Abstract:

The paper deals with statistical analysis of repeated pipe failures in Water Distribution Systems, affected by a time-dependent factor. Several systems observed over a given time period are considered. A particular parametric model, formulated in the counting process framework, is employed. This model accounts for system aging, harmful repairs and covariates. Practical issues in maximum likelihood parameter estimations in the presence of a calendar time-dependent covariate are discussed. The proposed methodology is applied to a real dataset. The results clearly show an improvement of estimates quality when a calendar time-dependent covariate is integrated to the model.