Modeling and control of a SES in various operational modes

Abstract

Accessibility and comfort of often go in compromise of cost efficiency of the vehicle for offshore commutation. The Surface effect ship (SES) is one of the most cost-efficient and comfortable vehicles currently available for offshore commutation. The Boarding Control SystemTM (BCS) damps vertical motions on a SES to increase accessibility to offshore structures for safe transfer of goods and personnel. Some of the controller parameters are currently tuned manually. This tuning process poses extra workload for the crew aboard and is a highly inefficient way of tuning the controller. This thesis presents a boarding control system where one (of two) control parameters are autonomously tuned as well as an extensive simulation model for a surface effect ship (SES) in zero forward speed and head-sea. The vessels heave and pitch motions are exclusively analysed in this study. The control system presented in this thesis, is an extension of the BCS (Auestad et al. 2015), using a machine learning approach. The simulation model is created by using the framework of Marine System Simulator MSS (Fossen & Perez 2016 (accessed April, 2018) and extending the hydrodynamic part of existing mathematical model of the SES with linear interpolation. Based on the simulation model, stochastic gradient descent is used for learning the system s response under various loading conditions and creates a signal which automates one of the tuning parameters. Results of the simulation model captured essential dynamics of the system earlier disregarded in the literature. Linear interpolation proved to be a powerful way to express time-varying potential coefficients which earlier were assumed to be constant. The results from the semi-autonomous boarding control system showed that the SGD algorithm was successful at creating a function that gives an adaptive tuning parameter, which enhanced the performance of the BCS. The simulation model is still to be verified with model-scale and full-scale experiments. The automated boarding control system still lacks an extensive stability analysis. Influences of unmodelled environments should also be considered for future research.