1. Introduction
Roll motion of a ship has always been a prime concern for many researchers in the marine field due to its importance and its possible consequence on the ship stability, operability, cargo handling activities, or comfort on board and sometimes even on the survivability of the ship. It is also considered as one of the least understood ship motion among all the six degrees of freedom of a ship as it was argued by Falzarano et al. [
1], who correlated this lack of understanding to the shortage in predicting the viscous roll damping. The damping effect is one of the most important aspects that should be taken into consideration when talking about roll motion prediction, especially if the ship is designed to sail in moderate or rough sea conditions, where the parametric rolling and possibly loss of stability are subjected to occur. Roll damping is mainly dominated by the viscous effect besides the interaction of the ship hull with free-surface, waves, and wind [
2]. Unlike the other degrees of freedom of a ship that can be predicted accurately using the potential based methods, roll motion requires a special modification to include the viscous effect. This could be applied based on either experimental or analytical based approaches. In the experimental method, the roll damping can be estimated based on one of three techniques: roll decay, forced roll, and excited roll tests. Various research can be found in the literature and all are aimed at predicting the roll damping based on the experimental approaches. Irvine et al. [
3] reported the free roll decay towing tank test results for the DTMB-5512 model at different Froude numbers in which they concluded that the damping period and the damping coefficient increases as the bilge keel is applied to the hull and also the damping effect increases as a result for the increment in both ship speed and the initial roll angle. The results also included the Particle Image Velocimetry (PIV) measurements for the flow field around the hull and the free-surface. These measurements were the reference data used for comparison in the roll decay case that was proposed for analysis in the Gothenburg Workshop on CFD in Ship Hydrodynamics [
4]. Moreover, these data are used for validating the numerically obtained results in this current study. Aloisio and Di Felice [
5] represented the results of the experimental analysis of the flow field around the bilge keel of the DTMB-5415 model in a forced roll decay test. The transfer of energy between the hull and the surrounding flow were highlighted by focusing on the vortical structures formation around the bilge keel at zero speed and at the medium speed range, i.e.,
Fr = 0.138. A free oscillation test has been performed and presented in [
6] to predict the roll damping of a floating ship-shaped structure model with various bilge keel dimensions and at different drafts. The measured data were used to validate the CFD study performed and presented in the same research, which concluded a proper agreement between the computed and measured data. A systematic series of roll decay tests with a single degree-of-freedom at various speeds for the DTMB-5617 model have been performed in a free decay and forced oscillation tests [
7]. The measured flow features resulted in a similar conclusion as the one reported by Aloisio in [
5], where the flow separations at the tip of the bilge keel and vortex formations during the roll period have been observed and reported. Roll decay test of intact and damaged ship model has also been reported in [
8] for different initial roll angles, results were also used to validate the CFD study for the same test conditions. Generally speaking, the experimental method is considered as the most accurate and straight forward technique; however, it has some drawbacks represented in the complex model preparation, time consumption, significant overall cost, and the fact that it is not flexible for optimization purposes, as multiple model configurations may be necessary.
On the analytical side, the historically considered effort in analyzing the roll damping started in the early 1950s using some analytical formulations deduced from experimental data. Despite the fact that those methods were outdated and mostly limited for zero speed roll motion, it can be said that these analytical formulations have established the basic principles in understanding the fundamentals of roll motion and its major influencing parameters [
9]. The highly recognized turnover in the roll damping researches returns to the late 1970s; thanks to the work of Ikeda and Himeno [
10] which resulted in the roll damping components hypothesis in which the total roll damping coefficient is divided into various components: friction; eddy; lift, wave, and finally, bilge keel. They could also provide empirical formulas deduced from extensive experimental tests on conventional ship models for predicting these components. Later, several corrections have been applied to improve these formulas, an effort that ended by the modified or simplified Ikeda method presented in [
11] which was proposed to predict the roll damping for any ship in the early design stage. Nevertheless, regardless of the popularity of these analytical methods, they are either limited for special types of ships or the components formulas were highly simplified.
