4.4 Discussion of Results

　

ESSO OSAKA:

　

 The sway force calculations for the ESSO OSAKA shown in Figures 3-4, show that the data from BMT, ITTC and the Mean Values are in good agreement. The Mean Value data shows higher drift forces at larger drift angles due to significant differences in the prediction of the non-linear derivatives. The Abkowitz method appears to underestimate the forces at higher drift angles and yaw rates, presumably due to the absence of higher order derivatives in the SI model. The importance of mathematical model selection in SI methods is further highlighted by the results from Rhee and Kim, who used the MMG model and predict drift forces significantly higher than the ITTC data. Note that this distorts the mean value. The BMT and ITTC predictions for sway force are very close in both the pure drift and pure yaw cases.

　

 Figures 3 and 4 show that the CFD predictions by El Moctar are in good qualitative agreement with the other results but here is still some discrepancy quantitatively, especially at larger drift angles and yaw rates.

　

 By contrast to the sway forces, the yaw moments determined using the Abkowitz derivatives show larger values than other data, although all of the data sets give higher yaw moments than the ITTC benchmark data.

　

 Figures 15-18 in the Appendix demonstrate the wide spread of results for the sway forces and yaw moments, with differences between data sets of up to 100% at higher drift angles and yaw rates. This highlights the difficulty we have in using the multi-source approach and the inherent difficulties of accurately predicting non-linear derivatives.

　

 The comparison of the full-scale turning circle performance with the BMT, ITTC and Mean Value datasets shown in Figure 7, demonstrates that all three 'models' have overestimated the size of the turning circle dimensions. The Mean Value data predicts a larger Advance and Transfer than that predicted by the BMT data, but is smaller than the equivalent ITTC data. However, the steady phase of the turning circle for the Mean Value prediction is larger than both the BMT and ITTC predictions.

　

 The mean value of each derivative can of course be distorted by spurious data, such as a change of sign, and hence care must be afforded to the selection of data sets that are used in the Mean Value calculation.

　

 A comparison of the Zig-Zag manoeuvre shows that the Mean Value data underestimates the overshoot angles and the ITTC overestimates. The BMT predictions show very good agreement in overshoot angles and response to helm with the full-scale data. All three models predict a quicker response to counter- helm than the full-scale vessel and hence a cumulative difference in the response time can be seen to build through the simulation. It is worth noting that this error is less in the ITTC predictions at the expense of the agreement in overshoot angles.

　

GOLDEN PRINCESS:

　

 For the GOLDEN PRINCESS, it was not possible to calculate the hydrodynamic forces and moments from the full range of prediction methods for reasons of applicability or insufficient information on the testing methods and regression analysis used.

　

 The comparison of forces in Figures 19 and 20 shows that the BMT and Mean Value data are in close agreement, with the exception of the Yγ force at higher yaw rates. Once again, this may be attributed to the prediction of the non-linear terms and the potential distortion of the mean value due to spurious data.

　

 The two datasets provide very similar turning circle predictions as shown in Figure 9, with the Mean Value data providing a slightly smaller Advance and Transfer in the transient stages, but a larger Tactical Diameter. The steady turning diameter for the Mean Value prediction is smaller than that predicted by the BMT method and this has the consequence of a greater speed loss (70% versus 63%) during the turn.

　

 The comparison of Zig-Zag results for the GOLDEN PRINCESS exhibits similar results to those for the ESSO OSAKA, in that using the Mean Value data results in an increase in response time to helm and counter-helm, although the overshoot angles are almost identical. No comparison with full-scale trials has been made in this case, as they were unfortunately subject to some significant environmental conditions and were not felt to sufficiently reliable for our present purposes.

　

 For a true assessment of manoeuvring performance and the accuracy of simulations, it has long been the author's opinion that one must consider realistic manoeuvres if our industry is to serve those who man the vessels that we model well. Indeed, this has become an interesting and challenging topic and much work is taking place on this ([27]).

　

 We have therefore compared our predictions with a slow speed turn (variable rudder angle and engine orders) recorded during a departure from Venice, Italy in May 2002. The authors attended the vessel and were able to record the full-scale data using the ship's Integrated Bridge System. We have simulated this manoeuvre using both the BMT derivatives and those predicted by the Mean Value approach. Figures 11 and 12 show the track-plots for this manoeuvre, with the simulated vessel shown as the ship's outline and the full-scale data as the heavy black line.

