1. Introduction

Since the late 1980s, 9–12% Cr creep strength–enhanced ferritic (CSEF) steels, especially Grade 91 steel, have been extensively used for fabricating critical boiler components in supercritical power plants [1], [2], [3], [4]. Many in-service steam components (headers and main steam pipes) fabricated by 9Cr steels are close to or already beyond their original anticipated design lives. Therefore, residual lifetimes of these in-service components need to be evaluated to ensure continued, safe operation of power plants [5], [6]. Welds are frequently the weakest link governing the service lifetime of at-risk components, and creep is a primary damage mechanism that must be properly assessed [7], [8]. Nucleation and growth of creep cavities along grain boundaries, especially prior austenite grain boundaries, was the most common damage during service, which led to cracks and failures in 9Cr steels and their welds [9]. Cavity nucleation and growth rates were greatly affected by the service temperatures and stress levels [10]. Premature failures, especially Type IV cracking in heat affected zone (HAZ), have been frequently reported in 9Cr steel welds [11], [12], [13], [14]. Formation of creep cavities was highly nonuniform across the welds, and creep cavities preferentially accumulated in the HAZ. Microstructural degradation in the HAZ, including the significant reduction of precipitation strengthening and extensive matrix grain recovery/growth accelerated cavitation and crack formation [8]. Prior research [11], [15], [16], [17] has consistently shown that localized creep deformation along with local cavitation in the HAZ is the main creep degradation mechanism in 9Cr steel welds. Determining creep rupture criteria of welds by tracking local cavity evolution (size and fraction) in the sub-regions of HAZ is useful for component lifetime managements, but very difficult with experiments. Instead, measuring localized creep deformation in the HAZ is a more feasible approach. Quantifying the strain localization in the HAZ with standardized cross-weld (CW) testing methods is still challenging and often impossible for conventional experimental setups. Many factors, including composition variations among filler metals and base metals, highly nonequilibrium thermal cycles of multi-pass fusion welding processes, and severe service and operation conditions result in a complex set of microstructures across the weld and within the HAZ [18], [19], [20]. Conventional extensometers can only measure nominal creep strain over a gauge section and are thus incapable of providing information relevant to local strain accumulation in the HAZ or its subregions. Recent research [21], [22] shows the nondestructive and noncontact technique of digital image correlation (DIC) provides a potential solution to overcome this technical difficulty. Application of this DIC approach to welds at the component-level may enhance current or future lifetime assessment methodologies.

Precipitation and coarsening of secondary phases (Laves phase Fe2Mo and Z phase Cr(Nb,V)N) that form in-service are considered to be one of the main factors causing loss of creep resistance in 9Cr CSEF steels and their welds [23], [24], [25], [26], [27], [28]. The formation of Laves phase (Fe2Mo) by consuming Mo decreased Mo’s solid solution strengthening, and the coarsened Z phase transformed from fine MX caused a reduced precipitation strengthening, which accelerated creep rate [29], [30]. Meanwhile, a high local stress concentration around these coarse precipitates also induced nucleation of cavities along grain boundaries [15]. Recent studies [31], [32], [33] also show that there are additional metallurgical factors, such as deviations in compositions, impurities, and heat treatments, that can significantly affect the creep performance of welded structures. Furthermore, many early failures could be, in part, linked to the improper control of trace elements and inclusions [7], [34], [35]. Unfortunately, the contents of trace elements and inclusions can vary significantly among 9Cr steels even though the stated composition may meet the ASTM and ASME standard specifications. Insufficient research exists to clarify these complexities and to properly assess the creep performance of welds fabricated with different 9Cr steels, because most studies on 9Cr steel welds investigate a simple combination (pipe–pipe, tube–tube, forge–forge) of the same material (the same heat). However, in-service components must be fabricated by welding at different heats, using different product forms, or both. Therefore, the assumption that “all 9Cr steel is the same” is an overly simplistic view that must be clarified to implement a meaningful life management strategy for welded structures.

In this work, remaining creep lifetime of a 9Cr steel weld taken from a 141,000-hour service-aged outlet header between an SA-182 forge 91 (F91) steel reducer to an SA-335 pipe 91 (P91) steel header was assessed using novel in-situ DIC technique. Meanwhile, multiscale metallurgical analyses were conducted to correlate microstructural features across this service-aged weld before and after the carefully designed creep tests to the measured strain accumulation data obtained using DIC. Insights into creep deformation and rupture mechanisms in 9Cr CSEF steel welds are discussed and supported by the available findings in this manuscript.