Computational Fluid Dynamics (CFD) in the past two decades has played a significant role in predicting ship hydrodynamic performance. Thanks to the significant development in physical and numerical modeling techniques, it is now possible to study and investigate more complex ship hydrodynamic problems on both levels for model and full scale ships. This development was also accompanied with a massive evolution in computational powers that resulted recently in the High Performance Computing (HPC) which opened the gate for using finer grids, less physical modeling assumptions and faster simulation turnaround time. Furthermore, the augmented awareness of computational errors and uncertainties that was established through the systematic verification and validation methods increased the confidence in the numerically obtained results and set the milestone of understanding and resolving numerical simulation problems. The integration between the numerical and experimental methods can help concurring more complex fields in ship hydrodynamics that require more understanding and disclosing such as the roll damping problem. Like the other problems in ship hydrodynamics, roll decay has received a special focus from CFD researches. Most of the performed studies were based on the unsteady Reynolds-Averaged Navier–Stokes (RANS) model with the majority of discretization based on Finite Volume Method (FVM) and few based on the Finite Difference Method (FDM). Grid generation was mostly based on the overset or sliding grid technique to stand for the large roll angles, while very few researches were based on deforming grids. Time step was selected mostly according to the criteria proposed by the recommended procedures of the International Towing Tank Conference (ITTC) [
12] or even a smaller time step for results boost. Verification and validation studies were performed based on a grid convergence and time step convergence criteria. A study for the roll decay of the DTMB ship model with and without bilge keel at different conditions is presented in [
13]. The results showed a good agreement with the Experimental Fluid Dynamics (EFD) data especially for the case when the ship was appended with the bilge keels. On the other hand, a high error was reported for the damping coefficient estimated for the ship without the bilge keel. The DTMB ship model was presented for CFD study in the roll decay condition in the Gothenburg 2010 Workshop [
4], with four different participants in the study using four different software the results were promising and the agreement between the presented results and the provided EFD data was satisfying, a fact that indicated the capability of the CFD method to predict accurately the roll decay of a ship [
4]. A consistent study for the DTMB ship model equipped with the bilge keels in the free roll decay condition is solved numerically using FVM, RANS solver with a deforming grid is presented in [
14]. Intact and damaged ship condition for the DTMB ship model and for a Burgundy section model are presented in [
15] and [
16], respectively. The outcome from both studies showed the capability of the CFD method in predicting the roll decay of the ship in both, intact and damaged ship conditions with a satisfying level of accuracy.
Following the same effort performed by other researchers regarding this topic, this study investigates the capability of CFD to predict the roll decay of the DTMB ship model at different ship speed and different initial roll angel. The free-surface and the flow configuration around the hull in different simulation conditions is thoroughly investigated in order to understand the mechanism of the free-surface deformation during the roll decay process and the vortex formations around the hull and more specifically around the bilge keels. To insure the consistency of the computed results, a verification and validation study for the numerical results is performed and reported in the following sections of this study.
2. Model Geometry, Characteristics, and Analysis Conditions
The ship model subjected to this study is the David Taylor Model Basin (DTMB), recently known as Naval Surface Warfare Center, Carderock Division (NSWC). The ship was introduced in the marine field in 1980 as a preliminary design of surface combatant. DTMB stands as an unconventional modern benchmark hull form, basically intended for flow explanation and CFD validation. The bare hull of the ship includes a sonar dome and transom stern; whereas the appended hull is equipped with two bilge keels, twin open-propellers driven by propeller shafts which are connected to the hull by means of brackets (struts) and finally, two rudders. In the current study, the hull is analyzed for roll decay. For this purpose, only the bare hull equipped with the bilge keels is introduced in the numerical simulations. The geometry of the bare hull model with bilge keels is depicted in
Figure 1, while the characteristic dimensions and main particulars of the DTMB-5512 model with a scale λ = 46.6 compared to the full-scale ship are tabulated in
Table 1.
The model was introduced as one of the three benchmark geometries in the Workshop on CFD in Ship Hydrodynamics that was held in Gothenburg in 2000 [
17] along with two other geometries for conventional commercial ships; the first is the 300 K modern tanker called KVLCC, while the second is a 3600 TEU container model known as KCS. Both hulls were designed by the Korea Research Institute of Ships and Ocean Engineering (KRISO) formerly known as the Maritime and Ocean Engineering Research Institute (MOERI). The full scale DTMB ship does not exist; however, various geometrically similar models were built, tested and reported all over the world. Unlimited data base is available for the ship model in the public domain from experimental tests reported by very well recognized towing tank tests facilities such as the Istituto Nazionale per Studi Ed Esperienze di Architettura Navale (INSEAN) in Italy, Iowa Institute of Hydraulic Research (IIHR) which is now titled as IIHR–Hydroscience & Engineering in the United States of America, FORCE Technology of Denmark, Maritime Research Institute of Netherlands, (MARIN) and many others. Towing tank tests were performed and reported for different aspects in ship hydrodynamics, such as resistance, free-surface, local flow, propulsion, seakeeping, and maneuvering, all aimed at understanding the flow physics and providing a verification and validation reference for CFD applications.
