Research On Intelligent Control For Large Ship Course

In this thesis, several steering controllers are designed for large ship autopilot system in three parts:ship motion mathematical model, ADRC control technology and CMAC neural network. The related research tasks this thesis has completed are listed as below.Firstly, MMG model theory is studied. In this part, with the XIN SHANGHAI, a large container ship as an example, mathematical models of fluid power and torque acting on the hull, propeller and rudder are established according to MMG theory, and then a brief introduction of calculations of ships affected by wind and flow is presented. To verify the legitimacy and availability of ship model, the simulation of turning test and Zig-zag test in Matlab based on real ship data is made for the design and simulation of the controller below.Secondly, the application of ADRC in large vessels course control is discussed. In view of the shortcomings which have been existed in classical PID control while it’s widely used, active disturbance rejection control theory raises an improved method, including disturbance estimation and dynamic compensation as its core function. In this part, based on different combinations of the two kinds of state error, two active disturbance rejection controllers are designed for the above established ship model, and their effects of course control are verified in three cases of disturbance environment.In the last part, CMAC theory is introduced to research ADRC and CMAC compound control. After introducing artificial neural network learning principle, CMAC neural network is expounded. CMAC is a kind of local approximation neural networks with fast learning, strong generalization ability and simple structure. It can be applied more widely than other type of neural network theory in real-time control field. In this paper, compound control mode, which combines the feedforward and feedback, is used to combine the advantages of CMAC to ADRC for online learning so as to achieve inverse dynamic model for the object to be controlled. From the simulation results, it can be seen that compared with separated ADRC, compound control system may help to effectively reduce system steady state error for more accurate course control.