nach oben

2014 | Buch

Kapitel lesen Erstes Kapitel lesen

Numerical Ship Hydrodynamics

An assessment of the Gothenburg 2010 Workshop

herausgegeben von: Lars Larsson, Frederick Stern, Michel Visonneau

Verlag: Springer Netherlands

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

This book assesses the state-of-the-art in computational fluid dynamics (CFD) applied to ship hydrodynamics and provides guidelines for the future developments in the field based on the Gothenburg 2010 Workshop. It presents ship hull test cases, experimental data and submitted computational methods, conditions, grids and results. Analysis is made of errors for global (resistance, sinkage and trim and self-propulsion) and local flow (wave elevations and mean velocities and turbulence) variables, including standard deviations for global variables and propeller modeling for self-propulsion. The effects of grid size and turbulence models are evaluated for both global and local flow variables. Detailed analysis is made of turbulence modeling capabilities for capturing local flow physics. Errors are also analyzed for head-wave seakeeping and forward speed diffraction, and calm-water forward speed-roll decay. Resistance submissions are used to evaluate the error and uncertainty by means of a systematic verification and validation (V&V) study along with statistical investigations. Post-workshop experimental and computational studies are conducted and analyzed for evaluation of facility biases and to draw more concrete conclusions regarding the most reliable turbulence model, appropriate numerical methods and grid resolution requirements, respectively.

Inhaltsverzeichnis

Frontmatter

1. Introduction, Conclusions and Recommendations

Abstract

The Gothenburg 2010 Workshop on CFD in Hydrodynamics was the sixth in a series started in 1980. The purpose of the Workshops is to regularly assess the state of the art in Numerical Hydrodynamics and to provide guidelines for further developments in the area. The 2010 Workshop was by far the largest one so far, with 33 participating groups of CFD specialists and a larger number of test cases than before. All participants submitted their computed results during the fall of 2010. The results were compiled by the organizers and discussed at a meeting in Gothenburg in December 2010. In Chap. 1 the background and development of the Workshops since the start are presented. The three hulls used in the 2010 Workshop are introduced and the computations requested from the participants are specified. Based on a questionnaire sent to all participants the details of their CFD methods are listed, and finally the general conclusions and recommendation for future Workshops are presented. The detailed results of the computations are discussed in subsequent Chapters.

Lars Larsson, Frederick Stern, Michel Visonneau

2. Evaluation of Resistance, Sinkage and Trim, Self Propulsion and Wave Pattern Predictions

Abstract

In Chap. 2 results in several areas are discussed. Resistance predictions were requested for all three hulls and there is a large number of submissions. It is found that for grid sizes larger than 3M cells all submissions are within 4 % of the measured data. The mean comparison error (data-simulation) is only −0.1 % and the mean standard deviation is 2.1 % of the data value, excluding self-propelled cases for which the error is larger. There is no discernible effect of the turbulence model; two equation models work as well as the more advanced ones. Sinkage and trim exhibit large comparison errors for the smallest Froude numbers, but this is likely to be due to measurement inaccuracies. For Froude numbers above 0.2 the mean comparison error is around 4 % and the standard deviation 8–10 %. Self-propulsion results are reported with real operating propellers as well as modeled ones. For K _T and K _Q the mean comparison errors are 0.6 and − 2.6 % resp. The standard deviations are 7.0 and 6.0 % resp. Comparing actual and modeled propellers it is seen that the actual propeller has a much smaller mean standard deviation. For K _T it is half, and for K _Q it is only 1/3 of that of the modeled propeller. There is no clear advantage when it comes to comparison error, however. In general, the wave contour on the hull and at the wave cut closest to the hull is well predicted. Further away from the hull the results differ considerably between the methods. The best submissions for all three hulls capture all the features of the waves out to the edge of the measured region.

Lars Larsson, Lu Zou

3. Evaluation of Local Flow Predictions

Abstract

In this Chapter, the computations performed by all the contributors are analyzed from the point of view of the local flow analysis in order to assess the level of agreement between computations and local flow measurements and identify the sources of errors. Pure resistance with or without free-surface, hull/propeller coupling, wave diffraction or roll decay configurations are reviewed in detail. One observes a general improvement of the agreement between simulations and measurements and a strong consistency of the simulations, compared to the previous editions of this workshop. When a reasonably fine grid is employed, similar turbulence models provide similar results, independently of the code used, which illustrates the state of maturity of modern CFD methodologies. For resistance and hull/propeller coupling configurations, a detailed comparison of the best statistical turbulence models and LES or hybrid LES turbulence closures, introduced for the first time in this workshop, is conducted to identify their respective impact on local flow simulations.

Michel Visonneau

4. Evaluation of Seakeeping Predictions

Abstract

Test cases related to seakeeping are studied in this chapter including heave and pitch with or without surge motion in regular head waves for KVLCC2 and KCS and wave diffraction and roll-decay with forward speed for DTMB 5415. For seakeeping, the total average error is E = 23 %D, comparable to the average error for previous seakeeping predictions. For resistance, the largest error values are for the 1st harmonic amplitude and phase (34 %D), followed by 0th harmonic amplitude (18 %D) and steady (7 %D). For motions, the largest error values are for the 0th harmonic amplitudes (54 %D), followed by 1st harmonic amplitude and phase (13 %D) and steady (9 %D). The errors for the CFD predictions are similar for the different geometries and wavelengths, the small and large amplitude waves, and for the cases with and without surge motion. The errors are larger for the cases with zero forward speed. Compared with potential flow, CFD showed larger errors for motions for the medium and long wavelengths. For wave diffraction submissions, the large grid size DES simulation has achieved an average error value of less than 10 %D, while for the small grid size URANS simulations the average error is 28 %D. For roll decay submissions, the average error values are 10 %D for resistance and less than 1 %D for roll motions.