Table 2 provides a summary for the basic and most famous test cases performed on the DTMB models and reported by the previously listed institutes and research centers.
From these tests, this study is an attempt by the authors to replicate the roll decay experiment performed by Irvine et al. in [
3] using the CFD method, in order to validate the capability of the RANS modeling to predict the roll performance of the ship. This work is an extension of the previous attempt by the author that was introduced in [
21] which was focusing on the same concept without taking into consideration the systematic verification of the CFD data, grid, and time-step effect on the numerical simulation, prediction, and validation of the viscous flow around the hull and particularly around the bilge keels and how it is affected by the turbulence model used. This is covered consistently in this study.
The analysis condition includes two levels of numerical simulations, the first intends to validate the numerical simulation against the experimental data, where the ship is sailing at a medium speed corresponding to a Froude number
Fr = 0.138 where
. This analysis condition covers different initial roll angle starting from 2.5 to 20 degrees with a step 2.5 degrees in every simulation. A special focus is applied for the 10 degrees condition, where the validation and verification of numerical result against experimental data takes place, following the same concept of case 3.6 from Gothenburg Workshop on CFD in ship hydrodynamics that was reported and summarized in the Workshop book [
4]. The second level is aimed at investigating the effect of the ship speed on the roll damping of the ship. For this purpose, a set of five ship speeds is analyzed and reported for corresponding Froude numbers
Fr = 0, 0.138, 0.20, 28, and 0.41 The viscous flow configuration around the ship hull and specifically around the bilge keel is also investigated and introduced.
5. Conclusions
The CFD simulation for the free roll decay of the DTMB 5512 surface combatant ship model sailing in calm water and appended with bilge keels was performed based on unsteady RANS where the closure to turbulence was achieved using the K-ω SST and the EASM models. The simulation is executed for different initial roll angles and for various ship speeds in order to investigate their influence on the roll decay of the ship. Standard deforming unstructured grid approach with a special refinement in the vicinity of the hull and close to the interface was implemented to count for the expected large ship motion rather than using overset or sliding grid, where they have some relative drawbacks compared to the standard grids represented in their complexity and the large required memory for simulation. The initial validation of the roll decay at different initial roll angles was qualitatively and quantitatively investigated in comparison with the EFD data showing that the resemblance between the CFD and EFD was satisfying, especially for the computed roll period which was estimated with less than 2% error; on the other hand, the discrepancies in roll amplitudes were significant for the second and sometimes the third roll period with an error up to 24.28%. The amplitude deviation was increasing as the initial roll angle increases.
Verification and validation (V&V) study based on the Richardson Extrapolation method was performed based on grid and time step convergence study with four geometrically similar grids and four time steps, respectively. The V&V study showed a global monotonic convergence for the predicted roll amplitude and occasionally local oscillatory convergence as the finest grids results were very close in values. Validation was not achieved at the validation uncertainty level where the average estimated error for computed data was 4.245% and = 3.378%; nevertheless, the average error was considered acceptable especially for the finest grids.
Since the roll decay is dominated by the viscous effect, a special focus on the flow in the vicinity of the hull and specifically at the bilge keel was presented by means of velocity field, TKE and vortex formation showing the interaction between the hull and surrounding flow during roll motion, which tends to impose an oscillation in the vortical tubes dissipated from the sonar dome and the keel vortices, both horizontally and vertically. On the other hand, the vortex formation around the bilge keels and its development was also represented showing the interaction between the bilge keel and the surrounding flow during the roll second period.
Roll decay influence on the free-surface was also illustrated showing the effect of the body motion on the Kelvin wave pattern during the simulation, which resulted in a slightly deformed Kelvin wave pattern due to the viscous and pressure effects imposed by the roll damping process. In addition, the effect of the speed on the free surface, as well as on the roll damping, which showed to be compatible with the same conclusions for the similar tank tests. For the low speed cases, it was concluded that only 60 cells/wave length is not adequate to predict the full development of the free-surface at that speed.
Overall, comparing the main objective of this study with the outcomes, it can be said that the implemented approach was successful in predicting the roll decay of the ship. Some discrepancies exist for different parameters, yet they are not very significant and they could be considered more than satisfying for initial design and for optimization purposes. Though the simulation time was relatively high, yet the flexibility of the simulation method makes it feasible compared to experimental method. Obviously, the integration between CFD and EFD is the key for improving the ship design process and for solving more complicated ship hydrodynamic problems.
Despite the fact that the turbulence model performed well, as it was illustrated in this study, more details in the flow could be captured using DES or LES models and can help in providing more understanding for the flow mechanism around the bilge keels. This can be a base for further investigation.