Assessment of Flaws in Non-Stress Relieved Carbon Steel Welds Caused by Hydrogen Attack

Abstract

Abstract Cracking due to high temperature hydrogen attack (HTHA) has been observed in non-PWHT’d carbon steel process equipment at conditions of temperature and hydrogen partial pressure below the original design limits recommended in API RP 941, necessitating changes to that standard. Consequently, flaw assessment procedures are needed to manage defects detected during inspection, or to establish appropriate inspection frequency. The latter typically involving estimation of the time for a detected or postulated crack to reach a critical size. This type of evaluation has been difficult to perform owing to the scarcity of data of fracture toughness as well as crack growth rate for steels in high temperature hydrogen. To address this gap, an experimental program was undertaken to help describe the ductile tearing characteristics of steel removed from service with various levels of HTHA damage. Near full thickness single edge notched tension SEN(T) specimens were machined from field samples and tested using existing “natural” cracks as the starter-crack. This provided insight into the behavior of real flaws subject to constraint conditions closely matching circumferential flaws in piping. Tests of undamaged steel were also performed in hydrogen at conditions designed to produce HTHA and compared with tests run in nitrogen. Crack growth tests obtained from the literature have been used to develop an empirical crack growth law for use in fitness for service assessments. The C* integral was also explored as a parameter for describing crack growth rate due to the strong similarity of HTHA damage to creep. The key results show substantial reduction in tearing resistance resulting from HTHA damage. A crack growth law similar to the Nikbin-Smith-Webster (NSW) model using the C* integral was found to show promise in describing the combined effects of creep and HTHA on crack propagation, although additional testing is needed to validate the correlation.