The physical degradation of water pipes can result in the loss of structural integrity (breaks/leaks), water quality deterioration and/or insufficient hydraulic capacity. This research, funded by the NSF, and undertaken by Polytechnic University, New York, NY in collaboration with the Cemagref, France, and the New York City Department of Environmental Protection evaluates whether and under what conditions a statistical approach, failure analysis, could model the structural degradation of a large and complex urban utility (New York City is used as the case study). Failure analysis introduces Weibull Proportional Hazard Models (WPHM); the risk factors of relevance, the parametric hazard and survival functions and the probability of failure can be generated.; The most promising models (Cox and Weibull PHMs, NHPP) have so far been tested at a few European utilities with the following characteristics: (1) Water networks in rural or suburban settings (non grid, few connections, long defined segments, low traffic and activity, low population density, little or no underground utility.) (2) Risk factors of relevance mostly pertaining to the pipes (age, diameter, material); traffic also sometimes shown to be relevant. (3) Long-term break data available (left truncation 20%).; Analysis of the New York City data set indicates that stratification on the material improves the results; stratification on the history of breaks does not. Short-term break data is an acceptable alternative when long-term break history is not available. Results were more significant with the ‘steel’ stratum. Risk factors pertaining to the pipes (diameter, length, material, date of pose) were, in some strata, statistically significant. Previous failures were in all cases the most significant variable. In some strata, risk factors pertaining to the environment (water zone, proximity to subway or highway) showed modestly significant results. Relevance of the models is estimated using validation indices.; When prioritization only is sought, a ranking of the ‘at risk’ pipes is obtained by fitting the data with a Cox PHM. A simple index, the Cox Index of Physical Condition (CIPC) is created and shown to produce a short-term forecast equivalent to the predictions obtained through Monte Carlo simulations on the Weibull survival curves.
展开▼