Study on Decision Fusion Identity Model of Natural Gas Pipeline Leak by DSmT

摘要

The acoustic detection was novel method for natural gas pipeline leak. In order to improve accuracy and stability detection system, redundant structures of multiple sensor was necessary. The complex background noise and various workingcondition adjective caused uncertainty, inadequacy and inconsistency of acoustic signal. In the process of multisource fusing identification, the high conflict among different sensor signal was inevitable. In this paper, the decision fusion model is built to identify natural gas pipeline leak. The decision fuse algorithm procedure includes signal preprocessing, feature extraction, basic relief assignment byBPneural network and decision fusion utilizing DSmT (Dezert–Smarandache Theory) and PCR5 rule. Experimental results showthat the decision fusion model is effective and feasible.The information conflict of among different acoustic sensors is resolved perfectively. The fusion results for 150 group test samples indicate that the accuracy of leak detection reach 94.7% under the given condition.