Proceedings of the

The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK

Evaluation of the Factors Determining Hydrogen Embrittlement in Pipeline Steels: An Artificial Intelligence Approach

Abhishek Subedi1,a, Alessandro Campari2,c, Antonio Alvaro3, Biraj Singh Thapa1,b and Nicola Paltrinieri2,d

1Department of Mechanical Engineering, Kathmandu University, Nepal.

2Department of Mechanical and Industrial Engineering, NTNU, Norway.

3SINTEF Industry, SINTEF, Norway.


Hydrogen is an emerging energy carrier with inherent environmental benefits. It has the potential to decarbonize industrial applications that require high-grade heat. In addition, hydrogen allows centralized clean energy production and distribution to remote end-use sites. For a smooth transition to hydrogen technologies, it is important to guarantee a safe and reliable distribution system. Hydrogen could be transported through the existing widespread pipeline network. Nevertheless, most pipeline steels were not designed for hydrogen service and are prone to hydrogen-induced degradation, which could result in sudden component failures and undesired releases with severe consequences. Hydrogen embrittlement depends on the interplay of three factors, i.e., the mechanical loading, the operating environment, and the material properties. The synergistic interaction of these parameters has significant safety implications. This study introduces a machine learning approach to evaluate the role of these factors in the occurrence of hydrogen-induced damages. Several pipeline steels have been assessed for embrittlement under different environmental and loading conditions. An extensive database has been created, and a decision tree model has been trained to predict the hydrogen embrittlement of materials. The main advantages of this model are its "white box" nature and simple interpretability. This artificial intelligence approach can ensure the safe application of hydrogen systems and allow advancements in inspection planning and predictive maintenance.

Keywords: Hydrogen embrittlement, Pipeline steel, Artificial intelligence, Machine learning, Decision tree, Inspection planning, Predictive maintenance.

Download PDF