The traditional approach to corrosion modeling is relatively straightforward. Corrosion is expressed as
a rate of penetration per year and failure is defined when a critical loss of section is predicted to have
occurred. Such an approach is well suited to applications involving ships and pipelines where a small
local leak can lead to more widespread problems. However, if this approach is applied to civil
structures then it generally proves to be overly conservative. Civil structures are usually not affected by
global failure due to a local event.
This paper describes the development of an alternative approach that takes into account the overall
variability in the environment and exposure conditions during the corrosion process. This generates a
basic data set that can be employed to represent the natural variability in extent and severity of
corrosion occurring on a structure. This was collated into cumulative frequency curves for several
common metallic systems and a Weibull distribution fitted to the data. The model has been verified
against several different types of structures. The results confirm that compared to the traditional
approach, the predictions generated by this method are less conservative, more realistic and greatly
improve confidence in the prediction of service lives.
Key words: corrosion, modeling, structures.