Prediction of Failure Frequency of Water-Pipe Network in the Selected City

Abstract

The paper presents the modelling results of failure rate of watermains, distribution pipes and house connections in one Polishcity. The prediction of failure frequency was performed usingartificial neural networks. Multilayer perceptron was chosen asthe most suitable for modelling purposes. Neural network architecturecontained 11 input signals (sale, production, consumptionand losses of water, number of water-meters, length andnumber of failures of water mains, distribution pipes and houseconnections). Three neurons (failure rates of three conduitstypes) were put to the output layer. One hidden layer, with hiddenneurons in the range 1-22, was used. Operating data fromyears 2005-2011 were used for training the network. Optimalmodel was verified using operational data from 2012. ModelMLP 11-10-3 was chosen as the best one for failure rate prediction.In this model hidden and output neurons were activatedby exponential function and the learning was done using quasi-Newton approach. During the learning process the correlation(R) and determination (R2) coefficients for water mains, distributionpipes and house connections equaled to 0.9921, 0.9842;0.8685, 0.7543 and 0.9945, 0.9891, respectively. The convergencesbetween real and predicted values seem to be, from engineeringpoint of view, satisfactory.