Modelling, identification and control of ships equipped with azipod propulsion systems

Abstract

The area of applying control engineering in a hydrodynamic context is increasing in importance by ongoing adoption in the industrial sector, for example in autonomous ship applications and companies have recently started marketing solutions to private consumers. In this thesis, a workflow for designing and calibrating a model-based dynamic ship control system is proposed and analysed. The workflow consists of a process to parameterise mathematical models to resemble a given ship's nonlinear dynamic behaviour using system identification, followed by using the parameterised models as a basis for a 2-degrees-of-freedom (2-DoF) control system. In the first major part of this thesis the system identification and model parameterisation are analysed. First, a suitable mathematical formulation for nonlinear ship behaviour was obtained from academic literature. A two-pod azimuthing podded propulsion drive configuration was assumed for its versatility but the approach is extensible to other types of propulsion. It was investigated what types of experiments need to be performed in the context of system identification to estimate sets of model parameters that resemble a given ship's dynamics with enough precision to be used in the controller design, as well as how to increase that precision further by supplying the parameter set found using a least-squares approach as an initial point for gradient-descent based optimisation. The results of the identification procedure are then presented and it is shown that a quadratic nonlinear model can be used to approximate higher order nonlinear ship dynamics and that by extending the model with linear terms, a single model can be used to reproduce either linear or nonlinear ship dynamics.The second major part concerns the process of using the parameterised models to design a control system. A 2-DoF controller was designed and tested in simulations, which consists of an exact-inversion based feedforward part and a full-state feedback controller tuned with a linear-quadratic-regulator approach to accommodate for the system's multi-input-multi-output characteristic by allowing that each DoF can be adjusted separately. Finally, the proposed workflow is validated by using an independently implemented nonlinear and higher-order ship model to generate measurement data for the system identification process, as well as to show that the resulting controller performs well in a simulation that additionally incorporates wind and wave forces as environmental disturbances.