FRAMEWORK FOR PROBABILISTIC REASSESSMENT OF AGEING JACKET PLATFORMS IN MALAYSIAN WATE

Due to advancement in current oil extracting technology, many ageing jacket platforms are pushed to be continued in-use beyond their design life. Thus, reassessment to determine reliability of these ageing platforms is pertinent. But no exact guide on reliability analysis is available. Furthermore, reassessment deals with numerous uncertainties. Hence, in this work probabilistic reassessment methods are developed. Simulation results from SACS finite element analysis are combined with numerical techniques; quadratic Response Surface and First Order Reliability Method to obtain reliability indices. The developed methods are validated prior to use. The novelty of this approach is in the simplicity of the procedure and the use of platform and environmental specific uncertainty variables to address the distinctive characteristics of the cases. Utilizing the developed methodology, further issues currently plaguing reassessment such as piled foundation ageing effect, recently proposed metocean load combination, effects of corrosion and statistical updating methods on reliability are investigated. Results show that combination of Failure Path and Simple Bound approach is able to establish relationship between component and system reliability, with some limitation. From the stepped up PSI curves using capacity ratio and average setup factor, reliability analysis was conducted. Minor improvements in the capacity and some variations in reliability indices were observed. The newly proposed metocean return period combination was found less conservative averaging from 17% - 95%. Linear corrosion model has produced polynomial reliability reduction function. Average reduction in reliability index in 40 years was observed to be 17.84%. From the Bayesian Updating, the updated probability of failure has shown significant increase beyond the design wave height. Truncation method is found efficient in updating probability of failure. The developed simplified approach in Linear and LOGEST methods have better correlation index ranging from 0.3695 - 0.7777 while LOG method has the smallest percentage error