Prediction of daily failure rate using the serial triple diagram model and artificial neural network

• Published in: Water science & technology. Water supply Vol. 22; no. 9; pp. 7040 - 7058
• Main Author: Kizilöz, Burak
• Format: Journal Article
• Language: English
• Published: London IWA Publishing 01.09.2022
• Subjects: Age; artificial neural network; Artificial neural networks; Asbestos; Cathodic protection; Cement; Charts; Daily; Decision making; Distribution; Drought; Failure; failure rate; Failure rates; Flow velocity; Geographic information systems; Geographical information systems; High density polyethylenes; Information systems; kriging; Methods; Model accuracy; Neural networks; Performance prediction; Polyethylene; Predictions; Rain; Regression analysis; Remote sensing; serial triple diagram model; Soils; Support vector machines; Variables; Water distribution; water distribution network; Water distribution systems; Water engineering; Water quality; Iran; Canada
• ISSN: 1606-9749; 1607-0798
• DOI: 10.2166/ws.2022.315

Abstract In this study, 41 models used for the prediction of daily failure rates in water distribution networks have been designed via the Serial Triple Diagram Model (STDM) and artificial neural network (ANN) methods. For this purpose, daily failure data measured coordinately in the water distribution system network in Gebze and transferred to the geographic information system (GIS) has been used. The data has been normalized through the min-max technique to scale it at regular intervals and develop the model prediction performance. In this study, certain meteorological variables such as temperature and precipitation have been taken into account as model input for the first time. According to the increasing values of these two variables, it is observed in the model results that daily failure rates tend to increase. The expected model accuracies in failure rate prediction could not be obtained through the suggested ANN models. The higher prediction performances have been obtained through the STDM, a structure that enables visualization of the model results by making inferences. The STDM method is a significant alternative approach to determine the relationship of the variables on the failure and to predict the failure rate. It is predicted that the suggested STDM charts will contribute to the decision-makers, experts, and planners to determine effective infrastructure management. Also, investment planning prioritization will be able to reduce failure rates by interpreting prediction charts.