Methodology And Application Of Pipe Condition Assessment In Urban Water Distribution System

As an important part of the urban infrastructure,continued efficient and safe operation of the water distribution systemis a basic guarantee of social production and people's daily life.Because of a large number of aging pipes,more and more pipe failures have occurred in recent years,which not only waste drinking water and increase costs,but also increase the risk of water safety and disturb social order.So it is necessary to make plan for the maintenance and renovation of the urban water distribution system.At present,water supply companies choose pipes to be maintenance and renovation according to the pipe material,failure times or experiences.Many factors have impacts on pipe condition,and the interaction mechanisms between various factors are not fully understood.In addition,pipe information is limited and the data quality is poor.All of these make it difficult to assess pipe condition.The purpose of this research is to identify the risk before pipefailures.The pipe condition assessment model based on intelligent machine learning algorithms was built on the water distribution system of a city in southern China.Based on the research results and experiences,the factors for pipe condition were summarized as physical factors,environmental factors and operational factors,which proposed standards for data collection and was helpful for modeling.In this case,pipe material,pipe diameter,pipe age,road load,operating pressure and stray current were chosen as factors.BP neural network algorithm and random forest algorithms were used to establish pipe condition assessment models.The model inputs were factors and the output was the probability of pipe failure.The ROC curve based on a 10-fold cross-validation was used to evaluate the stability and accuracy of the modelsimultaneously.The thematic maps were drawn on ArcGIS platform to show the spatial relationships of different condition pipes.Studies have shown that the accuracy ratesof the two models were both about 80%,basically consistent with the actual situation,which indicated that the model can effectively assess the condition of the pipe.On the basis of theoretical research,a pipe condition assessment system was established,applying.NET technology to complete the secondary development of ArcGIS.There were three functions,including information update,inquiry and statistics,data analysis.The system transformedthe scientific research into production practice and was helpful for asset management of water supply companies.