Modelling the pitting corrosion statistics for cast iron pipes

Research Doctorate - Doctor of Philosophy (PhD)Many major city water supply distribution networks consist of buried cast iron pipes. In many cases, the pipes are internally cement lined and the predominant corrosion is by external pitting. This may cause leakage and eventually structural failure. Predicting the remaining life of older pipes is, therefore, a matter of interest both for engineers and asset managers. An accurate predictive model requires an estimation of the probability of failure and would be useful for reducing the costs of maintenance and for protection against unexpected failures. Since the pipes are buried, and thus cannot be inspected visually, there is considerable uncertainty about their condition. Indirect inspection techniques also have a high degree of uncertainty. For these reasons, there is an interest in long-term asset management to develop realistic models for the depths and areas of pitting corrosion in order to aid the estimation of the probability of failure of older pipes. It is generally considered that the deepest pits are of most interest in relation to failure. It is thus conventional to use the Gumbel Extreme Value distribution to represent the statistics of the maximum pit depths and to use it to estimate the probability of pipe wall perforation. Herein, the data obtained for the maximum pit depths for large sized (1 2m long) samples of 24 pipes exhumed from different, apparently randomly selected, locations after 34 to 129 years of service are examined for consistency with the Gumbel probability distribution. This was the case for the deepest pits, but the data for less deep pits showed a consistent pattern of departure from the Gumbel distribution. Some extreme pit depth data, inconsistent with the rest, are interpreted as being possibly caused by material imperfections. The likely pit size is also of interest in estimating leakage losses, and for assessing the mechanical behaviors and failure probabilities of pipes. However, to date, limited statistical models for the sizes of the pits have been proposed. Current work aimed at statistically modeling the progression of pit sizes for cast iron water pipes is also reported, again using data from actual pipes exposed for long periods. The present study thus examines pit area and pit length data computed from pit depth data. To investigate the progression of pitting corrosion, the relationships between the pit areas and the pit depths of each pipe is described. These results, and the results of the Extreme Value Analysis of the maximum pit lengths of old cast iron pipes, supports previous observations that the growth of pits is not a continuous process but occurs in stages, with each stage involving lateral growth and the amalgamation of pits, and then the formation of new pits in the plateaus formed by earlier corrosion