Machine learning-aided risk-based inspection strategy for hydrogen technologies

Abstract

Although technically challenging, effective, safe, and economical transport is crucial for enabling a widespread rollout of hydrogen technologies. A promising option to transport large amounts of hydrogen lies in employing retrofitted natural gas pipelines. Nevertheless, H2-rich environments tend to degrade pipeline steels, reducing their load-bearing capability and accelerating crack propagation. Regular inspection and maintenance activities can preserve the pipelines’ integrity and guarantee safe operations. The risk-based inspection (RBI) approach is based on estimating the risk for each component item. It focuses most inspection activities on high-risk components to reduce costs while maximizing the plant’s safety and availability. However, the RBI standards do not consider hydrogen-induced degradations and cannot be adopted for industrial equipment operating in H2 environments. This study proposes a novel ad-hoc methodology for the risk-based inspection planning of hydrogen handling equipment. A machine-learning model to predict the fatigue crack growth in gaseous hydrogen environments is developed and integrated with the conventional RBI approach. The proposed methodology is validated on three pipelines transporting hydrogen and natural gas in different concentrations. The results show how similar operating conditions can determine different degradation rates depending on the environment and highlight how hydrogen-enhanced fatigue can reduce the pipelines’ lifetime.