Investigation of deterioration models for stormwater pipe systems

Like other engineering structures, buried stormwater drainage pipes deteriorate and fail over time in terms of pipe collapses due to structural deterioration or pipe blockages due to hydraulic deterioration. The deterioration of service infrastructure was a concern in Australian in recent times, where stormwater drainage pipes in Australia were rated as ‘poor condition’. The information on current and future condition of stormwater pipes is therefore important for making decisions on when and how to carry out maintenance and rehabilitation. As the major objective, this study attempted to develop separately structural and hydraulic deterioration models that can predict the condition changes of pipe population and condition changes of individual pipes as compared to the ‘like-new’ condition. The outcomes of the models can be used for planning annual budget and prioritizing repairs. Furthermore, this study aimed to identify the significant factors that affect the structural and hydraulic condition of stormwater pipes, which could support design and operation of stormwater pipes. To achieve these objectives, this study first considered an ideal deterioration model which recognized that pipes deteriorate differently due to their contributing factors such as pipe size and soil type. Based on the ideal deterioration model, five practical deterioration models were developed using statistical techniques and neural networks (NNs), and were calibrated using different optimization techniques in searching for the best suitable model. These deterioration models were selected considering the availability of snap-shot (or once only) inspection data and the ordinal grading system of pipe condition. The model inputs were contributing factors and the model output was pipe condition in ordinal numbers. Methods for assessing the predictive performance of these models and determining the significant input factors were considered. A case study with data collected from a City Council in Melbourne (Australia) was used to demonstrate the applicability of the models developed in this study. The results showed that the NN model and the Markov (statistical) model were the best models for predicting condition changes of individual pipes and pipe population respectively. Several factors such as pipe size and pipe location were found significant factors in these models. The significance of this study is the development of deterioration models that provide a basis for the construction of a comprehensive asset management system for stormwater pipes. The major innovation of this study is the exploitation of advanced modelling techniques for predicting the deterioration process of stormwater pipes.