicts the measured amount by over 10%, which is not insignificant. The computed sinkage agrees very well with the measured data and even the trim, although slightly under-predicted, is in good agreement. To predict the sinkage and trim so well, but under-predict the resistance as shown here is currently under investigation. Full resistance curves predicting sinkage and trim are usually run with grids on the order of two million points. To generate a curve for the entire speed range, a single long run, corresponding to an infinitely long tow tank where the model is slowly accelerated from rest to the highest speed of interest, is employed. Such an approximation may only be done computationally. The acceleration must be small enough that each speed represents the resistance that would be obtained for a steady state run and spot checks have confirmed the approach. Implementing the approach is convenient and requires little intervention once started. Because the acceleration must be small, long times, on the order of 1000 seconds full scale time, are required consuming about 5,000 CPU hours on 24~32 processors for these configurations. More details of resistance computations of Model 5594 as well as the Athena hull form at high speeds is documented in Miller et al. (2006). Figure 1. Model 5594 geometry Figure 2. Computed and measured total resistance, sinkage and trim for Model 5594 3.2. Added Resistance It is rare to have truly calm water in the ocean and ships routinely operate in waves. A first step beyond evaluating a ship design for calm water resistance is predicting the added resistance associated with operating in waves. Unsteady calculations were performed here for a catamaran with regular incident waves with the hulls fixed (diffraction) or free to move in response to the waves to demonstrate CFDSHIP-Iowa s ability to simulate such flows and as a mechanism to provide added resistance. For the regular incident head wave, a nondimensionalized amplitude corresponding to a SS6 (sea state 6) wave height and a wavelength of 1.411 ship lengths is used. Approximately 100 time steps per period
are used in the calculation. The solution is started abruptly, so that a few periods are required before a periodic solution is obtained. Because the ship is heaving and pitching, the wetted surface area changes in a cycle, which changes the resistance. Additionally, the interaction of the heaving and pitching ship with the wave field creates additional forces. The resistances vs. time results are shown in Figure 3. The steady state calm water resistance is shown as the thick black line of constant value. The diffraction solution has not yet obtained periodicity, but the solution indicates a linear type response to the regular incident input wave. The nonlinear response of the pitching and heaving catamaran is shown to have multiple oscillations occur for each period of the incident wave. The average oscillating force also appears to be greater than the steady state force. Figure 3. Catamaran resistance versus time in regular waves Shown in Figure 4 are some details of the pitching and heaving catamaran for two instances of a half period corresponding to extreme conditions when the hull has its minimum (top figures) and maximum (bottom figures) wetted surface area exposure. The figures in the right column show pressure contours on the wetted portion of the hull, the center figures show the computed free surface contours, and the figures on the left show a side view of the catamaran, in relation to the incident wave. Since pitch and heave of the catamaran are computed directly its position changes throughout the cycle. When the wetted area is at its minimum the pitch is close to its greatest bow up position at nearly 6 degrees and the heave is at its maximum of 0.04L. A trough of the incident wave is beneath the bow at this point, which is seen in the figure to the left. Both the pitch and the heave will begin decreasing at this point. At approximately ¼ period, the wetted area is about the same as its static or calm water. At this point, the vertical displacement is almost zero and still decreasing, while the pitch has nearly reached its most bow down position at approximately 2 degree. This is the position just beyond where a spike in both the drag and vertical force occurs, which might be a result of the bow re-entering (slamming) the water. When the wetted area is at its maximum the vertical displacement is near its maximum downward position of a negative 0.04L. The pitch is zero, but increasing. At this point, a crest of the incident wave is at the bow of the boat and water on deck can occur depending on incident wave height. Typical run times with 2.5 million grid points on 24 processors were approximately 1500 CPU hours for simulations of 34 seconds full scale time. 3.3. Maneuvering This exercise demonstrated CFDSHIP-Iowa s capability to predict the forces and flow field about a HSSL catamaran concept resulting from prescribed planar maneuvers in calm seas and in regular incident head waves. Typically, to evaluate a model for maneuvering information tests are run with a planar motion mechanism (PMM) machine, which imparts oscillating or zigzag motions as a model is towed down a tank. This is a corresponding computation where the sway and yaw motion, given by sway = 0.07218 sin(2.039 t ) and yaw = -8.4 cos(2.039 t), is imparted and the change in forces computed. About 160 time steps per period are used. The motion was started abruptly, so that some transients will likely appear in the results. The hull is in the fixed static orientation. The plots on the right side of Figure 5 show the calculated lateral force for three periods of the prescribed planar motion. The solid black curves show the results for the maneuver in calm seas and the dashed red curves show the results in regular head waves. The calculations show that extreme side forces occur at the extreme sway locations, where the turning is the greatest, and the least occur when the boat is crossing the centerline of the maneuver. The incident waves have a large effect on the lateral forces at some places doubling the magnitude and other places changing the direction. The figure on the left illustrates the flow field when the catamaran is just moving across the centerline. The large asymmetric bow wave may be seen due to the forward and sideward motion plus the turning at the bow. Beneath the hull the figure illustrates the vorticity created by a combination of the sideward motion and the rotation of the boat. At this instant, vorticity can be seen on one side of the boat in the forward region and on the other side in the aft region.
