Calibration of Model Uncertainties for Fixed Steel Offshore Platforms Based on Observed Performance in Gulf of Mexico Hurricanes

Data obtained during and after major hurricanes in the Gulf of Mexico (GoM) in the past three decades provide valuable information about the design model uncertainties for fixed steel offshore platforms because multiple offshore platforms were loaded close to and/or beyond the predicted capacity and, in particular, because multiple damages/failures were observed in the jacket superstructures, and a three-pile jacket system was well documented and extensively studied in the postfailure stage. The objective of this paper is to extract information on the current fixed offshore platform design practice by comparing the predicted versus observed performance of 18 offshore platforms in five major GoM hurricanes using Bayes’ theorem. The Bayesian calibration indicates that the American Petroleum Institute design recipe, on average, (1) is close to being unbiased for assessing the safety margin for the jacket superstructure performance, (2) tends to be unbiased for the pile lateral capacity using the upper bound lateral bearing factor in $p\text{-}y$ curves with the mean yield strength of pile steel, (3) is slightly conservative by 5% for the pile axial capacity in clay, and (4) is conservative by close to 50% for the pile axial capacity in sand. It is concluded that a jacket system with four or more legs/piles is less likely to fail in the foundation system than in the superstructure.