Numerical Ship Hydrodynamics

The Tokyo 2015 Workshop on CFD in Hydrodynamics was the seventh 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 2015 Workshop offered 16 test cases for three ship hulls. A total of 36 participating groups of CFD specialists submitted their computed results during the fall of 2015. The results were compiled by the organizers and discussed at a meeting in Tokyo in December 2015. In this chapter the background and development of the Workshops since the start are presented. The three hulls used in the 2015 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 from each chapter and recommendations for future Workshops are presented. The detailed results of the computations are discussed in subsequent Chapters.

In this Chapter, measured data of resistance, sinkage, trim, self-propulsion factors, longitudinal wave cut and detailed flow are summarized for Japan Bulk Carrier (JBC) with and without an energy saving circular duct. JBC is a newly designed ship for CFD validation of a ship with an energy saving device (ESD). Resistance and self-propulsion tests are conducted in towing tanks of National Maritime Research Institute (NMRI) and Osaka University (OU) using model ships of different sizes. Detailed local flow data are acquired using SPIV (Stereo Particle Image Velocimetry) for several cross sections at towing tanks of NMRI and OU. The local flow data are also measured by SPIV in a wind tunnel at Hamburg University of Technology (TUHH).

This chapter is dedicated to the experimental results of resistance, sinkage, trim, and self-propulsion tests of the KCS hull form. The KCS was conceived to provide data for both explication of flow physics and CFD validation for a modern container ship with a bulbous bow. The Korea Research Institute of Ships and Ocean engineering (KRISO) performed towing tank experiments to obtain resistance, mean flow data and free surface waves. Self-propulsion tests were carried out at the National Maritime Research Institute (NMRI) and later, additional resistance tests were also reported in Gothenburg 2010 Workshop on CFD in ship hydrodynamics. The main information of the KCS hull and its propeller is introduced in Sect. 1.1 and then the results of resistance and hull attitudes measurement are followed by self-propulsion results and some local flow measurements.

Evaluation is performed of the data used for T2015 test cases for the KCS captive added resistance σ_AR and ONRT free running course keeping/speed loss in head and oblique waves. For KCS calm water resistance, the individual facility N-order level testing uncertainty is U_Xi = 1%D and the multiple facility standard deviation based on three institutes using 2 model sizes (L = 7.3 and 6.1 m) is SD = 0.74%D, such that the individual facility MxN-order level testing uncertainty is U_Di = 1.75%D. U_Xi was not reported for sinkage and trim; however, the SD = 4 and 7%D, respectively. For KCS head waves, the analysis was based on three institutes with model sizes L = 6.1, 3.2 and 2.7 m. For the 6.1 m model, FORCE provided U_Xi = 8, 4 and 4%D, respectively, for σ_AR and first harmonic heave z₁/ζ₁ and pitch θ₁/ζ₁k amplitudes, whereas for the 2.7 m model, FORCE/IIHR provided U_Xi = 7/18, 8/2 and 9/5%D, respectively. In consideration of the differences in model sizes and rigid vs. surge free mounts the agreement between the facilities is reasonable: for the primary variables, SD = 3, 20 and 10%D for z₁/ζ₁ and θ₁/ζ₁k amplitudes and σ_AR, respectively; and for secondary variable SD = 66%D for first harmonic resistance C_T1. For KCS oblique waves, the data is only available from IIHR for tests in 2015 and 2016. The agreement is good, i.e., SD values are comparable to the corresponding values for head waves: for primary variables, SD = 6 and 63%D for z₁/ζ₁ and θ₁/ζ₁k amplitudes and σ_AR, respectively; and for secondary variable SD = 13%D for C_T1. For ONRT self-propulsion and head and oblique waves, the data is only available from IIHR and only preliminary uncertainty analysis is available. The evaluation showed reasonably reliable data for both KCS and ONRT. However, clearly it is desirable to have data from more facilities including uncertainty analysis for assessment of facility biases, which will provide more robust data and uncertainty analysis for CFD validation.

JBC predictions are presented for four cases: towed and self-propulsion for the bare hull and a hull with an Energy Saving Device (ESD). A statistical evaluation is made based on the 88 resistance predictions submitted. The comparison error is defined as the difference between the measured data and the numerically predicted value. This error is analyzed in different ways. It is seen that the mean signed error is as small as the measurement accuracy, while the mean absolute error is about twice as large. This represents the typical error in the prediction. A similar accuracy was found in the 2010 Workshop. The number of grid cells has increased since 2010 and the required grid size for a given uncertainty of 4% has increased from 3 M cells to 10 M cells. Reasons for this are discussed. As in the previous workshop the two-equation turbulence models produce more accurate results than the more advanced models, although the more and more popular Explicit Algebraic Stress Model (EASM) is close. A surprising result of the analysis is that methods with wall-functions give significantly smaller resistance errors than those with a wall-resolved flow. Grid convergence is discussed and it is shown that the vast majority of results converge with grid refinement, but that the achieved order of accuracy is often far from the theoretical one. Sinkage is much better predicted than in 2010 and the trim results are also improved. Finally, the wave pattern prediction is discussed. The best methods in 2010 predicted the waves extremely well, in fact better than in the present workshop. A consistent shift in the predicted wave phase relative to the data could indicate a measurement error.