2. Materials and experiments

In this study, a 9Cr steel girth weld was extracted from an outlet header component between an SA-182 forge 91 (F91) steel reducer and an SA-335 pipe 91 (P91) steel header (Fig. 1a). This header piece was in use from 1991 to 2015, accumulating a total of 141,000 h of operation and 3300 starts at a nominal steam temperature of 568 °C and a nominal steam pressure of 17 MPa (design conditions of 575 °C and 17 MPa). The outer diameter and wall thickness are 308 mm and 33 mm, respectively. Table 1 summarizes the chemical compositions of the F91 and P91 base metals. Fig. 1b shows a macro-graph of cross section of the entire girth weld, including P91 steel on the left, weld metal (WM) in the center, and F91 steel on the right. A feature cross-weld (CW) creep specimen was designed for creep testing (Fig. 1c). The gauge section has a dimension of 30 mm width × 10 mm thickness × 70 mm length, which is the entire girth weld, including the cap reinforcement on the top and root passes at the bottom.

1. Download : Download high-res image (575KB)

2. Download : Download full-size image

Fig. 1

Table 1. Chemical compositions* of the studied forge 91 steel and pipe 91 steel (Wt.%).

Materials	C	Mn	P	S	Si	Ni	Cr	Mo
Forge 91 (F91)	0.103	0.40	0.009	0.0105	0.242	0.11	8.148	0.89
Pipe 91 (P91)	0.099	0.43	0.021	0.0003	0.262	0.17	9.125	0.98
Materials	V	Nb	Ti	Co	Cu	Al	B	Fe
Forge 91 (F91)	0.200	0.108	0.002	0.012	0.124	0.033	<0.0005	Bal.
Pipe 91 (P91)	0.200	0.077	0.003	0.007	0.032	0.007	<0.0005	Bal.
Materials	N	As	Pb	Sb	Sn	Ti	Zr	W
Forge 91 (F91)	0.0378	0.010	0.005817	0.0013	0.008	0.002	<0.002	<0.002
Pipe 91 (P91)	0.0398	0.004	0.00003	0.0002	0.001	0.003	<0.002	<0.002

*Also analyzed with no measurable amount: Bi, Ca, La, Ta.

Creep tests were performed in an ATS 2330 series lever arm creep testing system. The specimens were tested at 600 °C, 625 °C, and 650 °C with stress levels from 60 to 120 MPa. A DIC camera system was integrated with the creep frame to capture images of the creep specimen at different time intervals during each test. More details of this creep-DIC setup can be found in a previous work [21]. The DIC image acquisition rates are 30 min per frame and 5 min per frame throughout the creep test with an in-house developed code. Postprocessing of the DIC data (images) was conducted with the VIC-2D software to determine creep strain (exx, Lagrange) distribution along the loading direction. The local strain for each subregion was the average strain across the subregion’s width. The subregion width was determined by hardness distribution and optical microstructure analysis. Creep strain curves over the gauge length and the local regions of interests were processed from the DIC data.

Microhardness (Vickers) measurement (mapping and line scan) with a load of 0.5 kgf with a dwell time of 10  s and an interspacing of 0.15 mm was conducted on the CW specimens before and after creep tests. For microstructural analysis, the F91-P91 CW specimens before testing and after testing were polished using a conventional mechanical polishing method. Multiscale microstructural analyses were characterized with a Zeiss AXIO optical microscope and a TESCAN MIRA3 XMH Schottky field emission scanning electron microscope equipped with EDAX energy-dispersive X-ray (EDS) and electron backscatter diffraction (EBSD) system. The EBSD data were postprocessed using OIM Analysis software to obtain orientation, and local misorientation maps. Transmission electron microscope (TEM) and EDS were used to identify precipitates. A particle analysis on inclusions and coarse precipitates was conducted with SEM images using ImageJ software. A cavity analysis on the interrupted creep specimen was also performed with a laser microscope. Thermodynamic modeling of precipitate fractions at the equilibrium condition was performed for the compositions given in Table 1 for F91 and P91 using Thermo-Calc with the TCFE9 database. The equilibrium calculation considered the following elements: C, Si, Mn, Ni, Cr, Mo, V, Nb, S, P, Ti, Co, Cu, Al, B, and Fe. The following phases were considered: α, γ, M23C6, MX, Laves phase, and Z phase.

