Probabilistic modeling of cast iron water distribution pipe corrosion

Abstract

Due to the random nature of the corrosion process, stochastic approaches are more appropriate to mathematically represent the corrosion depths on metallic objects. Based on data from 202 pipes, a model was developed to compute the probability of finding maximal corrosion depth in a given interval of values for 150-mm cast iron water distribution pipes. Only the age of pipes was taken into account as an explanatory variable to compute this probability, since the soil characteristics were not available in the close surroundings of the inspected pipes. The model combines two functions: (1) a Weibull distribution function to represent the distribution of pipe ages at the time when the maximal corrosion depth reaches 100% of the pipe wall thickness; and (2) a generalized extreme value (GEV) distribution function, with the location parameter varying as a function of pipe age, to represent the distribution of maximal corrosion pit depths on pipes that did not reach a maximal corrosion pit equal to 100% of their wall thickness. The developed model offers a good representation of the distribution of observed maximal corrosion depths for Quebec City's 150-mm cast iron water pipes.