DATA-DRIVEN COMPUTING FOR THE STABILITY ANALYSIS AND PREDICTION OF FLUID-STRUCTURE INTERACTION

Abstract

The primary aim of this thesis is to explore data-driven methods for the problems of stability analysis and prediction of fluid-structure interactions. Accordingly, the current work is categorized into two main parts. The first part deals with the application of data-driven stability analysis for various canonical problems related to the offshore industry. The proposed data-driven approach relies on the Eigensystem Realization Algorithm (ERA) to design ROM models in a state-space format. The second part of the thesis deals with the problem of prediction using fluid dynamics data. The dynamical prediction has a great role in the design of real-time control and monitoring systems. Two hybrid deep learning models are proposed in this work and they are applied for various problems related to the offshore industry. In the first model, the low dimensional features are obtained via proper orthogonal decomposition (POD) as POD modes. The temporal evolution of modal coefficients is learned via recurrent neural networks. A more complete nonlinear reduced order model is developed by replacing POD with convolutional neural networks for obtaining the low dimensional features. The integration of CNN with RNN results in a convolutional recurrent autoencoder network model and is found to be more efficient.