Advanced sewer asset management using dynamic deterioration models

The main purpose of this study is to develop Dynamic Deterioration Model that is capable of doing individual and group predictions. The model considers the potential effect of location related attributes (e.g. land use, highway crossing) in addition to physical and operational attributes (e.g. root problem, grease problem). To avoid a uniform treatment to the entire network, the new approach includes clustering model that groups sewer pipes based on their location related attributes and detail operational conditions. Later, the results from group prediction models are applied in pipe material selection based life cycle cost analysis while the results from individual prediction models are used in risk assessment model for the prioritization of future pipe replacement. The study shows that the patterns of deterioration among pipes within a network are indeed vary depending on their physical and operational conditions. The utilization of location related attributes and operational condition data are shown to be helpful to efficiently group pipes with a network into several clusters representing different pattern of deterioration. The comparison among three modeling techniques shows that decision tree has the best accuracy over regression and neural networks models. In the case of pipe material selection, a location sensitive life cycle cost analysis proven to generate a better recommendation that helps utilities to avoid making a wrong decision that may cost the agencies unnecessary expenses throughout the asset life. The uses of deterioration model for individual prediction in conjunction with semi parametric survival analysis is proven to be capable of quantitatively measures the criticality of individual pipe segment. Overall, the optimal utilization of agency's owned data is not only improves the accuracy of the outcomes but also shows how the agency can benefit from data collection efforts throughout the years.