3. Results

3.1. Preexisting creep damages in the service-aged F91-P91 weld

Fig. 2 shows the hardness distribution near the transition regions between P91 base metal (F91-BM, left), weld metal (WM, middle), and F91 base metal (P91-BM, right). A typical hardness profile across the mid-thickness of the weld is also shown in Fig. 3a. The WM maintains the highest hardness of 230–250 HV0.5, and the hardness of P91-BM and F91-BM are 200 HV0.5 and 180 HV0.5, respectively. Continuous soft zones (blue regions) with a lower hardness are observed near the visible edge of the HAZ in both the F91 and P91. In the soft zone, a reduction in the hardness value to ∼170 HV 0.5 is observed on the F91 side and is wider than the P91 side where the hardness value is ∼180 HV 0.5. SEM images in Fig. 3 show there are still large microstructural variations in the HAZs and BMs after a 16-year service aging. Three regions in the HAZ were identified: the coarse-grained heat affected zone (CGHAZ), the creep-damaged zone (CDZ), which had the most cavities; and the soft zone (SZ), which had the lowest hardness. By tracing the location of coarse precipitates, prior austenite grain boundaries are still visible in the CGHAZ, as shown in Fig. 3b. Fine equiaxed grains (<5 µm) with coarsened precipitates are observed in the CDZ. TEM and EDS analysis in Fig. 4 shows Cr-rich M23C6 carbides, MX-(V,Nb)CN carbonitrides, and Laves phase are present after service. Creep cavities were observed not only in the HAZ but also in the BM on F91 side. The area fractions of cavities in the F91-BM and F91 CDZ are 0.29% and 0.42%, respectively. The CDZ (Fig. 3g) rather than the SZ has the highest density of creep cavities. Globular cavities (1–4 µm) formed along grain boundaries in F91 CDZ. In the F91-BM (Fig. 3i), cavities nucleated and grew along prior austenite grain boundaries and other martensite boundaries. The number densities of cavities in the F91-BM and F91 CDZ are 2.28 × 104 counts/mm2 and 2.58 × 104 counts/mm2, respectively. The average sizes of cavities in the F91-BM and F91 CDZ are 407 nm and 453 nm, respectively. No cavities (>1 µm) were observed on the P91 side. These microstructure features indicate the F91 side experienced a much faster microstructural degradation than the P91 side.

1. Download : Download high-res image (885KB)

2. Download : Download full-size image

Fig. 2

1. Download : Download high-res image (1MB)

2. Download : Download full-size image

Fig. 3

1. Download : Download high-res image (724KB)

2. Download : Download full-size image

Fig. 4

3.2. Creep testing with in-situ DIC

Microstructural characterization results above clearly indicate the onset and evolution of creep damages is different for the F91 side and P91 side. Fig. 5 exhibits a typical example of the heterogeneous distribution of creep strain across the test specimen evaluated at 650 °C and 80 MPa. This strain inhomogeneity existed not only between the HAZs on both sides of the weld but also within each HAZ. The results clearly show that the F91-HAZ experienced more strain accumulation, and within the HAZ, the strain was preferentially localized near the cap and root regions. The highest strain accumulated in the F91 HAZ near the cap region was ∼7% three minutes before rupture occurred after a total test duration of 135.5  h. In the P91 HAZ, three high strain bands were captured by the DIC as well. These bands are related to the complex HAZ microstructures produced by the multi-pass welding process.

1. Download : Download high-res image (882KB)

2. Download : Download full-size image

Fig. 5

Creep curves in each microstructural constituent were extracted from the DIC measurements to further understand the deformation response during the test. Fig. 6 compares creep strain curves with decreasing resistance to deformation: P91-BM < F91-BM < WM < P91-HAZ < F91-HAZ. The strain of P91-BM was only about 0.5% before rupture, which is relatively low compared to 1.2% of the F91-BM. Similarly, the strain in the F91-HAZ was ∼7% immediately before rupture, compared to that of the P91-HAZ at only 2%. These results suggest that creep resistance of the HAZ is proportional to the strength of corresponding base metals in this header weld.

