Hybrid models for handling variability and uncertainty in probabilistic and possibilistic failure analysis of corroded pipes

• Likelihood of failure of corroded pipes is calculated using inspection data.
• Inspection data simultaneously suffer from variability, uncertainty and imprecision.
• Two hybrid probabilistic–possibilistic models for handling imperfect data are presented.
• Strengths and weaknesses of different approaches are discussed.

The pipes carrying oil and gas are inspected for any damage that can lead to leakages and ruptures. Unfortunately, the data collected from these inspections is always inflicted with a number of imperfections, like variability, uncertainty and imprecision. These imperfections can be handled using the traditional probabilistic approach or the possibilistic approach. Both of these approaches have their own advantages and disadvantages. The probabilistic approach tends to give very low probability for events resulting from the low probability input values. This can often give a false sense of security, especially when the “weakest” link may cause a major accident. On the other hand the possibilistic approach is rather imprecise and may give over conservative and uneconomical recommendations. The recent advances in the field of fuzzy-probabilistic modelling offer an improvement in the way the calculations have been traditionally carried out by using the strengths of both techniques. This paper presents two different hybrid approaches for calculating the likelihood of failure of corroded pipes under internal pressure.