Comprehensive Evaluation of NBR Sealing Performance Under Medium-Pressure Hydrogen Aging: Experimental Analysis and Predictive Model Using Convolutional Neural Network

Abstract

The study investigates the sealing performance of nitrile butadiene rubber (NBR) seals in hydrogen facilities, with a focus on safety and reliability through extensive experimental analysis and accelerated aging simulations. This research involved a range of characterization tests, including tensile strength, swelling, hardness, and resistivity measurements, as well as scanning electron microscopy (SEM) morphological examinations, to assess material degradation under hydrogen pressures up to 5000 psi at room temperature. The results revealed significant changes in mechanical and physical properties, including reductions in tensile strength and hardness, increased resistivity, and material swelling, all of which were attributed to hydrogen-induced structural degradation. Morphological analysis via SEM highlighted surface roughening and microcrack formation at higher pressures. A Multivariate regression model was developed to predict key properties such as tensile strength, resistivity, weight, and hardness as functions of hydrogen pressure. Additionally, a hybrid machine learning model incorporating convolutional neural networks (CNNs) for feature extraction with fully connected layers for regression was also developed, demonstrating strong predictive performance for NBR’s material properties using both numerical features and SEM images. These findings contribute to the development of safer and more reliable sealing materials, crucial for the hydrogen energy industry, ensuring that seals can withstand the unique challenges posed by hydrogen storage and transportation.