　

 It can be seen that the correlation for this relatively simple manoeuvre is very good with the maximum cross-track error being less than 5% of the 'Advance' of the vessel. The use of the Mean Value derivatives appears to further improve the prediction for the completed turn.

　

 Figures 13 and 14 show the speed and heading of the vessel during the manoeuvre for the two mathematical models and full-scale vessel. In this case, the agreement with the full-scale vessel is better when using the BMT derivatives, although the correlation is very good in both cases.

　

 This is an important result with regard to the validation of ship specific mathematical models for simulation purposes as pilots and ship's masters will in our experience, judge the simulator model on known quantities to them, such as ship speed and heading for a given rudder command and scenario. It is therefore very important that we are able to reflect these parameters accurately. Positional accuracy is far harder for the human eye to determine and a qualitative approach is often used. We have no doubt all heard comments such as;

　

 "For 10 degrees rudder, I would be in the middle of the channel with this ship"

　

 As described earlier, the sway forces and yaw moments determined from the two datasets are very similar with some difference appearing at higher drift angles and yaw rates. Based on the comparison of the slow speed manoeuvre, it would appear that it is the non-linear derivatives that are key to the prediction of the hydrodynamic forces and moments. The linear components are is very close agreement and hence the traditional method of assessing manoeuvrability by examining course stability is perhaps questionable in realistic manoeuvring scenarios, where the non-linear hull forces dominate.

　

5.1 Ship Modelling Requirements

　

 The recent advent of the Standards of Training, Certification and Watchkeeping for Seafarers, 1995 (STCW'95) has led to a requirement for Masters and Senior Officers to undertake ship-handling training for their particular vessel. This is further compounded by the development of more manoeuvrable ships, different control devices (e.g. podded propulsors) that affect how the ship should be handled and the growing requirement for vessels to use new ports that are becoming smaller and more remote.

　

 It is now very important that we are able to supply the users with accurate ship models of their own vessels to allow the STCW training to be performed and to provide better use of simulators for manoeuvre rehearsal and port design applications.

　

5.2 The Future of Ship Modelling?

　

 BMT is currently leading a major EU funded research project called SEA-AHED with three aims:

　

　

 The navigational predictor will record the full-scale manoeuvring data for the vessel in order to compare its predictions with the actual values achieved and thus validate its own predictions over time. The by-product of this is access to full-scale data for every ship in which the system is used with no incremental cost for running specific sets of trials. This data can then be used for the validation of the simulator models and ultimately for the advancement of our prediction methods for hydrodynamic forces and moments.

　

 This study forms the beginning of a long term project that will result in new prediction methods for the hydrodynamic derivatives being derived. This will include a re-appraisal of the currently available methods to derive new regression formulae based on the successful validation of a wide range of ship models. It is hoped that the full scale data will be used to undertake an analysis of full-scale hydrodynamic forces and moments for the eventual prediction of the linear and importantly, the non-linear hydrodynamic derivatives.

　

 The review of the available prediction methods for hydrodynamic derivatives has demonstrated that predictions for the linear derivatives are more readily available and are generally more accurate than those for the non-linear components.

　

 When predicting the derivatives for a specific ship, it may be possible to use a combination of methods, provided sufficient attention is given to the quality and format of the data.

　

 The prediction of hydrodynamic forces and moments shows a significant spread in the results and is highly dependant on the quality of the input data.

　

 The results of the simulated turning circles and zig-zag manoeuvres show that results may be improved with the multi-source approach in certain circumstances. The authors believe that the results are promising in this regard, but would express caution in applying this approach until fully investigated.

　

 The simulation results for the GOLDEN PRINCESS slow speed turn again demonstrate the potential advantages of the multi-source approach, with the proviso that one must consider the relative importance of each parameter (cross-track error, speed, heading etc) to be validated before implementing.

　

 The prediction of the non-linear components of the hydrodynamic forces acting on the hull is the most important factor in achieving accurate ship models for realistic, slow speed manoeuvring applications.

　

 The present paper is essentially a summary of a larger body of work at BMT but demonstrates that simulation results may be improved through the careful selection of hydrodynamic derivatives from a range of sources, providing one has sufficient confidence in the quality of the data and its derivation. It is our opinion that this approach be continued in order to re-assess the formulae for the hydrodynamic derivatives and the eventual derivation of new regression analyses based on full-scale performance.

　

[1] Proceeding of the 23^rd ITTC-Vol II, Specialist Committee on Esso Osaka, Final Report.

