The Research Of The Neural Network Control For Ship Steering Autopilot Under High-wave-level Environment

Under high-level-wave environment, it is impossible to test the operating conditions of the ship steering autopilot system with real ship experiment, and the effect of the ship's directional control either. This paper emphasizes the ship steering autopilot system's nonlinear characteristics such as backlash, saturation, dead area and so on. It re-creates the operating conditions of existing PID control system under high-level-wave environment with the help of simulation. The result shows that the existing PID control system was unsatisfied to the ship steering autopilot system. According to the nonlinear characteristics of the ship steering autopilot system under high-level-wave environment, it improves the effect of the ship's directional control under high-level-wave environment with advanced neural network controller.Firstly, the model of the ship steering autopilot system is founded. Ship dynamic equations of yawing, rolling and swaying are founded based on momentum theory and momentum torque theory. The force and moment on the ship are studied, including the force of water, on the rudder, and the disturbance of wave. The calculating model of these forces and moment are given.Secondly, the nonlinear model of the rudder hydraulic servo system is founded. The basic composition structure of the rudder hydraulic servo system is analyzed, and the models of every part are founded. The system's existing nonlinear factors are analyzed under practical operating conditions.Then, the simulations of existing ship steering system is carried out. The control system simulations in different sea conditions and encounter angles are carried out. The curves and statistic values of yawing angle, rudder angle and flap-rudder angle are given. Digital simulation results show that the rudder angle which under the control of the existing controller constantly reaches saturation. In essence, the system is in the uncontrolled (open-loop) status in this process. Thus, the ship steering control accuracy requirements are unsatisfied. At last, the Neural Network Control for Ship Steering Autopilot is designed and simulated. According to the nonlinear characteristics of high-level-wave environment, the neural network controller is designed. The process includes determining neural network controller structure, and setting the initial value, designing neural network training algorithms. The design direction of the control system simulation curve is presented, and compared with the existing control system, the simulation results show that the neural network controller in the direction control accuracy and resolution rudder angle on the issue of saturation is obviously improved above design conditions on the sea.