| Ship manoeuvrability is the hot topics and common issues for shipping,shipbuilders,societies and International Maritime Organization.Model is a key element for ship manoeuvrability research.For the nonlinear ship motion and extensive distributed data,grey box identification modeling methods based on free running test are proposed for an effective and practical model.Based on the hypotheses of the deep water and the independence between ship manoeuvrability and seakeeping,ship manoeuvring based on grey box models and algorithms are developed for prediction and simulation of real sea ship manoeuvring.Ship manoeuvring identification modeling is a kinds of vital modeling method.Models for navigation are different from marine engineering for restriction of test scenario.For the autonomous navigation and collision avoidance,the grey box identification model based on reference models and the improved algorithms are expected to be established for ship manoeuvring on the calm water and the real sea.The models and algorithms are validation by the free running model test(FRMT)and the full-scale sea trials.The main research works and contributions of this dissertation are as shown follows:1.The analysis of the rudder force,heeling and data distribution for ship reversed rudder manoeuvring.The analytic expression of rudder force for the reversed rudder is deduced and is validated by simulations methods,and the ship heeling and the main engine loading is analyzed.Additionally,the probability density of zigzag test and turning circle test data are compared.The results show that(1)the analytic expression is a kind of accurate hyperbolic function.(2)And the reversed rudder causes rudder force adding 100%,the 400,000 DWT very large ore carrier adding 240%.Also the reversed rudder aggravates the ship heeling and fluctuates the main engine loading.(3)The probability density of test data is calculated and distribution of zigzag test is more extensive.These contributes a conflicting between data distribution and manoeuver safety,and the standard manoeuvres is suitable for identification.2.Grey box identification modeling based on reference model(RM)for FRMT.With calculating weights of main particulars by partial least squares,the vector of ship particulars is optimized.Then the suitable RM is selected via correlation values.Along with the similarity rules to diminish the particular bias between RM and target ship,the RM modeling is proposed.The random forest(RF)and the bayesian neural networks(BRN)compensate the acceleration errors between the RM and the target ship respectively,and the RM identification modeling is proposed.The KVLCC2 free running model verifies the methods.The results show that(1)the+35°turning circle ratios of advance,tactical diameter,time to change heading 90°and time to change heading 180°between RM and FRMT are 1.028,0.967,1.051,1.052.(2)The RM-RF can predict zigzag test precisely,but not turning circle.To solve the above problem,the inputs and outputs of RF is optimized by manoeuvring principle and RM-IRF is achieved.The ratios between RM-IRF and FRMT have been improved to 0.991,0.957,1.014,1.013 respectively.The validations demonstrate the effectiveness of the grey box identification modeling and the generalization ability improved by optimized inputs and outputs.3.The hyperparameters tuning of grey box identification modeling and its application to white box identification.The support vector machine(SVM)is utilized as identification algorithm.The precise of RM-SVM is analyzed by traversing box constraint,kernel scale and insensitive band values.For the discontinuous,discrete and non derivable objective function of RM-SVM,correlation values 1 norm based objective function is solved by pattern search(PS)and the RM-PSM-SVM is achieved.The whole ship hydrodynamic coefficients is identified based on RM-PSM-SVM prediction,and grey box model is converted to white box model.The results show that(1)The objective function reachs 0.99 and approachs max value while the iterations is 20.(2)The+35°turning circle ratios of advance,tactical diameter,time to change heading 90°and time to change heading 180°between RM-PSM-SVM and FRMT are 1.004,0.983,1.003,0.985.(3)For whole ship model precise compared with captive model test,the sway linear coefficient reaches 50%the turning linear coefficient reaches 60%,the rudder angle 3 order coefficient reaches 70%.4.Full-scale sea trials correction and manoeuvring modeling based on grey box.Firstly the measurements and precise of instruments are verified for selecting the useful data source.Secondly,based on the hypothesis of steady turning and uniform flow field,the drift forces and distances of wind and wave are estimated preliminarily.Finally,the coefficients for wind wave and current are optimized.The wind wave and current disturbs are corrected,and the correction method for full-scale sea trials is proposed.The results show that(1)in this sea trial the current is the primary part,the wind takes second place,the wave is the minimum.The correction method is effective.(2)The corrections method in other literatures are special example of the proposed method.(3)The correlation value for+20° turning circle reach 0.85.The trajectory prediction errors for-35°turning circle is less than 50m.The corrrection method and grey box model is validated in the study case.The programming in this dissertation are executed on Matlab/Visual C++.The established grey box model validates the road map from model test to full-scale sea trial.The research enhances navigation safety guarantee. |
PDF Full Text Request