This chapter is devoted to an analysis of the local flow field for test cases 1-3, 1-4, 1-7 and 1-8. The general objective of this analysis is to assess the present computational approaches according to their ability to represent the influence of a duct on the local flow field in presence or absence of a propeller, thanks to the existence of local flow measurements mainly performed by NMRI. General conclusions based on the numerous contributions, are drawn concerning the reliability of turbulence closures in these peculiar flow configurations, the influence of the propeller representation on the local flow field and the ability of CFD to represent the local influence of a duct on the flow at the stern region and in the wake of a ship.

This chapter discusses the results of resistance predictions including trim and sinkage (the workshop case 2.1) and the results of self-propulsion predictions (the workshop case 2.5) for KCS. The resistance predictions for 6 different Froude numbers in trim and sinkage free condition were requested. The comparison error at the design speed (Fr = 0.26) is −0.2% and the standard deviation is 1.5% of the data value. The mean comparison error for all 6 speeds is 0.43% and the mean standard deviation is 2.48%. The increased error and standard deviation is caused by the results for low Fr simulation submissions. The submitted trim and sinkage also shows larger comparison error in low Fr region (Fr < 0.2). Self-propulsion results are reported with both towing force (FD) fixed and propeller revolution (rps) fixed. The participants using a body force propeller model based on potential theory selected FD fixed condition for self-propulsion and the participants who adopted actual propeller rotating simulation preferred rps fixed condition. The mean comparison errors of KT and KQ are 0.5 and −3.5% respectively. The standard deviations are 2.7 and 2.4% respectively. Self-propulsion parameters are slightly better predicted by body force models (FD fixed). However, local flow characteristics are better predicted by actual propeller rotating simulation (rps fixed).

CFD is assessed for added resistance for KCS (captive test cases 2.10 and 2.11) and course keeping/speed loss for ONRT (free running test cases 3.9/3.12/3.13) in head and oblique waves. The number of submissions were 10, 2, and 8 for test cases 2.10, 2.11, and 3.9/3.12/3.13, respectively. The assessment approach uses both solution and N-version validation. The former considers whether the absolute error \(\left| {E_{i} } \right| = \left| {D - S_{i} } \right|\) is less, equal or greater than the validation uncertainty, which is the root sum square of the numerical and experimental uncertainties, i.e., \(|E_{i} | \le U_{{V_{i} }} = \sqrt {U_{{SN_{i} }}^{2} + U_{D}^{2} }\). The latter considers whether the absolute error is less, equal or greater than the state-of-the-art SoA_i uncertainty, i.e., \(\left| {E_{i} } \right| \le U_{{SoA_{i} }} = \sqrt {U_{{V_{i} }}^{2} + P_{{\left| {E_{i} } \right|}}^{2} }\) where \(P_{{\left| {E_{i} } \right|}} = k\sigma_{\left| E \right|}\) is the uncertainty due to the scatter in the solution absolute error. Errors and uncertainties are normalized using both the data value D and its dynamic range DR. The captive resistance C_T and free running self-propulsion propeller revolutions RPS \(\overline{\left| E \right|} < 2{\% }{\text{D}}\) with U_D less than but comparable \(P_{{\left| {E_{i} } \right|}} < 3{\% }{\text{D}}\) such that 3/1 solutions were validated but 8/4 codes/solutions were N-version validated for C_T/RPS. The head waves captive and free running heave and pitch \(\overline{\left| E \right|}\) is less than 8%DR with U_D less than 5%DR and \(P_{{\left| {E_{i} } \right|}}\) less than 13%DR such that about 5 for captive and 2 for free running solutions were validated and about 7 for captive and 5 for free running codes/solutions were N-version validated. The errors for added resistance and speed loss were less than 13%DR with U_D less than 7%DR and \(P_{{\left| {E_{i} } \right|}} = 25\) and 13%DR such that about 4 for captive and 1 for free running solutions were validated and 7 for captive 4 for free running codes/solutions were N-version validated. For captive head waves, the errors and scatter are smaller than those for potential flow. The oblique waves captive and free running motion errors \(\overline{\left| E \right|}\) are less than 10%DR except for roll with U_D large 23%DR for captive pitch and other wise small <2%DR. The errors for added resistance and speed loss were <8%DR with U_D < 9%DR. The largest errors were for roll, which had errors for captive of 12%DR and for free running of 20%DR. None of the KCS and ONRT test cases were able to achieve a programmatic requirement of 5%D and 5%DR for calm water and waves, respectively. Note that not all the experimental uncertainties are able to meet this requirement either such that both experiments and CFD need to reduce their uncertainties and in addition CFD its errors. Nonetheless, in view the comparable CFD capability for ONRT free running vs. KCS captive conditions, the prognosis for CFD capability is excellent. Future CFD assessment should extend the current test cases for consideration of added power in regular and irregular waves. Verification should be required, and more limited/most important validation variables should be confirmed and used for the V&V and SoA assessment.