Comparison study of linear and nonlinear controllers and observers for Dynamic Positioning systems
MetadataVis full innførsel
- Institutt for marin teknikk 
A dynamically positioned ship is continuously exposed to environmental disturbances like wind, waves and current. To maintain the desired position and heading or to perform lowspeed maneuvering of a ship, by using its own propellers and thrusters it is necessary to use a dynamic positioning (DP) system. A DP system needs an observer to produce lowfrequency estimates of the vessel motion. The observer includes a model of the waves, and is therefore able to lter out the wave-frequency motion. This is known as wave ltering. Position reference sensors, combined with wind sensors, motion sensors and gyro compass, provide information to the computer pertaining to the vessel's position and the magnitude and direction of environmental forces aecting its position. This thesis deals with a commonly used DP system for a vessel, and compares two types of controllers in combination with a nonlinear passive observer. In order to demonstrate the performance of this system, a simulation system in MATLAB/Simulink and a model experiment in the Marine Cybernetics Laboratory (MCLab) are developed and tested. The two types of controllers considered in this thesis are a nonlinear PID controller type and a multivariable PID controller type. The nonlinear PID controller has shown good performance for changing the vessels setpoints both in the simulations and the model tests. However, the yaw angle should have been in the range of ±3 [deg] which is not the case for this controller. This error is considered when developing the multivariable PID controller, which shows a better performance than the nonlinear PID controller. For the simulations with station keeping control, the maximum position deviation is 3.5 [cm] in surge, 3.25 [cm] in sway and 5.5 [deg] in yaw. For the model tests with Cybership III the maximum position deviation is at its initial position is 40 [cm] in surge, 7 [cm] in sway and 5 [deg] in yaw. Comparisons between the model tests and the simulations show that the same controller gains are working well in both settings. Two of the observer gains are tuned down for the model test, which provides accurate separation of the LF and WF induced vessel motions. For the model tests, the bias model gave a drift in the separation of the WF and LF vessel motions. This is probably due to the bias integrator. The bias time constants can be tuned in order to get better results.