Design Loads Generator: Estimation of Extreme Environmental Loadings for Ship and Offshore Applications.

Abstract

High-fidelity hydrodynamic loads computation systems have become available due to developments in fluid dynamics and computer science. However, the use of these programs during the concept design stage of novel marine systems remains relatively unpopular, partly due to prohibitive costs. Addressing this issue requires approaches in at least two directions. The first would be to improve the accuracy and speed of computation, while the second would be to find rational and accurate ways of determining design events. Compared to the many efforts being made in the former direction, the latter has often been considered an open question or an area of future research. The aim of the current dissertation is to address this very question.

Design processes can be highly subjective and vary significantly depending on projects. However, any rational design process should include the identification of design life and operating environments. The extreme response of a marine system under an operating environment for a finite time period should then be studied not as a deterministic event, but as a stochastic process. Therefore, it is fitting and proper to consider the distribution of extreme environmental loadings.

The distribution of extreme responses can be calculated without extensive Monte Carlo type computations through a probabilistic process, here designated as Design Loads Generator (DLG), redeveloped in this dissertation. More specifically, DLG is a process that can construct an ensemble of short input time series, the extreme responses of which follow the theoretical extreme value distribution of a Gaussian random variable for a given exposure time. The input time series are calculated based on the assumption of a linear system and a Gaussian seaway, which are both deemed fit and proper especially during the concept design stage. Moreover, the exposure time associated with the Gaussian process becomes a good measure by which the associated nonlinear responses can be bounded. This dissertation presents several examples that show, through the use of this strategy, how the distribution of even a highly nonlinear, non-Gaussian process can be bounded, suggesting DLG can supplement or even replace the rule-based design approach.