An advanced approach for spectral fatigue analysis of offshore platforms

The work presented in this thesis makes a number of significant advances in the field of spectral fatigue analysis of offshore platforms. The original objectives of the work were focused on one specific type of spectral analysis method called the deterministic-spectral method. All of the original objectives of the work were achieved, as follows. The spectral fatigue analysis method used is the so called deterministic-spectral method, so named because a deterministic approach is used to evaluate the system transfer functions which are then used in a conventional spectral analysis. Spectral methods are a very elegant and computationally efficient means of assessing structures where the response can be assumed to be linear, Gaussian, stationary and random. For such an analysis, computation of the response is straight forward once the transfer functions have been established. Classical methods of solution for the transfer functions become more difficult and for all but very simplified structures an approximate method must be used. The deterministic-spectral method is one such method. Because of the many assumptions which are made when using this method, consideration must be given to the errors. A comprehensive analysis of the limitations of the method was carried out and an important improvement to the implementation of the method, concerning the correct choice of the fundamentally important transfer functions, was made. The assumption of linearity was thoroughly evaluated and a new implementation of the spectral method was developed which enables the approach to be used on shallow water platforms. The new deterministic-spectral fatigue approach is based on the Longuet-Higgins wave height-period joint probability density function. It is used to produce better estimates for the important base wave cases in order to improve the accuracy of the deterministic-spectral method. These base wave cases have a fundamental influence on the calculated transfer functions because of system nonlinearities. This new approach makes the method more accurate and efficient. The deterministic-spectral method was used to assess the fatigue damage of typical shallow water platforms which experience a wide range of random loading condition due to the stochastic nature of the ocean waves. A new time domain based approach was used to compare with the spectral and deterministic methods. In addition, the effect of uncertainties of some important input environmental parameters needed to predict the fatigue life of offshore platforms was investigated. All fixed offshore platforms are fixed on foundations and these usually have nonlinear characteristics because of the surrounding ground conditions. Some typical soil models which have strong nonlinear stiffness were modelled using nonlinear springs in order to describe the behaviours of some typical soils. Their influence on the structural dynamic response of offshore structures with piled foundations was investigated. The primary limitation of the spectral method is that it assumes linearity of both the structural system and the wave loading mechanism. A neural network technique was developed for this nonlinear situation. It was used for calculating the response spectra of nonlinear structural systems. The results show that this new method is a very promising alternative method for nonlinear spectral analysis. A trained network for a specific offshore platform application was developed. This approach is a novel and original implementation of neural computing