Page 9 - Дисертаця Венгринюк

9

Based on the established relationships between hydrogen charging intensity

and mechanical loading rate and their effect on the fracture toughness characteristics

of 17H1S pipeline steel, determined using the J-integral method, the feasibility of

applying a low loading rate for evaluating fracture toughness under hydrogen

exposure has been substantiated. This approach ensures sufficient time for hydrogen

diffusion into the fracture process zone. Reducing the loading rate by two orders of

magnitude from the standard value (from 0.5 to 0.005 mm/min) significantly

increased the sensitivity of the steel’s fracture toughness parameters to hydrogen.

The steel in the as-received condition exhibited the highest susceptibility to

hydrogen embrittlement in terms of the β parameter (the slope of the J–R curves),

whereas the exploited steel showed the highest sensitivity in terms of J₀ (the J-

integral at crack initiation).

A criterion for reaching the limit state of transmission gas pipeline steel in terms

of fracture toughness under conditions of repurposing the pipeline for hydrogen

transportation was substantiated based on established relationships describing the

influence of hydrogen and loading rate on the fracture toughness characteristics of the

steel, as well as on the calculation of stress intensity factors at the tips of possible but
undetected crack-like defects in the pipe wall identified during technical inspection,

depending on their geometry and internal hydrogen pressure.
On this basis, an computational-experimental method for assessing the

degradation of steel in an in-service transmission gas pipeline under hydrogen
exposure was developed. It was demonstrated that, within the typical operating

pressure range of 3.5–7.5 MPa, the reduction in fracture toughness of steel after 38

years of service in the pipeline does not exceed allowable limits.

Based on an analytical review of machine learning methods (neural networks),

their optimal architecture was identified, and a computational model was developed

for predicting the temporal distribution of hydrogen concentration in the pipe wall

steel using physics-informed neural networks, providing high prediction accuracy.

Using this model, the evolution of hydrogen concentration distribution in the pipe wall

steel over time was evaluated.