Application of SVM and ELM Methods to Predict Location and Magnitude Leakage of Pipelines on Water Distribution Network
A.Ejah Umraeni Salam, Muh. Tola, Mary Selintung and Farouk Maricar
Abstract
In this research, the system of leakage of pipelines detection will be done by a computerized technique by using analysis of pressure monitoring as a determinant of presence of pipeline leaks in the water distribution network. The pressure data obtained from EPANET software, namely a modeling in a hydraulic system. This study uses two methods, artificial intelligence, namely Support Vector Machine (SVM) and Extreme Learning Machine (ELM) which the results can be compared in order to predict the magnitude and location of leakage. Overall, both of these methods can be used to predict the magnitude and location of leakage. The accuracy of predictions for the magnitude and location of leakage of these methods is based on the value of NRMSE. In this case the results obtained by using the method of ELM are more accurate compared than the method of SVM of the entire pipeline systems.
Keyword
Leakage of pipelines , EPANET, SVM, ELM, NRMSE.