| Summary: | This thesis concerns how the dynamic positioning (DP) control system can better handle transient events, where the loads experienced by the vessel change significantly over a short time frame. The material is intended for DP systems controlling surface ships, but it is also relevant for other DP vessels, as well as other motion control applications for marine vessels. The thesis is a collection of papers, with some introductory chapters to set the context of the problem. The thesis contains contributions on the fundamental level, where different load models and observer algorithms are fairly compared and analyzed. The study investigates the effect of including nonlinear damping in the model. The results show that when using models where the residual loads are modeled as a current, then nonlinear damping improves performance. For the models where the residual loads are modeled as a superimposed load vector, then the effect of nonlinear damping is less apparent. The different observer algorithms show surprisingly similar performance, which indicate that DP is dominantly a linear process. Two different augmentations of existing observer design have been proposed for better transient performance, while maintaining good steady-state performance. A time-varying model-based observer is presented and analyzed, where aggressive gains are used during transient for responsiveness, and relaxed gains are used in steady state for lower oscillations in the state estimates. The performance is verified through high-fidelity closed-loop simulations and on experimental full-scale data from a cruise with the research vessel R/V Gunnerus. In addition, on the cruise with R/V Gunnerus, a partial closed-loop validation with integrated DP observer and controller was performed. The other design is a hybrid observer combining model-based and kinematic observers. The hybrid observer switches to the kinematic observer in transient conditions, and to the model-based observer in steady conditions. The observer performance is verified through model-scale closed-loop experiments, and on full-scale experimental data from R/V Gunnerus. For the control design, integral action is compared to other ways of compensating the environmental and unmodeled loads, with special focus on transient conditions. The results show that the best solution is to use the estimate of the environmental and unmodeled loads from an observer with tuning optimized to estimate these loads. This method outperforms the other methods in transients, and has equal performance to integral action in steady state. In addition, using an observer to find the estimate alleviates anti windup issues, and contrary to tuning integral action, an observer can be tuned open loop – which is a large benefit. Hybrid integral action is proposed to improve performance in transient conditions and still keep relaxed and satisfactory performance in steady conditions. This is achieved by high gains in the integrator in transients to better compensate the loads in transients, and relaxed gains in steady conditions to not induce unnecessary oscillations. Pseudo-derivative control (PDF) is proposed as an alternative to traditional proportionalintegral- derivative (PID) control. The PDF control algorithm does not need a reference filter as the PID does, as the references are generated internally, and the PDF control algorithms is better at mitigating integral windup, compared to the PID control algorithm. Performance of the PDF control law is shown through a simulation study, and through full-scale closed-loop trials with R/V Gunnerus.
|
|---|