ABSTRACT

In recent years, computer automation has helped to reduce the time and cost associated with the preparation of design criteria. The advantages of automation have offsetting disadvantages that need to be noted, such as the inherent limitations of wind, wave and current databases and the methods employed to estimate extreme criteria. The purpose of this paper is to illustrate the effects of the more common approaches to criteria development on the resulting estimates of extreme forces, which in turn have direct effect on the cost of a marine project. This paper examines three extreme criteria methods, two of which use parametric extreme value distributions such as Gumbel and Fisher-Tippett III and the third is a non-parametric method due to Nolte.

Parametric extreme value distributions such as Gumbel and Fisher-Tippett III assume that historical data constitutes a truly random series of extreme events. Whereas storm events such as hurricanes are unpredictable, they mayor may not be truly stochastic in mathematical sense. These distributions, when applied to real world data, yield divergent extreme values because the Gumbel distribution favors the lower end of the hindcast data sample and the FT-III distribution tends to favor the upper end. Depending on the trends in the upper and lower values in the hindcast data sample, the difference between Gumbel and FT-III distributions could be very large.

In a non-parametric method such as Nolte's, the distribution of wave height is calculated directly without curve fitting using entire hindcast storm history rather than the annual or storm maxima. In mathematical terms, the Nolte method computes the non-parametric annual exceedance probability P(x> x100 I hindcast storm histories). When the two curve fitting methods discussed above diverge from each other, the Nolte method extreme values are usually in between them. As real world data often do not satisfy the conditions necessary for the validity of those mathematical models, the Nolte method could provide a means of avoiding assumptions about the mathematical validity of a given model.

INTRODUCTION

A small percentage reduction in the design criteria for a large deep water structure could result in a substantial reduction in project cost. The cost of a deep water structure can easily exceed \$100 million and, assuming cost to be proportional to design forces, a few percentage point change in design criteria can change the present value of a deep water project by a few million dollars. Hence, the need for the lowest possible design forces consistent with safety requirements cannot be over-emphasized.

The estimates of extreme values of design criteria, for parameters such as wind speed, wave height and current speed, are highly dependent upon the method of estimation. For example, the search for a universally applicable extreme value distribution, which is central to estimating extreme criteria, generates heated debate at gatherings of statisticians and engineers. A committee of the International Association of Hydraulic Research (IAHR) is looking into this matter (References 1–3).

This content is only available via PDF.