Sewer networks are essential to the functioning of urban areas, and the deterioration of their pipes makes it increasingly important to anticipate changes in their condition. This can help improve the planning of maintenance and rehabilitation operations, particularly when inspection data are limited. The proposed approach is applied to 43 sewer collectors in the Algiers sanitation network, using data from a single inspection campaign, including pipe age, slope, length, material type, and observed condition. In this study, ordinal logistic regression and a Markov chain are combined to estimate the transition probabilities between different condition states of sewer pipes.
Pipe condition is described using three ordered states: Good, Fair, and Poor. Ordinal logistic regression is first used to examine the factors associated with pipe condition and to estimate the probability of each state. A global model is initially fitted using age, slope, length, and material type. Based on a 10% significance level, a reduced model is then obtained, with age and slope retained as the explanatory variables. The individual probabilities estimated from this model are subsequently grouped according to the initial condition of the pipes to obtain a global transition matrix.
The results indicate that pipes initially in Good condition have a 62.9% probability of remaining Good and a 35.6% probability of moving to Fair condition. For pipes initially classified as Fair, the probability of remaining in the same state is 74.3%, while the probability of transitioning to Poor condition is 25.7%. Overall, the proposed approach provides a simple probabilistic way to anticipate changes in pipe condition from a limited inspection dataset, and can support the planning of future maintenance and rehabilitation activities.