摘要详情

ID / 提交时间

96 / 2025-04-15 17:03:41

标题

Model for Natural Gas Pipeline Leak Detection Based on Temporal Convolutional Network

关键字

Natural gas pipeline leak detection, dilated gated convolution, self-attention mechanism Introduction

主题及专题

T05 状态监测与故障预测

状态

全文待审

作者

海清刘 / 四川轻化工大学

兆飞李 / 四川轻化工大学

世淋罗 / 四川轻化工大学

朝斌唐 / 四川轻化工大学

摘要

Abstract—Due to the high noise interference in leak signals and the strong similarity in signal patterns of different types of leaks, traditional methods struggle to accurately identify and classify leaks. To address this, this paper proposes a deep learning model based on Temporal Convolutional Network (TCN) for natural gas pipeline leak detection. The model integrates dilated gated convolutions and a self-attention mechanism to effectively capture the temporal features and key leak information from pipeline sensor data. The core of the model is the dilated gated convolution, which is used to extract temporal features, while the self-attention mechanism dynamically adjusts the feature weights. Each convolutional layer expands the receptive field through the dilation factor, allowing the model to capture long-range temporal dependencies. This structure enables the model to capture local features while also handling dependencies between distant time steps. Experimental results show that the method achieves an accuracy of 93.05% on the public GPLA-12 dataset across 24 leak categories, effectively distinguishing between normal and leak states, and providing a reliable solution for pipeline safety monitoring.