Abstract

In this paper, an approach is proposed to conduct reliability analysis on an offshore jacket considering corrosion degradation under extreme load cases. Corrosion degradation is considered as thickness wastage of the jacket element, which is seen as time-dependent variables. One probabilistic corrosion in literature is adopted by using different distribution models. Also, three different inspection cases (environmental conditions) of the corrosion are studied. The reliability assessment is evaluated by Crude Monte Carlo simulation based on the trained surrogate model. Deep neural networks are used to train the surrogate model, because they are not limited by the distribution and dimension of variables. The results show that using different corrosion distribution model, the probabilities of failure of the jacket are different, even though they have the same mean and standard deviation values. In addition, with same assumption of the distribution model in corrosion, the reliability of the jacket changes a lot concerning different inspection cases. Furthermore, it is noted that the inspection cases have more influences on the reliability analysis of jacket than different corrosion distribution assumptions. At the end, two recommendations are derived from this work.