New insights from probabilistic modelling of corrosion in structural reliability analysis

• Published in: Structural safety Vol. 88; p. 102034
• Main Author: Melchers, Robert E.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 01.01.2021; Elsevier BV
• Subjects: Chemical analysis; Computational fluid dynamics; Concrete structures; Containment; Corrosion; Corrosion products; Crude oil; Deterioration; Extreme value analysis; Extreme values; Fiber reinforcement; Fibers; Glass fiber reinforced plastics; Glass fibers; Glass fibres; Infrastructure; Linearity; Localized corrosion; Mathematical models; Papermaking; Petroleum pipelines; Pipelines; Pitting; Pitting (corrosion); Probabilistic models; Reliability analysis; Reliability engineering; Seawater; Stainless steel; Stainless steels; Steel plates; Steels; Structural reliability; Water analysis; Water injection; Structural reliability; Pitting; Glass fibres; Extreme value analysis; Steels; Deterioration
• ISSN: 0167-4730; 1879-3355
• DOI: 10.1016/j.strusafe.2020.102034

•Data trends for extremes on Gumbel plots are not linear as conventionally assumed.•Data show clustering of pit depths or defect sizes that can be associated with the step-wise development.•Observations provide new insights for the proper representation of the probability distribution of maximum pit depth and defect sizes.•Probability distributions for pit depths in water injection pipelines and in oil production pipelines shows similarities not normally suspected.•Broad similarities between probability distributions for pit depths and for glass fibre defects. Infrastructure intended to contain or exclude fluids or gasses, or reinforced with glass or other fibres may be subject to failure caused by pitting corrosion of the containment system or of the reinforcement. To assess the reliability of the infrastructure then requires probabilistic models for the deterioration processes involved. Typically the critical deterioration is through localized pitting and in particular the most extreme pit depth as this governs containment capability and also fibre strength. The ‘arch-typical’ distribution for representing maximum pit depth is the Gumbel extreme value distribution, for which complying data plot linearly on a so-called Gumbel plot. However, there is increasing evidence that large, homogeneous datasets show significant deviations from linearity. This is demonstrated for steel plates continuously immersed in seawater, for the interior corrosion of water injection pipelines and for crude oil production pipelines. Further examples are given for the maximum pit depth on stainless steel rollers in the papermaking industry and for localized corrosion of glass fibre reinforcement in concrete structures. In each case the non-linear trends obtained permit a re-interpretation of, and new insights for, the underlying physico-chemical-material mechanisms. The results are important for accurate representation of deterioration processes and for best-practice structural reliability analyses.