Identifying biases in deterioration models using synthetic sewer data

The assessment and validation of sewer deterioration models is difficult because reliable data are missing. This makes it hard to find the most suitable model for a particular application. A network condition simulator (NetCoS) is used to generate synthetic sewer data for defined test scenarios. Thereby, the deterioration and replacement of pipes, the expansion of the sewer network, and classification errors are considered. Based on such synthetic data, deterioration models are calibrated and their results compared with the predefined scenario. While this approach is not capable of proving that a model performs correctly on a real application, it highlights the strengths and weaknesses of a model. The influence of condition classification errors and the age of the sewer system is investigated for two deterioration models. The results show, that classification errors can introduce substantial biases in the parameter estimation of the Markov model while in comparison the applied cohort model is fairly robust. Young sewer systems with fewer pipes in bad condition states on the other hand, have a very strong influence on the parameter uncertainties of the cohort model while the Markov model proved to be less sensitive.