This paper contrasts two different models for describing the risk of failure of offshore platforms due to waves and other forces. The first model considered, in wide use in the industry, is based on the non-encounter probability or return period concept.
The second model, described in this paper, is based on the classical reliability function. In consist to the first model mentioned, the reliability model may be readily modified to treat time varying strength histories. A simple example analysis is presented for a jacket-type offshore structure. The advantages of using the reliability function approach are described.
Due to the random nature of natural hazards, economical and reliable design of drilling and production structures requires the application of probabilistic methods. It is generally recognized that the acceptance of a certain risk of failure is inevitable and tolerable in such structures. The accurate numerical evaluation of the order of magnitude of this risk is one of the key problems in risk analysis.
This paper is applicable to the general offshore risk problem, but only the risk due to waves is specifically treated here. Other types of risk like earthquakes and tsunamis may be treated similarly by this model with only slight modifications. Previous approaches to the structural reliability problem are not discussed here in detail, but they are reviewed briefly below.
Based on Gumbel's1 return period concept, Borgman2 introduced the notion of the non-encounter probability, which is now widely used among engineers. (See Appendix.) Since then work by Marshall3, Freudenthal and Gaither4,5, Russell6, Bea7,8, Stahl9, Schueller10, and Moses11 has expanded on these ideas.
In the past, load and strength uncertainties have been treated separately in many cases. This has been done in part for analytical simplicity, but the result is often a quite unrealistic risk evaluation. Numerous non-failures of Gulf of Mexico platforms exposed to more than design conditions (some water in the cellar decks) imply the presence of conservative factors in design practice. These conservative factors are probably related mostly to structural (including foundation) strength under-estimation.
The purpose of this paper is to introduce an easily understood, integrated and rational treatment of both the variability of wave loads for waves of a given height and the distribution of expected member and platform strengths. The method is based on the classical reliability function and easily permits treatment of member strengths which may significantly decrease with time, as is the case in fatigue-producing rough sea areas like the North Sea. The reliability function approach also can be used to arrive at a more rational design basis than that resulting from application of arbitrary safety factors. The standard analytical treatment relies on safety factors which often cannot be directly related to the actual probabilities of an overall structural failure. It is shown here that this pitfall may be avoided in an analysis based on the reliability function.
