This paper demonstrates the statistical approach for describing failures and lifetimes of water mains. The statistical approach is based on pipe inventory data and the maintenance data registered in the data base. The approach consists of data pre-processing and statistical analysis. Two classes of statistical models are applied, namely counting process models and lifetime models. With lifetime models, one can estimate the probability which a pipe will fail within a time horizon. With counting process models one can see the deteriorating (or improving) trend in time of a group of “identical” pipes and their rates of occurrence of failure (ROCOF). The case study with the data base from Trondheim municipality (Norway) demonstrates the applicability of the statistical approach and leads to the following results: 1). In the past 20 years, Trondheim municipality has experienced approximately 250 to 300 failures per year. However, the number of failures per year will significantly increase in the near future unless better maintenance practice is implemented now. 2). Unprotected ductile iron pipes have a higher probability of failures than other materials. The average lifetime of unprotected ductile iron pipes is approximately 30 to 40 years shorter than the lifetime of a cast iron pipe. 3). Pipes installed 1963 and 1975 are most likely to fail in the future; 4) The age of a pipe does not play a significant role for the remaining lifetime of the pipe; 5). After 2 to 3 failures, a pipe enters a fast-failure stage (i.e., frequent multiple between failures).
Estimation in some Counting Process Models with Multiplicative Structure