1. Download : Download high-res image (346KB)

2. Download : Download full-size image

Fig. 6

A cavity analysis was conducted on a CW specimen tested at 625 °C and 60 MPa and interrupted immediately before failure was assessed in an attempt to correlate the measured local strain with the actual creep damage (Fig. 7). There exists significantly more damages in the F91-HAZ compared to the P91-HAZ. The onset of micro- and macro-cracking in the F91-HAZ is nonuniform through the wall thickness, and there is extensive damage in the form of a high density of creep voids through the entire HAZ. An image analysis shows the F91-HAZ cap region near the cracks has the highest number density of cavities (1.22 × 104 counts/mm2) with an average size of 3.09 ± 1.98 μm (note: fine cavities below 0.5 um were not counted). During creep testing, the DIC suggests cracks and failure initiated from the F91-HAZ in the cap region of the weld, as shown in Fig. 5m. The F91-HAZ in the cap region exhibits the largest micro-cracks. The results suggest there is a relationship between the local strain and crack formation. It follows that the locations where cracks are first to evolve are those with the highest density of creep voids. Note, however, this study does not provide the necessary detail for an interrupted creep test specimen at a lower strain or life fraction to fully elucidate this point.

1. Download : Download high-res image (856KB)

2. Download : Download full-size image

Fig. 7

Table 2 summarizes the full-size CW creep-DIC testing results for the F91-P91 header welds. All CW specimens failed in the F91-HAZ. The cracks initiated from the cap region (creep stress <120 MPa) or root region (creep stress ≥120 MPa). Fig. 8 compares the nominal creep strain curves over the gauge length and the local creep strain curves of the identified crack initiation regions (cap or root regions). At higher creep temperatures (600–650 °C) and higher stresses (80–120 MPa), the welds failed in <500 h and with a low nominal strain of <3.5% before rupture. The observed creep strain is quite low compared to other lab-scale Grade 91 CW samples without thermal aging, which have generally showed larger strain values at failure. This observation is linked to the higher susceptibility of the F91 heat to damage development and emphasizes the potential influence of creep ductility on the susceptibility to Type IV cracking observed in real components [7] .

Table 2. Summary of the creep-DIC testing results.

Test ID	Specimen type	Temperature (°C)	Stress (MPa)	Rupture life (h)	Fracture location	Crack initiation location
1c	Full size cross-weld	625	60	1843	F91 HAZ	HAZ cap region
2a	Full size cross-weld	650	80	136	F91 HAZ	HAZ cap region
2b	Full size cross-weld	625	100	200	F91 HAZ	HAZ cap region
2c	Full size cross-weld	600	120	295	F91 HAZ	HAZ root region
B	Full size cross-weld	625	120	84	F91 HAZ	HAZ root region

1. Download : Download high-res image (562KB)

2. Download : Download full-size image

Fig. 8

Table 3. Statistical analysis of Laves phase and inclusions in as-received service aged F91-BM and P91-BM.

Particles	Location	Number density [counts/mm2]	Area fraction [%]	Mean size [nm]
Laves phase	F91-BM	60,520	0.64	333 ± 154
	P91-BM	71,646	0.74	329 ± 147
Inclusions	F91-BM	3,928	0.24	889 ± 559
	P91-BM	205	0.03	1125 ± 1011

3.3. Posttest characterization

Fig. 9 shows photographs of the failed CW specimen (test ID: 2b, 625 °C-100 MPa) and optical images of the cross section on F91 side in other fractured specimens. Fig. 9b confirms the brittle nature of the failure including limited necking at the fracture edges (e.g., cap and root regions). The characteristics of low creep ductility and fracture location are consistent with similar observations of Type IV cracking. Optical macrographs show the distance from the fusion line to the fracture edges is not uniform along the thickness direction (from the cap to root) (note the red arrows in Fig. 9d–g). The fracture location shifts toward the F91-BM when increasing the applied creep stress from 60 MPa to 120 MPa. Based on the strain maps, the high strain region (green and red color) is narrow and under 60 MPa and 80 MPa (Fig. 9h and i), whereas the high strain region is wider under 100 MPa and 120 MPa (Fig. 9j and k) The regions with the highest strain shown in the DIC contour maps (Fig. 9h–k) are the crack initiation locations, where the crack initiated from the F91-HAZ cap region in most of the specimens, except the specimen tested at 120 MPa (F91-HAZ root region). In terms of the specific region within the HAZ, the crack initiation locations are the identified CDZ (intercritical HAZ, ICHAZ) under a stress of 60 MPa and 80 MPa. While this location shifted into the over-tempered base metal under a creep stress of 100 MPa and 120 MPa. This observation is consistent with prior works showing the low-ductile Type IV cracking occurred under a low stress level (<100 MPa).

