The Robust Neural Network Control Of The Responding Nonlinear Ship Model

In the ship motion control, there are two important research fields including the modeling of ship motion dynamics and the design of its controller. It is vital to build a mathematical model, which possesses appropriate complexity and required precision, for the research on the performance of the closed-loop control system. The responding nonlinear ship motion mathematical model is constructed on the basis of the state space ship motion mathematical model by the systematic method of input-output information. The model not only deals with the nonlinear infection of the process but also is easy to calculate and has obvious physical meaning, which meets with the requirement of the model precisioaIt is required by the ship controller's design that rudder angle δ overcomes the disturbance of environment in course keeping and changing, and makes the course ψ keep to the set course. At the present there have been many neural network course keeping methods, such as direct control, supervised control and internal model control and so on. Though many effective course keeping methods are provided with the application of neural network control, these methods are still deficient in some aspects. For example, the disturbance of process parameters and environment disturbance can't be made up and the robust performance of the system isn't satisfaction for the feedback being not in the direct control system. But the robust performance is improved obviously when the closed loop gain shaping control algorithm is introduced into the ship motion system on the basis of neural network control in the thesis. The main contents of the thesis are as follows.l.The conception and theory of the Neural Network control and closed loop gain shaping control algorithm are introduced simply.2. The responding nonlinear ship motion mathematical model is constructed on the basis of the state space ship motion mathematical model by the systematic method of input-output information.3. A ship course keeping method of BP Neural Network is brought forward and the controller with preferable control performance and robustness is set up when closed loop gain shaping control algorithm is introduced into the ship motion system in the thesis.Firstly, the turning test of the Norrbin nonlinear ship motion mathematical model is carried out by using VB as the tool of programming, and the datum are obtained. Then the parameters' identification of the responding nonlinear ship motion mathematical model is realized by using Matlab to program on the basis of least square method. All simulating experiments are accomplished through the Matlab's Simulink. The simulating conclusions demonstrate that the method has the advantages of simple design procedure and clear physical meaning. Moreover, the good control performance of the course outputs of the robust Neural Network were satisfaction in the experiment, in which input signals being changed, nonlinear links being joined, control objects being remodeled or internal and external disturbance were involved, and the good control performance of the system had been validated.