On The Modeling Of Ship Manoeuvring Motion In 4 Degrees Of Freedom Based On Support Vector Machines

Sea transportation by ships has a distinctive preponderance in ratio of freight volume to energy consumption. The sea shipping industry has been playing an irreplaceable role in the national economy of China. In recent years, the International Maritime Organization has launched a series of new standards and new critiaria, which put forward a much higher requirement on the energy efficiency of ships. Highly intelligent green ships with better navigating ability will surely be the development direction of the sea transfortation industry.For warships and other ships with low metacentric height(e.g., container ship), ship manoeuvring motion is usually accompanied by roll motion of remarkable amplitude. The roll motion not only affects ship navigation safety and the tactical and technical performance of warships, but also has directly impacts on ship manoeuvrability. Besides, roll motion must be considered in course keeping, designing the autopilot and/or rudder roll stabilization system for a sea-going ship. Therefore, investigation of ship manoeuvring motion with the influences of the roll motion being taken into account is urgently needed.In this thesis, system identification method based on support vector machines for modeling of ship manoeuvring motion in 4 degrees of freedom is investigated. Firstly, three modeling methods, i.e., white-box modeling, grey-box modeling and black-box modeling are applied respectively for modeling of ship hydrodynamic models(including whole-ship model and modular model) in 4 degrees of freedom. Secondly, based on the linearized equations of ship manoeuvring motion in 4 degrees of freedom, the coupled response models for ship steering and roll motion are derived. Then, Sensitivity analysis of the hydrodynamic coefficients in the hydrodynamic models is implemented. The mathematical models are simplified by omitting the coefficients of smaller sensitivity according to the results of sensitivity analysis. Finally, the fruit fly optimization algorithm is used to optimize the parameters of support vector machines, which is applied in ship manoeuvring motion prediction.In white-box modeling, identification formulas are reconstructed, so that the hydrodynamic coefficients can be identified directly, and the limitation that the number of the coupled hydrodynamic coefficient terms must be equal is overcome. In grey-box modeling, the hydrodynamic coefficients are not needed to identify. The high-order vectors of motion state variables are taken as input data according to the mathematical model of manoeuvring motion. The prediction model of the support vector machines is established and used to predict the ship manoeuvring motion. In black-box modeling, independent of the the mathematical model of manoeuvring motion, the state variables at last time step are taken as the input data to establish the prediction model of the support vector machines, which is then used to predict the manoeuvring motion at next time step. The zigzag test and turning circle manoeuvre are predicted to demonstrate the effectiveness and the good generalization performance of these three modeling methods. Finally, the proposed modeling methods are analyzed and compared with each other in aspects of prediction accuracy, computation speed and known conditions required. The appropriate modeling method can be chosen according to the intended application of the mathematical models and the available data needed for system identification.Based on the linearized equations of ship manoeuvring motion in 4 degrees of freedom, the coupled response models for ship steering and roll motion are derived. The response relations between the yaw motion, sway motion and roll motion to the rudder angle are obtained, and the complete expresions of the manoeuvrability indexes are presented. By using the manoeuvrability indexes calculated from the RPMM test data of a container ship, the coupled response models are used to predict the yaw motion, sway motion and roll motion in zigzag tests. The validity of the coupled response models is demonstrated by comparing the prediction results with those predicted by the hydrodynamic models. At last, based on the free-running model data of the DTMB5415 model, ε-support vector machines and least squares support vector machines are used to identify the manoeuvrability indexes in the coupled response models, respectively; and zigzag tests are predicted by using the identified manoeuvrability indexes. Good predictive ability and generalization performance of the coupled response models are demonstrated by comparing the prediction results with the experimental data.Sensitivity analysis using the direct method and indirect method is implemented for the hydrodynamic coefficients in the hydrodynamic models. The mathematical models are simplified according to the results of sensitivity analysis, and the number of the hydrodynamic coefficients in whole-ship model is reduced from 102 to 61, while the number of the hydrodynamic coefficients in modular model is reduced from 41 to 34. Then by analyzing the original simulation data, the hydrodynamic coefficients in the simplified models are identified by white-box modeling method using support vector machines, and the manoeuvring motions are predicted by using the identified simplified models. The predicted results are compared with the simulation data, which shows the validity of the sensitivity analysis based on the direct method and indirect method, the reasonability of the simplified model based on the results of the sensitivity analysis and the effectiveness of the modeling method using support vector machines. The simplified hydrodynamic models can be used to guide the design of physical model tests and is convenient to analyze and apply.The influences of different structure parameters and kernel parameters of support vector machines on the accuracy of predicted manoeuvring motion are compared. In order to reduce the difficulty in choosing the structure parameters and kernel parameters of support vector machines, the fruit fly optimization algorithm, a new kind of swarm intelligent algorithm, is applied to optimize the parameters to establish the black-box model for predicting the ship manoeuvring motion in 4 degrees of freedom. The comparison of the predicted results with those obtained by the parameters from the cut and try method shows that it is more accurate by using the optimized parameters. The comparison of different methods for parameter optimization shows that compared to the particle swarm optimization and the grid search method, the proposed algorithm has the advantages of more simple to construct, less parameters to adjust, not prone to falling into local minima and more easy to obtain the global optimal solution.Through the present study, the validity of the modeling method for ship manoeuvring motion in 4 degrees of freedom by using support vector machines has been demonstated. It provides a comprehensive technical support for modeling of ship manoeuvring motion in 4 degrees of freedom by applying system identification method. According to the results of sensitivity analysis, the simplified hydrodynamic models are obtained, which are more convenient to analyze and apply. Based on the linearized equations of ship manoeuvring motion in 4 degrees of freedom, the coupled response models are derived, which provide conveniences in study of course-keeping under effect of roll motion, design of roll stabilization by rudder and autopilot for a sea-going ship.