Frederick Stern, Hamid Sadat-Hosseini, Maysam Mousaviraad, Shanti Bhushan

5. A Verification and Validation Study Based on Resistance Submissions

Abstract

In Chap. 5 the database of ship total resistances submitted to the workshop is used to evaluate the error and uncertainty by means of a systematic verification and validation (V&V) study along with statistical investigations. Three representative methods are applied for verification: Grid Convergence Index, Factor of Safety and Least Squares Root. Validation of the results is carried out by the ASME V&V 20-2009 Standard. It is found that the iterative convergence is an important aspect in the numerical computation due to its contribution to the numerical uncertainty and its influence on the determination of discretization uncertainty. A limit for the iterative error is proposed. In the grid convergence study, unstructured grids are shown to more seldom achieve monotonic convergence than the structured grids. 2 to 10 million grid points and a grid refinement ratio 1.2 are most common among the research groups. In the study of structured grids using different verification methods, most solutions achieve monotonic convergence and are in the vicinity of the asymptotic range. Similar uncertainties are then predicted by the three methods. For cases further from the asymptotic range the methods predict quite different uncertainties. The scatter in solutions is an issue which is shown to significantly affect the determination of the grid convergence and the order of accuracy. In the validation study, the numerical error is mostly larger than the experimental error. Most solutions are estimated to have a smaller comparison error than validation error, implying that the modeling error is buried in the numerical and experimental noise.

Lu Zou, Lars Larsson

6. Additional Data for Resistance, Sinkage and Trim

Abstract

In this Chapter, additional resistance, sinkage and trim data are presented against Froude number for KVLCC2, KCS and DTMB 5415. Comparisons are made with the original data used at the Workshop. The purpose is to provide additional information useful for future validation of CFD results and to estimate the uncertainty in the data from the different facilities. However, due to lack of information about precision in most measurements only bias errors are estimated. For KVLCC2 and KCS one additional set of data is added to that used at the Workshop. The estimated bias errors in residuary resistance are very small, around 0.2 % of the mean total resistance, while those of sinkage and trim are considerably larger: 6–11 % of the mean values across the Froude number range. For 5415 new data from three organizations are presented. Bias errors in residuary resistance are 0.9–1.6 % of the mean total resistance. Sinkage errors are in the range 3–6 % of the mean value and trim errors around 0.01°.

Lu Zou, Lars Larsson

7. Post Workshop Computations and Analysis for KVLCC2 and 5415

Abstract

The Workshop submissions for the local flow predictions for straight ahead KVLCC2 and 5415 were on large disparate grids ranging from 0.6M to 300M, which made it difficult to draw concrete conclusions regarding the most reliable turbulence model, appropriate numerical method and grid resolution requirements. In this chapter, additional analysis including grid verification study is performed on intermediate grids to shed more light on these issues. Second order TVD or bounded central difference schemes are found to be sufficient for URANS, whereas fourth or higher order schemes are required for hybrid RANS/LES (HRLES). Resistance predictions show grid uncertainties £ 2.2 % for URANS on 50M grid and HRLES on 300M grid, which suggests that these grids are approaching asymptotic range. URANS with anisotropic turbulence model perform better than URANS with isotropic turbulence model. Grid with 3M points are found to be sufficient for resistance predictions, however, grids with up to 10s M points are required for local flow predictions. Adaptive grid refinement is helpful in generating optimal grids; however available grid refinement technique based on the Hessian of pressure, fails to refine the grid further downstream along the hull. HRLES simulations are promising in providing the details of the flow topology. However, they show limitations such as grid induced separation for bluff body KVLCC2 and inability to trigger turbulence for slender body 5415. Implementation of improved delayed DES and/or physics based RANS/LES transition is required to address these limitations. Grid resolution of 300M shows resolved turbulence levels of > 95 % for bluff body, thus such grids seem sufficiently fine for HRLES. The free-surface predictions do not show significant dependence on boundary layer predictions, and accurate prediction for 5415 at Fr = 0.28 is obtained using just 2M grid points. The free-surface reduces pressure gradients on the sonar dome, causing weaker vortical structures than single phase. Flow over 5415 shows three primary vortices, and all of them originate from the sonar dome surface. Onset analysis shows that all the three vortices have open-type separation, and separate from the surface due to cross flow. Further investigation of the cause of differences in KVLCC2 CFD submissions and experimental data suggests that it could be due to differences in the sharpness of the stern.

Shanti Bhushan, Tao Xing, Michel Visonneau, Jeroen Wackers, Ganbo Deng, Frederick Stern, Lars Larsson

Titel: Numerical Ship Hydrodynamics
herausgegeben von: Lars Larsson
Frederick Stern
Michel Visonneau
Verlag: Springer Netherlands
Electronic ISBN: 978-94-007-7189-5
Print ISBN: 978-94-007-7188-8
DOI: https://doi.org/10.1007/978-94-007-7189-5

Premium Partner

Marktübersichten

Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.

Zur Marktübersicht