Early Warning Method for Gas Pipe Leakage Based on Multi-source Heterogeneous Data Fusion

More and more pipeline leaks are causing irreparable effects. We present an early warning method for gas pipeline leakage based on multi-source heterogeneous data fusion. It the data layer, we categorize data and perform data preprocessing operations separately. In the feature layer, SVM model and DCNN model are trained to extract the features of the two types of data and perform discriminative classification respectively. In the decision-making layer, based on the D-S evidence theory, the two types of model discrimination results are fused and decisions are made to realize the real-time warning of gas pipeline leakage. We have effectively realized the efficient fusion of heterogeneous data from multiple sources, eliminated the uncertainty of data sources, and effectively avoided the property loss and environmental hazards caused by gas pipeline leakage.