Grade 91 steel is widely used in tubes, headers and piping of superheaters and reheaters because of its higher strength at elevated temperature compared to low alloy steels such as Grade 22. The improved strength is a result of a tempered martensitic microstructure with a fine distribution of carbonitride precipitates. This microstructure is achieved through careful heat treatment: normalizing, tempering, and subsequent forming and post weld heat treatments. If these heat treatments are not performed properly, then the strength of the material essentially reverts to that of a low alloy steel like Grade 22, and is usually accompanied by a reduction in hardness, leaving the Grade 91 material in a so-called “soft” condition.
This article summarizes a case study for Grade 91 material in the “soft” condition, which was responsible for a steam leak after only 5 years of operation, illustrating how this material condition can result in forced shutdowns and safety hazards. It is because of these consequences that it is recommended to have a Grade 91 life management program to understand if your plant may have such vulnerability. This case study provides general background to the steam leak and describes the subsequent metallurgical evaluations performed to verify that mal-heat treatment of the Grade 91 steel was the root cause of the leak. A follow-on article (the next issue, Volume 45, of News and Views) will provide additional insight into local stresses and analytical prediction of such failures, as well as highlighting key aspects of a Grade 91 life management program. Suffice it to say if this plant had implemented such a program, the vulnerability of the affected spool would have been identified and mitigating actions could have been taken to avoid the leak.
A leak was detected in the main steam piping of the combined cycle plant after approximately 35,000 hours of operation. Upon investigation, the leak was located at a radiographic testing (RT) plug, apparently caused by excessive swelling (bulging) of the pipe section. The bulging was observed in the straight section of pipe downstream of a 90o bend. Figure 1 shows a picture of the bulged pipe with dashed lines illustrating the degree of swelling. Also shown in Figure 1 is a schematic of the failed component geometry with axial centerline and circumferential clock positions identified; these are referenced throughout this article.
After the leak was identified and the unit brought off line, in-situ (field) hardness testing was performed by another contractor (not Structural Integrity) on the bulged region with hardness values found to range from 158 to 209 Brinell hardness (HB). Based on these low values and the degree of visible swelling, the decision was made to replace the entire pipe section during the forced outage.
The removed spool was submitted to Structural Integrity (SI) for laboratory analysis to further assess the cause of the low hardness and localized swelling. SI made diameter measurements and performed additional hardness tests along the spool length to determine the extent of swelling and “soft” material. Two window sections were also removed to evaluate the microstructure, chemistry, and through-thickness hardness. One window was removed from the bulged area, between 137” and 147”, and the other was removed remote from the bulging near Field Weld #1, between 2” and 6”. The cracked pipe section was specified as ASME SA-335, Grade P91 with 22.0 inch OD and 2.004 inch minimum wall thickness. The design pressure and temperature were reported as 2646 psig and 1050°F, respectively.
Figure 2 presents the measured diameters along the pipe length, showing that the swelling was confined to the straight section of the pipe spool between 110” and 177”. The maximum diameter was at the 140” position and was approximately 12% greater than the non-bulged diameter.
Due to the significant scatter in the field hardness data obtained by others (values ranging from 158 to 209 HB), in-situ hardness testing using a portable device (UCI probe) was first performed on the swelled region to verify the “soft” condition and evaluate its axial extent. The results showed uniformly low hardness (~160 HB) in the swelled region. The region of low hardness extended approximately 10 inches into the upstream bend. The center of the bend exhibited hardness values in the acceptable range for Grade 91 (~200 HB), which correlates with swelling being confined to the downstream end of the spool.
Hardness testing was also performed at both field girth welds to assess the adjacent upstream and downstream spools. The measured hardness values on the adjacent spools were in the acceptable range for Grade 91 piping, indicating the welding and post-weld heat treatment procedures were performed correctly.
After bounding the extents of low surface hardness and swelling, through-thickness hardness testing was performed on the two window sections. The results were consistent with those reported on the pipe surface: the through-thickness hardness in the swelled section ranged from 159 to 165 HB, while the through-thickness hardness in the section remote from the swelling ranged from 200 to 210 HB.
Close-up views of the external and internal surfaces on the window section removed from the swollen area revealed axially-oriented oxide cracking, shown in Figure 3. The oxide cracking was consistent with significant in-service swelling of the pipe and did not extend significantly into the base material. The average thickness of the window section was 1.654 inch, which was below the specified minimum value of 2.004 inch. The minimum thickness measured remote from the bulging near Field Weld #1, which should be representative of the thickness in the straight sections prior to swelling, was 2.184 inch, in excess of the minimum specified value.
Cross sections from the window sections were examined using a metallurgical microscope to assess the microstructure; Figure 4 shows the microstructure from each window section. The microstructure on the swollen section consisted of dispersed alloy carbides in a ferrite matrix, which is not the desired microstructure for Grade 91, but is consistent with the low hardness values and mal-heat treatment. Isolated creep voids were observed across the entire thickness of the section. The microstructure on the window section removed remote from the swelling consisted of the desired tempered martensitic microstructure for Grade 91 piping.
Chemical analysis was performed on a sample from the bulged window section; the results for significant impurity elements copper, nickel, and tin are presented in Table 1 along with the chemical requirements for ASME SA-335 Grade P91 and the more stringent requirements recommended by EPRI in their publicly available report 3002006390 (2015).
The results showed the sample met the chemical requirements of ASME SA-335 Grade P91, but the copper, nickel, and tin contents were greater than the EPRI-recommended values. These higher impurity element concentrations indicate this heat of material likely exhibited lower creep strength and lower damage tolerance (greater creep cavitation susceptibility) than average. However, the mal-heat treatment that caused the degraded hardness results in a reduction of strength well beyond the compositional effects.
The final area evaluated on the spool was the leak associated with the RT plug. Figure 5 shows the crack along the toe of the RT plug weld; the area surrounding the crack was ground, polished, and etched in preparation for metallographic replication. Figure 6 shows two high-magnification views of the crack as seen in the replica, which was located at the toe of the RT plug weld to the main run pipe. The location of cracking was as expected given the swelling of the pipe section. The crack was oxide-filled and consisted of several crack segments linking together; minor creep voids were observed adjacent to the primary crack. The typical microstructure observed in RT plug weld metal was the desired tempered martensite, and the average hardness was 234 HB, which is in the acceptable range.
The fact the low hardness values were confined to the bent pipe spool and were not in adjacent piping upstream and downstream of the field welds indicates the improper processing leading to low hardness and poor creep strength occurred during manufacturing of the bent spool, rather than during field erection. The large amount of swelling and creep voids observed in the window section clearly indicated that the “soft” pipe section was at the end of its useful life. The leak occurred at the RT plug because of a local strain intensification created by the stronger RT plug in the “soft” pipe spool. If the RT plug had not been present, continued swelling would have eventually resulted in rupture of the base metal and a significantly greater release of high temperature steam.
This case study highlights the need for careful quality control and detailed verification of fabrication processes for creep strength enhanced ferritic steels such as Grade 91. If possible, those steps should be performed during the various steps of the fabrication and erection process, or as part of an in-service Grade 91 life management program. In the present case, screening lifetime calculations to identify locations vulnerable to failure if in the “soft” condition would have identified this location as presenting high risk, and subsequent field hardness testing could have been timed to verify if the material was in the “soft” condition. This could have avoided the leak, lost generation time, and significant safety risk.