A stochastic corrosion model for structural steels subject to sea waters

In order to perform reliability analysis of steel structures subject to sea waters and thus to corroding marine environments, it is necessary to use: a) a model describing physical behavior of corrosion in structures placed in these circumstances and b) a stochastic model describing probabilistic behavior of corrosion phenomenon along time. In this study, Melcher's model is employed as a physical corrosion model. Furthermore, a probabilistic model is proposed in which the corrosion process is discretized into a few steps each of which representing a jump/an increment in corrosion quantity. Despite having a pre-defined deterministic small duration, each step is modeled with an independent random variable following a gamma distribution. In this paper, in contrast to the conventional gamma process, the shape and scale parameters of the involving gamma distributions are different for each step. This will result in an accumulating corrosion process which is a sum of n different gamma distributed random variables in any arbitrary point of time. Herein, in addition, a new application of "maximum likelihood method" for estimating of model parameters is proposed. Panama Canal data for carbon and low-alloyed steels is used for the purpose of parameter estimation. Reliability analysis of a structure subjected to an environment similar to that of Panama Canal is then investigated. Implementation of the model proposed herein to data available indicates that it could be sufficiently flexible to show the mean and variance of the marine corrosion processes.