Probabilistic Methods: Risk Based Design and Assessment

Abstract Structural integrity assessments typically aim to calculate the integrity of a component under nominal or best estimate conditions. To account for potential variability and uncertainty present in the system, safety factors are often applied to assessment inputs and outputs. This approach does not allow the level of conservatism present to be quantified, often leading to over-conservatism or inadvertent non-conservatism. Probabilistic assessments explicitly calculate the probability of failure based on distributions of the input parameters and hence quantify the margin present in the assessment, leading to a greater understanding of the system. In this study a creep-fatigue damage assessment of a transiently loaded piping component is used as a vehicle to investigate some of the challenges and benefits of probabilistic assessments. A probabilistic assessment of the component life is compared to a lower-bound deterministic calculation to identify the mismatch in margin between the two results. The potential inaccuracies introduced when reducing the computational burden of Monte Carlo simulations with response surface methodologies are explored and tested. Finally, two challenges when attempting to underwrite a very low probability of failure are tackled: the inference of the shape of a distribution’s tails from limited experimental data and the uncertainty of extreme percentiles of finite Monte Carlo samples.