Deep Learning-Enabled Self-Sensing E-Gasket for Early Warning of Transportation Pipeline Leakage

In advancing structural health monitoring (SHM) of industrial pipeline networks, the real-time assessment of bolt-gasket-flange connections (BGFCs) is crucial for preventing hazardous fluid leakage. The most critical indicator of sealing integrity is the spatial distribution of contact pressure at the gasket-flange interface. However, conventional monitoring methods are inherently limited by environmental susceptibility, compromised mechanical sealing, and insufficient spatial resolution in contact pressure distribution. To address these limitations, this study presents a continuous self-sensing electronic gasket (e-gasket) based on a piezoresistive carbon black/nitrile butadiene rubber (CB/NBR) composite for the early warning of sealing failure. Featuring a uniform percolation network within a hyperelastic matrix, the e-gasket simultaneously performs structural load-bearing functions and enables spatial self-sensing of contact pressure. Systematic evaluations of sensing performance reveal that the composite achieves the highest sensitivity of −0.2185 MPa–1 within the range of 0–1.9 MPa. To overcome the ill-posed problem caused by the soft-field effect and crosstalk errors, a data-driven hybrid strategy is proposed. Utilizing a customized U-Net architecture and a validated finite element model, this methodology maps the resistance matrix to a high-resolution contact pressure distribution. In the application demonstrations involving single- and double-bolt loosening, the method achieves an image correlation coefficient (ICC) exceeding 0.97 and a relative image error (RIE) below 0.15. Experimental results verify the effectiveness of the e-gasket in precisely localizing potential leakage pathways, enabling a reliable and noninvasive solution for comprehensive pipeline SHM.