Long-term fatigue assessment of deepwater risers in time domain
Date of Issue2017-12-27
School of Civil and Environmental Engineering
With the increasing demand for oil and gas, offshore industry moved offshore structures towards deeper waters and severe environmental conditions. Offshore structures are subjected to many environmental loads during their service life, such as waves, winds, currents and ice. Such environmental conditions are non-stationary processes. However, in common practice, it is always assumed that these environmental actions are statistically stationary for short-term sea state, typically for three hours. Riser system is one of the most critical offshore structure component and is exposed to a large number of sea states during its life span, especially in deep water environment. Therefore, only stationary analysis for the riser system is not sufficient, and a global dynamic analysis is indispensable to obtain its response due to the fluid-structure interactions. During the global dynamic analysis, extreme response and fatigue life are two important design concerns. Extreme response predictions can be obtained based on either short-term or long-term approaches. However, fatigue predictions can only be based on long-term methods due to their accumulative characteristics. The main objective of this study is to investigate various existing methods for their accuracy, limitations and efficiency, and to develop advanced methods in calculating the long-term fatigue damage of deep-water risers as well as in constructing a suitable statistical model for ocean environmental random variables. The traditional way of solving long-term fatigue problem is using a wave scatter diagram. By lumping several sea states into a small number of bins, one sea state is selected as a representation of each block to ensure that the damage quantity yielded by this sea state is not lower than the original sea states. It is apparent that such block method has a lack of accuracy. Therefore, there is a need for a fast and precise method for long-term fatigue analysis especially in time domain. The first part of the author’s research is to develop a new efficient approach for long-term riser fatigue analysis in time domain. This new approach allows one to simulate the wave amplitudes from a distribution different from the original one. Then, the results for many sea states can be obtained by simply altering the importance sampling weighting function. This proposed approach is enhanced by a succession of additional techniques to reduce the sampling variability. The study of this proposed new efficient approach is presented in Chapter 3. The stochastic nature of environmental loads such as wave and wind applied on the structure is quite complicated. In the random process related structural analysis, a suitable statistical environmental condition model is essential and should be carefully developed for the prediction of long-term performance of offshore structures. As a common practice, a conditional joint distribution model is used for the statistical environmental condition, which is popularly used in the industry. However, in such model, only a linear relationship between pairs of random variables is considered. If the actual dependency is nonlinear, those traditional models no longer estimate the relationship appropriately. Chapter 4 proposes a new approach by using a classical multiple dimensional copula model to treat the joint distribution of random environmental conditions. This classical multiple copula model is slightly superior to the traditional model by model selection, and the influence of different statistical models in the long-term fatigue analysis is studied. Due to the limitation of multiple copula type, an advanced copula model is developed in Chapter 5 to construct more flexible statistical models based on the copula concept, estimated by model selection methods. The application in the long-term fatigue analysis is also compared with traditional models and classical copula models. Compared with traditional models, the biggest advantage of multiple copula models is their ability to model nonlinear relationships among different environmental random variables.