1. Download : Download high-res image (2MB)

2. Download : Download full-size image

Fig. 9

A fractography analysis in Fig. 10 uncovers more characteristics of the creep failures in the F91-HAZ. Overall, the fracture surface is rough, as shown in Fig. 10a. The fracture surfaces are highly nonuniform along the thickness direction. The magnified images from the cap to root show a transition from a brittle intergranular failure to a ductile transgranular failure. The cap region exhibits a low creep ductility with large and deep voids, and individual grains, as shown in Fig. 10c. The root regions shows a typical ductile failure with uniform dimples, as shown in Fig. 10i. The low-magnified image in Fig. 10f shows the mid-thickness region exhibits mixed features of the cap region and the root region. Fig. 10d and 10j show coarse precipitates distributed on the grain surfaces or inside the dimples. An EDS analysis was conducted on the fracture surfaces (Fig. 10d and j) and the results are presented in Figs. 11 and 12. Based on the precipitates identified in Fig. 4, the Cr, Mo, Nb, Al maps represent four different particles, including Cr-rich M23C6, Laves phase (Fe2Mo), NbCN, and Al oxides, respectively. The Cr-rich M23C6 has the largest number fraction and the smallest size among the four. The Al oxides possess the largest size. The precipitates on the grains surface in the cap region seems to be slightly coarser than those inside the dimples in the root regions. All these fractography features confirm that the cap region was the weakest region. The crack initiated in the cap region and propagated towards the root region.

1. Download : Download high-res image (1MB)

2. Download : Download full-size image

Fig. 10

1. Download : Download high-res image (1MB)

2. Download : Download full-size image

Fig. 11

1. Download : Download high-res image (919KB)

2. Download : Download full-size image

Fig. 12

Microhardness distribution across a failed weld was measured in Fig. 13. Fig. 13a details the hardness distribution profile across the entire weld near the cap region. Fig. 13b and c compare hardness distributions near the HAZs on both sides. After creep testing, the WM has the largest hardness reduction, but maintains the highest hardness of 200–230 HV0.5. The F91-BM exhibits a hardness of 165–175 HV0.5. The hardness of the P91-BM is essentially unchanged, and the hardness reduction of the P91-HAZ is also relatively small (from 200 to 190 HV0.5). The soft zone on the P91 side shows a hardness of 170 HV0.5. The softening on the F91 side is more obvious. The fractured F91-HAZ region exhibits a hardness of 150 to 155 HV0.5. The soft zone (blue region) adjacent to the fracture region on the F91 side exhibits the lowest hardness at 140 to 150 HV0.5. Note, the ultimate failure location is not within the soft zone, but near the boundary of the soft zone. The lowest hardness region is associated with the over-tempered region in the HAZ.

1. Download : Download high-res image (970KB)

2. Download : Download full-size image

Fig. 13

Additional information regarding the fracture location on the F91 side is given by the EBSD analysis in Fig. 14. The inverse pole figure in Fig. 14b shows the fracture occurred in a region with fine equiaxed grains, instead of the region adjacent to the F91-BM where tempered martensite is observed. A grain orientation spread map in Fig. 14c presents the distribution of recrystallized grains (blue grains) with lower dislocation densities. The failure evidently did not occur in the soft zone with the largest number of recrystallized grains and the lowest hardness. Fig. 14d and e clearly exhibit globular cavities along the boundaries of equiaxed grains. No elongated grains were observed on the fracture edge. This intergranular creep damage eventually led to the low creep ductility feature of the Type IV cracking.

1. Download : Download high-res image (1MB)

2. Download : Download full-size image

Fig. 14

4. Discussion

The assessment of full-size welds in so-called feature type tests is essential to inform life management and damage evolution in welded structures. Because components are fabricated from large sections including multiple product forms, material suppliers, or both, the complex multi-pass weld configurations, and subsequent heat-specific response to local stress states in the time-dependent regime will introduce large variability, suggesting that local stress concentration can significantly affect the weld life. The header weld was serviced under a pressure load, and the overall main stress states were hoop stress (dominant stress) with axial and radial stresses. During long term service, the non-uniform stress distribution occurred because of the creep strength difference among HAZ, weld metal, and base material. As the HAZs are of less creep resistance than the corresponding base materials, the HAZs creep faster and get mechanically constrained by the strong base metal and weld metal. This results the strong triaxial stresses and lower effective stress in HAZs. These details were observed in the present study as the response in the P91-HAZ and F91-HAZ was markedly different even though these constituents were exposed to the same operating conditions and location in the header from which the assessed girth weld was removed.

