System identification theory approach to cohesive sediment transport modelling
Two aspects of the modelling sediment transport are investigated. One is the univariate time series modelling the current velocity dynamics. The other is the multivariate time series modelling the suspended sediment concentration dynamics. Cohesive sediment dynamics and numerical sediment transport model are reviewed and investigated. The system identification theory and time series analysis method are developed and applied to set up the time series model for current velocity and suspended sediment dynamics. In this thesis, the cohesive sediment dynamics is considered as an unknown stochastic system to be identified. The study includes the model structure determination, system order estimation and parameter identification based on the real data collected from relevant estuaries and coastal areas. The strong consistency and convergence rate of recursive least squares parameter identification method for a class of time series model are given and the simulation results show that the time series modelling of sediment dynamics is accurate both in data fitting and prediction in different estuarine and coastal areas. It is well known that cohesive sediment dynamics is a very complicated process and it contains a lot of physical, chemical, biological and ocean geographical factors which are still not very well understood. The numerical modelling techniques at present are still not good enough for quantitative analysis. The time series modelling is first introduced in this thesis to set up cohesive sediment transport model and the quantitative description and analysis of current velocity and suspended sediment concentration dynamics, which provides a novel tool to investigate cohesive sediment dynamics and to achieve a better understanding of its underlying character.