Leakage Detection in Pipeline Network Based on Random Forest Fusion

Abstract

For solving the difficult problem of leakage detection and diagnosis in pipeline, a method with random forest fusion and independent component analysis was proposed. Firstly, the theoretical background of random forest and independent component analysis was introduced. The ICA was used to reducing noise of the negative pressure wave signal and the flow signal. Then the RF fusion model was set up. The random forest data mining method is used to determine the characteristic parameters as the input parameters of random forest fusion to classify the working conditions of the pipeline network. The experimental results show that this fusion method can improve the accuracy and effectiveness of pipeline leakage diagnosis.