The pretest characterization results in Figs. 2 and 3 show the preexisting creep voids are highly nonuniform in this F91-P91 girth weld after 16-year service exposure. Therefore, a feature CW test specimen that included the entirety of the through-thickness weld profile (Fig. 1c) was designed and assessed. The local strain maps in Fig. 9 show a bias of strain accumulation to the cap region where the cracking first initiates before propagating through the HAZ and leading to failure. The preferential creep degradation in the HAZ of subcritically post-weld heat treated welds is a primary concern for premature creep failures in 9Cr steel welds [36]. The in-situ DIC approach introduced in this work demonstrated the ability to quantify and partition the observed strain in a heterogenous welded test specimen (Fig. 5). Furthermore, the observed strain was not uniform in the HAZ on either side of the weld, indicating heat dependency for the strain accumulation. A clear discontinuous strain field was observed with the HAZ, and was clearer in the P91 HAZ, indicating the potential role that nonuniform microstructures in the HAZ created by the multi-pass welding deposition procedures may ultimately have on failure.

The DIC measurements not only provide details about the sequence of events leading to failure but also provided a means to develop creep strain curves for each of the WM, BM, and HAZ constituents (Fig. 6). This detail is helpful to assess local strain accumulation, supports model development and validation, and provides a means to extract a significant amount of information from a single test (as opposed to individual tests with each of the constituents). Fig. 8 shows it is difficult to predict when the weld will fail when the strain curves are averaged over the gauge length because the deformation is most extensive in the relatively narrow HAZ region where cracking and failure occur. The local creep curves show a threshold level of 2% local strain before an acceleration of creep strain accumulation was observed in the F91 HAZ. Such local strain measurements and threshold values could be beneficial to informing maintenance schedules or as input to future methods or means for assessing local strain in welded structures operating at high temperatures. The considerable difference between the local strain measurements and the cross-weld strain from regular lab creep testing could also be valuable to informing component designers for additional considerations to minimize local stress concentration when designing components using the cross-weld creep testing results.

Cavitation in creep failures, especially Type IV cracking in the HAZ, was one of the typical metallurgical characteristics in 9Cr steel welds. Fig. 7 clearly presents a high density of cavities concentrated in a narrow region in the HAZ. Micro-cracks formed by connecting those coarsened cavities in the identified CDZ (intercritical HAZ), as shown in Fig. 14. The observed cracking failure with a low creep ductility resulted from the accumulation of intergranular cavities instead of matrix grain deformation. The weld was loaded and held at a constant tensile load (with the direction perpendicular to the centerline of weld) during testing. The main stress state is tensile stress for the weld coupon. Under tensile creep testing, the non-uniform stress also developed because of the creep strength differences among sub-regions in the weld. It is believed that the CDZ was under high triaxial stresses which was originated from the mechanical constraint effect from the adjacent base metal, CGHAZ, and weld metal [36], [37]. This high stress triaxiality accelerated grain boundary sliding which led to nucleation of cavities [38]. Meanwhile the insufficient precipitation strengthening along grain boundaries in the CDZ due to dissolution and coarsening of precipitates (mainly M23C6) during welding and service cannot inhabit cavity growth. The coarsened precipitates (M23C6, Laves phase) and inclusions also acted as preferential nucleation sites [15]. Coarse particles on the fracture surfaces in Fig. 11, Fig. 12 show their association with cavities. Eventually creep deformation was accommodated by the formation of cavities in the weakest region (CDZ) on the F91 side (Fig. 14).

