Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.284729
Title: Ship steering control using feedforward neural networks
Author: Unar, Mukhtiar Ali
ISNI:       0000 0001 3540 9402
Awarding Body: University of Glasgow
Current Institution: University of Glasgow
Date of Award: 1999
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
One significant problem in the design of ship steering control systems is that the dynamics of the vessel change with operating conditions such as the forward speed of the vessel, the depth of the water and loading conditions etc. Approaches considered in the past to overcome these difficulties include the use of self adaptive control systems which adjust the control characteristics on a continuous basis to suit the current operating conditions. Artificial neural networks have been receiving considerable attention in recent years and have been considered for a variety of applications where the characteristics of the controlled system change significantly with operating conditions or with time. Such networks have a configuration which remains fixed once the training phase is complete. The resulting controlled systems thus have more predictable characteristics than those which are found in many forms of traditional self-adaptive control systems. In particular, stability bounds can be investigated through simulation studies as with any other form of controller having fixed characteristics. Feedforward neural networks have enjoyed many successful applications in the field of systems and control. These networks include two major categories: multilayer perceptrons and radial basis function networks. In this thesis, we explore the applicability of both of these artificial neural network architectures for automatic steering of ships in a course changing mode of operation. The approach that has been adopted involves the training of a single artificial neural network to represent a series of conventional controllers for different operating conditions. The resulting network thus captures, in a nonlinear fashion, the essential characteristics of all of the conventional controllers. Most of the artificial neural network controllers developed in this thesis are trained with the data generated through simulation studies. However, experience is also gained of developing a neuro controller on the basis of real data gathered from an actual scale model of a supply ship. Another important aspect of this work is the applicability of local model networks for modelling the dynamics of a ship. Local model networks can be regarded as a generalized form of radial basis function networks and have already proved their worth in a number of applications involving the modelling of systems in which the dynamic characteristics can vary significantly with the system operating conditions. The work presented in this thesis indicates that these networks are highly suitable for modelling the dynamics of a ship.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.284729  DOI: Not available
Keywords: TC Hydraulic engineering. Ocean engineering Ships Offshore structures Automatic control Control theory
Share: