Nonlinear response prediction of spar platform using artificial neural network / Md. Alhaz Uddin

Md. Alhaz , Uddin (2012) Nonlinear response prediction of spar platform using artificial neural network / Md. Alhaz Uddin. Masters thesis, University of Malaya.

Preview

PDF (Thesis M.A) Download (6Mb) | Preview

Abstract

Due to global energy demand, offshore industries are moving towards deep and ultra-deep waters for oil and gas exploration in ocean environment. Floating offshore structures such as spar platform is considered to be the most economic and suitable offshore structure in deep water regions. During oil and gas exploration, floating offshore structures may sometimes be affected by critical environmental forces. Quick decision must be taken either to continue or to stop production, on the basis of response prediction of offshore structures under forecasted environmental conditions. Finite Element Method (FEM) is an important technique to predict the response of offshore structures considering all nonlinearities. However, FEM is a highly time-consuming process for predicting the response of platforms and usually used as a final analysis tool. On the other hand, Artificial Neural Networks (ANN) can predict response in rapid mode. ANNs are also capable of providing efficient solutions to problems such as damage detection, time series prediction and control where formal analysis is highly complex. This study presents nonlinear response prediction of spar platform for various environmental forces using ANN. The neural network has three layers, namely the input, output, and hidden layer. A hyperbolic tangent function is considered in the present study as an activation function. Environmental forces and structural parameters are used as inputs and FEM-based time history of spar platform responses are used as targets. Feed-forward neural networks with back-propagation algorithm are used to train the network. After training the network, the response of the spar platform is obtained promptly for newly selected environmental forces. The response obtained using ANN is validated by conventional FEM analysis. It has been observed that using completely new environmental forces as input to ANN, the time history response of spar platform can be very accurately predicted. Results show that the ANN approach is very efficient and significantly reduces the time for predicting response time histories.

Actions (For repository staff only : Login required)