The F91-HAZ was the most creep damage–susceptible constituent in the assessed girth weld. HAZ properties are closely related to the condition of the parent base metal (e.g., the composition and heat treatment history) [32], [39]. Nucleation and growth of creep cavities are closely associated with the large particles in the matrix, including coarse precipitates (M23C6, Laves phase) and inclusions [31], [40]. To assess why the F91-HAZ was more susceptible to the evolution of damage, an analysis of the M23C6, Laves phase and inclusions was performed in the F91 and P91 base materials. The influence of the composition variation present in Table 1 on the equilibrium phase fraction of precipitates (M23C6, MX, Laves phase, Z phase) was simulated using Thermo-Calc (Fig. 15). The phase fractions show no significant difference in the base metal compositions. For the service temperature of 568 °C, the fractions of M23C6 carbides are 2.20% and 2.18% for P91-BM and F91-BM, respectively. The Laves phase in P91-BM (0.70%) is higher than 0.40% in the F91-BM. The experimental results are consistent with these predicted results. The statistical analysis results of the Laves phase and inclusions in Fig. 16 and Table 3 exhibit a comparable distribution of Laves phase in F91-BM and P91-BM; the mean size for Laves phase is ∼330 nm with an area fraction of 0.64% and 0.74% for F91 and P91, respectively. The analysis of the inclusion density and type, however, reveals a more profound observation. In the F91-BM, there are more inclusions (mainly aluminum oxides and MnS) compared to the P91 BM. The area fractions of inclusions are 0.24% in F91-BM and 0.03% in P91-BM. The inclusions in F91-BM are smaller (889 nm) with a number density of 3926 counts/mm2, and the inclusion size in P91 BM is 1125 nm with a number density of 205/mm2. These finer and higher density of inclusions in the F91-BM increases the susceptibility to the evolution creep damage. Table 1 also shows the concentrations of inclusion-forming elements (S, Al) and trace elements (As, Cu, Pb, Sb, and Sn) in the F91-BM are higher than those in the P91-BM. Although this study did not pursue the potential relationship between inclusions, trace elements, and damage, recently completed or active studies are revealing similar findings (e.g., that the poorest performing heats in a population have significantly higher densities of inclusions and concentration of trace elements) [32], [40]. The selection of “purer” 9Cr steels with low densities of inclusions and controlled levels of deleterious trace elements is necessary to reduce the onset of time-dependent damage in at-risk components.

1. Download : Download high-res image (394KB)

2. Download : Download full-size image

Fig. 15

1. Download : Download high-res image (1MB)

2. Download : Download full-size image

Fig. 16

5. Conclusions

In this work, a novel approach including feature test specimens extracted from an ex-service P91-F91 girth weld were monitored with an integrated in-situ DIC system to assess time-dependent response of the primary microstructural constituents, and creep damage and the onset of failure in a representative 9Cr CSEF steel weld. Three main findings are summarized here:

• (a)

  Nonuniform preexisting creep voids are observed in the ex-service P91-F91 girth weld removed from a superheated outlet header. Large hardness differences were observed among the WM, P91-BM, and F91-BM. The soft zones with a lower hardness adjacent to the HAZ on the F91 side of the weld is softer and wider. Globular cavities formed on the HAZ and BM on F91 side, but no cavities were observed on the P91 side. The CDZ with the most cavities is not associated with the soft zone but with a region with fully equiaxed grains about 0.5 mm from the soft zone.

  
  

• (b)

  A design for a feature cross-weld creep specimen including the entire weld thickness and profile a full wall thickness is successfully evaluated for the first time using an in-situ DIC technique. Creep deformation of the weldment microstructural constituents follows the order: P91-BM < F91-BM < WM < P91-HAZ < F91-HAZ. The HAZ near the cap region on F91 side was the most susceptible region to crack initiation, ultimately leading to a rapid failure through a uniform field of creep damage. The DIC measurements confirm the low creep ductility nature of Type IV cracking including a low observed nominal strain < 3.5% while the local strain value in the F91-HAZ was < 15%. The onset of accelerated damage progression was obvious once the local strain reached a threshold strain of 2%, suggesting the necessity of cautious considerations about local stress concentration in remaining life assessment and new component design.

  
  

• (c)

  Creep failures of this weld initiated in the F91-HAZ near the cap region and very rapidly propagated to the root. The fracture did not occur in the middle of the identified soft zone, which has the highest number of recrystallized grains and the lowest hardness. Most cavities are associated with large particles, especially inclusions and Laves phase. The higher number of inclusions in the F91-BM is believed to significantly contribute to the lower creep resistance of the F91 HAZ as the number density and size of Laves phase and M23C6 were markedly similar for the P91 and F91 heats.