A new sensor fault diagnosis method for gas leakage monitoring based on the naive Bayes and probabilistic neural network classifier. (15th May 2022)

• Record Type: Journal Article
• Authors: Tan, Qiong; Mu, Xiaowei; Fu, Ming; Yuan, Hongyong; Sun, Jinhua; Liang, Guanghua; Sun, Lu
• Is Part Of: Measurement. Volume 194(2022)
• Publication Date: 2022-05-15
• Subjects: Gas pipeline safety monitoring -- Anomaly data identification -- Naive Bayes classifier -- Probabilistic Neural Network -- Fault diagnosis; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111037
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A new sensor fault diagnosis method for gas leakage monitoring is proposed derived from the Naive Bayes Classifier and Probabilistic Neural Network. The effectiveness of the developed method is verified in the urban gas pipeline monitoring systems. The global accuracy of sensor fault diagnosis is over 95%. Abstract: Gas monitoring sensor is prone to failure and its fault type is difficult to identify due to harsh working condition. In this work, a new sensor fault diagnosis method for gas leakage monitoring has been proposed derived from the Naive Bayes Classifier (NBC) and Probabilistic Neural Network (PNN). Firstly, NBC is used to identify the abnormal safety monitoring data. Then PNN is employed for sensor fault classification. The feasibility and effectiveness of this method are verified by applying it to the urban gas pipeline leakage monitoring system. It is shown that the abnormal monitoring data can be online distinguished, and sensor fault type can be effectively recognized. The global accuracy of abnormal data identification and the global accuracy of sensor fault diagnosis can reach 85% and 95%, respectively. This work can provide a guideline to improve the reliability of the urban gas pipeline monitoring systems.