3.4. Seakeeping Seakeeping is an important component of ship evaluation and much of it is done in regular waves at various orientations to the path of the ship. Head wave conditions have already been demonstrated for the added resistance computation. Additionally, for the PMM simulations in regular waves the orientation of the bow to the wave changes as it goes through its prescribed motion. Separately, computations have been performed for both the catamaran and a trimaran concept at various headings in relation to the regular wave field. In addition to regular waves, irregular wave simulations using a multidirectional Bretschneider spectrum have been done using CFDSHIP-Iowa. Shown in Figure 6 is the wave field computed for the trimaran in 45 degree quartering waves at 43 knots in SS6 giving an indication of the complicated interactions of the wave field with the ship. Other computations have been done with 135 degree quartering waves at 20 knots. For these irregular waves trimaran runs, the total grid size was 4.5 million grid points. Each case used 48 processors with total wall clock time of 100 hours (4800 CPU hours) to predict ship motions and flow field for 3.5 nondimensional seconds (27 seconds full scale time). These are representative times for this type of run. Further study needs to be done to determine how many periods should be run for regular waves to obtain meaningful data. For irregular waves repeatability is not expected and studies usually need to be run for long periods of time to investigate the potential for extreme events, slamming, and other nonlinear effects. 4. Summary High speed sea lift ship configurations provide many of the same challenges associated with traditional monohulls as well as additional challenges related to hull interactions and other design considerations where experience is limited. Consequently, computational tools are needed to predict the hydrodynamics of these configurations. In particular, computational approaches requiring a minimum of empiricism are desired as there is a limited experimental database available. To achieve this, efforts are underway to apply high-end URANS computations to these configurations in nearly all aspects relevant to their hydrodynamics analysis and design. Shown here are demonstrations with the URANS solver CFDSHIP-Iowa for catamaran and trimaran hull forms for a variety of speeds in calm water as well as waves. Computations are demonstrated related to areas of resistance, maneuvering, and seakeeping. Sinkage and trim are well predicted. The resistance itself is currently under-predicted, but the changes with speed are captured. This is presently being investigated. The results clearly demonstrate the ability of the software to numerically predict the response of the hulls to various ship speeds and sea states. Significant efforts are still underway and must continue to be pursued to demonstrate the accuracy of the computations in all these areas as well as what may be done to efficiently achieve good accuracy. This is especially important for the behavior in waves where it is necessary to evaluate performance for a large variety of headings and wave fields that could require hundreds if not thousands of simulations, imposing a significant drain on available HPC resources. Such a computational approach is clearly needed for the design and analysis of future US Navy ships having HSSL capability. Acknowledgements The authors are grateful to the US Department of Defense s High Performance Computing Modernization Program Office (HPCMPO) that provided the computer resources at NAVO on the IBM P4 and at ASC on the SGI 03k as part of this Challenge Project. The research is supported by the Office of Naval Research under the administration of Dr. L. Patrick Purtell. The NSWCCD effort is also partially supported by In-House Laboratory Independent Research (ILIR) funds under the administration of Dr. John Barkyoumb. References Boger, A.D., User s Manual for USURP, Unique Surfaces using Ranked Polygons. Penn State University, Applied Research Laboratory Manual, 2006. Campana, E., D. Peri, Y. Tahara, and F. Stern, CFD Based Local Optimization of a Surface Combatant With Experimental Validation. Proc. 25 th ONR Symposium on Naval Hydrodynamics, St Johns, Canada, 2004. Carrica, P., R. Wilson, and f. Stern, Linear and Nonlinear Response of Forward Speed Diffraction for a Surface Combatant. CFD Workshop Tokyo 2005, National Maritime Research Institute, Mitaka, Tokyo, Japan, 9 11 March 2005. Carrica, P.M., R.V. Wilson, R. Noack, T. Xing, M. Kandasamy, and F. Stern, A Dynamic Overset, Single-Phase Level Set Approach for Viscous Ship Flows and Large Amplitude Motions and Maneuvering. to be presented at 26 th symposium of Naval Hydrodynamics, Rome, Italy, 2006. Kandasamy, Xing, Wilson, and Stern, Vortical and Turbulence Instabilities in Unsteady Free-Surface Wave-Induced Separation. 5 th Osaka Colloquium, Osaka, Japan, 14 15 March, 2005. Miller, R., P. Carrica, M. Kandasamy, T. Xing, J. Gorski, J. and F. Stern, Resistance Predictions of High Speed Mono and Multihull Ships with and without Water Jet Propulsors using URANS. to be presented at the 26 th Symposium on Naval Hydrodynamics, Rome, Italy, Sept. 2006.
Noack, R., SUGGAR: a General Capability for Moving Body Overset Grid Assembly, 17 th AIAA Computational Fluid Dynamics Conf., Toronto, Ontario, Canada, 2005. Simonsen, C. and F. Stern, RANS Maneuvering Simulation of Esso Osaka with Rudder and a Body-Force Propeller. HYDRONAV 03, Gdansk Poland, 2003. Wilson, R., P. Carrica, M. Hyman, and F. Stern, An unsteady Single-Phase Level-Set Method with application to breaking waves and Forward speed diffraction problem. Proc. 25 th ONR symposium on naval hydrodynamics, St Johns, Canada, 8 13 August 2004. Wilson, R., N. Sakamoto, and F. Stern, RANS Simulations for High Speed Ships in Deep and Shallow Water. Proceedings 8 th International Conference on Fast Sea Transportation, Saint Petersburg, Russia, 27 30 June 2005. Wilson, R. and F. Stern, Unsteady RANS Simulation for Surface Combatant Roll Decay. Proc. 24 th ONR Symposium on Naval Hydrodynamics, Fukuoka, Japan, 8 13 July 2002. Figure 5. Catamaran simulation of PMM Figure 6. Trimaran in 45 degree irregular waves (43 knots) Figure 4. Catamaran in regular head waves