Intelligent detection method of low-pressure gas system leakage based on semi-supervised anomaly diagnosis

作者：

Highlights：

• Semi supervised leak detection method without leakage data training.

• Deep autoencoder based on LSTM network and attention mechanism.

• Two-dimensional anomaly diagnosis method supporting expert decision.

• Based on actual pipe system data, leak detection accuracy can reach 85.3%.

摘要

•Semi supervised leak detection method without leakage data training.•Deep autoencoder based on LSTM network and attention mechanism.•Two-dimensional anomaly diagnosis method supporting expert decision.•Based on actual pipe system data, leak detection accuracy can reach 85.3%.

论文关键词：Leak detection,Low-pressure pipeline,Semi-supervised,STS autoencoder,Abnormal diagnosis

论文评审过程：Received 30 December 2021, Revised 30 June 2022, Accepted 1 August 2022, Available online 5 August 2022, Version of Record 11 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118376