[2] Barr, R. A. 'A Review and Comparison of Ship Manoeuvring Simulation Methods', Trans. SNAME, v.101, 1993.

[3] El Moctar, O.& Bertram, V. "Selected Topics of CFD for Ship Manoeuvring", 36^th WEGEMT Summer School, Manoeuvring and Manoeuvring Devices, May 2001.

[4] Tanaka, S. "On Hydrodynamic Forces acting on Ship with Large Drift Angles", Proc. MAN 98, 1998.

[5] Abkowitz, M, "Measurement of Hydrodynamic Characteristics from Ship Manoeuvring Trials by System Identification", Trans, SNAME, vol 88, 1980.

[6] Rhee, K.P and Kim, K. "A New Sea trial Method for Estimating Hydrodynamic Derivatives", Proc SOTECH, vol 3, no. 2, 2001.

[7] Burnay, S. "An Investigation into Different Types of Manoeuvring Models and the Prediction of Hydrodynamic Derivatives", BMT SeaTech Report No. A899.01, 2000.

[8] Crane, C.L "Manoeuvring Trials of a 278,000DWT Tanker in Shallow and Deep Waters", Trans. SNAME, vol 87, 1979.

[9] Hydronautics, "Model Test and Simulation Correlation Study Based on the Full-Scale ESSO OSAKA Manoeuvring Data", Hydronautics Report No. 8007-1, 1980.

[10] Dand, I. and Hood, D. "Manoeuvring Experiments using Two Geosims of the Esso Osaka", NMI Report No. R163, 1983.

[11] Clarke, D et al, "The Application of Manoeuvring Criteria in Hull Design Using Linear Theory", Trans. RINA, v.124, 1982.

[12] Inoue, S et al "Hydrodynamic Derivatives on Ship Manoeuvring", Int'l Shipbuilding Progress, v.28, No.321, 1982.

[13] Oltmann, P. & Sharma, S:'Simulation of Combined Engine and Rudder Manoeuvres Using an Improved model of Hull-Propeller-Rudder Interactions', 15^th ONR Symposium on Naval Hydrodynamics, Hamburg, September 1984.

[14] Fujino, M, "Keynote Lecture: Prediction of Ship Manoeuvrability: State of the Art", Proc. MARSIM '96, 1996.

[15] Kijima, K. et al "On the Manoeuvring Performance of a Ship with the Parameter of Loading Condition", Trans. Japanese Soc. of Naval Architects, November 1990.

[16] Ishiguro T, et al "A study on the Accuracy of the Recent Prediction technique of Ship's Manoeuvrability at Early Design Stage", Proc MARSIM '96, 1996.

[17] Ankudinov, V. "Design Workbook on Ship Manoeuvrability", Chapter 6, SNAME T&R Bulletin 1-44.

[18] Khattab, O. "Multiple Regression Analysis of the Hydrodynamic Derivatives for Manoeuvring Equations", BSRA Report No. W1090, 1984.

[19] Dand, I. "The Revised Manoeuvring Prediction Program (MPP)', BMT SeaTech Ltd Report for the CRS-MAN Working Group, May 1994.

[20] MARIN, "Computer Program MPP, Theory and User Manual", CRS-MAN Report, 1995

[2l] Dand, I. "On Modular Manoeuvring Models" Proc. Int'l Conf. On Ship Manoeuvrability; Prediction and Achievement, RINA, London, April/May 1987.

[22] Khattab, O. :'Ship Handling in Harbours Using Real Time Simulation', Int'l Conf. On Ship Manoeuvrability; Prediction and Achievement, London, April/May 1987.

[23] Hooft, J. P. & Nienhuis, U. : 'The Prediction of the Ships Manoeuvraility in the Design Stage', Trans. SNAME, v.102, 1994

[24] Dand, I. "An Enhanced Ship Manoeuvring Simulation Model" NMI Report No. 251050, 1984.

[25] Clarke, D. & Horn, J. R. : 'Estimation of Hydrodynamic Derivatives', Proc. 11^th Ship Control Systems Symposium, v.2, University of Southampton, April 1997.

[26] Burnay, S. "Further Investigations into the Prediction of Hydrodynamic Derivatives for Ship Manoeuvring Simulation", BMT SeaTech Report No. 5004_2.1_1, 2003.

[27] Hwang, W-Y et al, "An Exploratory Study to Characterize Ship Manoeuvring Performance at Slow Speed", To be pub.., MARSIM 2003, 2003