Improving the benefits of sewer condition deterioration modelling through information content analysis

As urban sewer infrastructures age, it becomes increasingly important to make effective decisions to maintain the structural condition of the sewers at an acceptable level. To support the decision-making process, the utility manager can apply sewer deterioration models. However, the quality of the decision support from such models is dependent on the accuracy and reliability of the predictions, and previous research has shown that the sewer deterioration predictions can be unreliable. In this paper it is shown, by numerical experiment and analysis of information content, how the accuracy of sewer deterioration models is inhibited by data heterogeneity. The data heterogeneity arises when the condition class is used as a response variable, because the condition class is an aggregation of different failure modes, and contains information that does not describe structural deterioration. Based on these findings, the paper suggests changes to be implemented in the condition classification standard, which can mitigate heterogeneity and improve prediction reliability. The suggestions for improvement include distinguishing between structural and functional defect codes, defining new condition metrics better suited for deterioration modelling, and registration of detailed defect codes to allow distinction of different failure mechanisms.