MODELLING OF SERVICEABILITY CONDITION OF BURIED STORMWATER PIPES USING NEURAL NETWORKS

One of the major problems local government councils face today is the issues associated with proactive identification of maintenance, renewal and upgrade needs of stormwater drainage networks. Existing buried pipes are replaced today when they become under capacity or when they deteriorate beyond their performance levels. There are two types of deterioration taking place in stormwater pipes, namely structural deterioration and serviceability deterioration. Assessing pipe condition using CCTV units are very expensive and most local government councils cannot afford to use CCTV to assess all pipes. Hence it is vitally important to develop a reliable model to predict future pipe condition, both serviceability and structural, for the purpose of identifying maintenance, renewal and upgrade programs and budget forecasts. Recently few such models have been developed for predicting deterioration of pipe networks as well as individual pipes, base on limited data sets, though they are not yet available in the market. However the reliability of these models is still questionable as councils have not used these models yet. This paper describes the work that has been done so far, using Neural Networks, to develop improved deterioration models to predict serviceability deterioration of individual stormwater pipes. The resulting models were used in the Bentleigh catchment in Glen Eira City Council in Victoria, Australia