PSI - Issue 68

Hamidreza Rohani Raftar et al. / Procedia Structural Integrity 68 (2025) 1066–1073 Hamidreza Rohani Raftar et al./ Structural Integrity Procedia 00 (2025) 000–000

1067

2

capability encounters several obstacles, particularly regarding the storage of compressed gaseous hydrogen within a hydrogen-centric energy system [5,6]. Low-alloy and carbon steels having a body-centred cubic (BCC) structure emerge as the top choices due to their affordability. Nonetheless, an understanding of how hydrogen affects these steels remains incomplete [7–10]. In this context, various research studies have conducted investigations into hydrogen-assisted fatigue crack growth (HAFCG) [11–13]. There was a prevailing belief that low or medium-strength steels exhibit a degree of resilience against hydrogen embrittlement (HE) induced by fatigue, attributed to their high threshold stress intensity values observed in fracture toughness tests [14]. However, FCG test results highlight that these materials can exhibit HE susceptibility even at stress intensity levels below the threshold for HAFCG [15,16]. In environments containing hydrogen, the trend of FCG in steels exhibits a remarkable correlation with the stress intensity at the crack tip. This phenomenon is commonly explained as an interaction or combination of sustained load stress-corrosion cracking and mechanical fatigue [17–19]. To investigate HE induced by fatigue in low-carbon steels with diverse chemical compositions, extensive FCG tests were conducted under different hydrogen gas pressure conditions [14,20,21]. However, comprehensive studies that compare these experimental results and facilitate the prediction of FCG trends under hydrogen exposure are currently lacking. Additionally, understanding the factors that influence HAFCG can help develop effective predictive models and reduce damage caused by hydrogen in low-carbon steels. Recent advancements in machine learning techniques have proven effective in estimating the mechanical properties of various materials, such as high-temperature alloys and steels [22–24]. This advanced data analysis method provides a fresh approach to speed up and simplify materials design. Machine learning also uncovers deep insights from complex datasets in experiments and simulations, while reducing development costs, risks, and time. Despite these advancements, there has been limited exploration into using machine learning to predict hydrogen embrittlement trends in low-carbon steels and analyze factors influencing FCG. In this study, which incorporates a literature review and experimental results, a workflow involving data-driven methods, correlation analysis, and prediction (illustrated in Fig. 1) has been implemented. Additionally, experimental results are presented to analyze FCG in a hydrogen environment and validate the predictive model. The focus is on predicting FCG trends in low-carbon steels under hydrogenating environments at various pressure levels. Utilizing a data-driven approach, the analysis aims to assess the significance of factors such as hydrogen pressure and alloy elements in influencing FCG behavior. By conducting comprehensive data analysis and modeling, this research aims to identify key parameters and insights that can support the development of effective strategies for predicting hydrogen embrittlement in low-carbon steel. Moreover, this research involves analyzing FCG based on experimental data in pipeline steels, presented as a case study.