Pittaras, Athanasios; (1993) Adaptive signal prediction with application to ship motions. Doctoral thesis (Ph.D), UCL (University College London).

Text Adaptive_signal_prediction_wit.pdf Download (16MB)

Abstract

The purpose of this research work was to develop adaptive signal processing algorithms suitable for ship motion prediction. The prediction of signals generated by a system is based on the assumption that a suitable mathematical model for the system under study exists. This mathematical model may then be suitably transformed to arrive at a predictor for the system. The transformation is straightforward and what is more interesting is the problem of system modelling. If the underlying system generating the signals to be predicted is time-varying, the modelling task becomes more difficult since a single model cannot adequately describe system behaviour. The approach proposed in this thesis is to attempt to estimate a mathematical model for the system of interest in real time and use this for signal prediction. Sea vessels are an important class of physical systems where there has been considerable interest with regard to the prediction of their motion over a short prediction horizon (10s to 30s ahead). These motion predictions may be used to improve the safety of operations at sea, the transfer of loads onto or from ships and the deployment of ROVs (Remotely Operated underwater Vehicles) or diving bells in the splash zone. The research work described in this thesis is naturally divided into two parts. In the first part the abstraction of systems by mathematical models is considered and linear models and their properties are discussed. The optimal transformation, in terms of the minimisation of the squared prediction error, of a linear model into predictor form is then given. Next a general framework for parameter estimation algorithms for linear models is presented with details of the implementation of such an algorithm for a specific model form. The first part is concluded with a survey of techniques used for model structure and order selection and model validation. In the second part the estimation and prediction algorithms discussed in the first part are applied to the prediction of ship motion signals obtained from actual sea trials and ship model tests.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item