BY S. SHI, J. C. LIPPOLD, AND J. RAMIREZ

Significant differences in hot ductility behavior and repair weldability wereobserved among Nb-bearing stainless steel castings after exposure to

similar service conditions

Hot Ductility Behavior and RepairI Weldability of Service-Aged, Heat-ResistantI Stainless Steel CastingslII

i

Materials and ExperimentalProcedures

Materials. Heat-resistant HP-Nb mod-ified (for simplicity, they are referred tohere as HP-Nb alloys) and 20-32Nb alloysare the most commonly used cast high-temperature furnace tube alloys. Thespecification of HP-Nb alloys is describedin ASTM A297 (Ref. 6); and the 20-32Nballoy is covered under ASTMA351/A351M-05 (Ref. 7), but are usuallyidentified by their trade names such asKHR32C (Ref. 8) and CR32W (Ref. 9).The compositions of the alloys evaluatedin this study are provided in Table 1. Thenumber of years of service exposure at ap-proximately 815°C (1500°F) is also indi-cated in the table.

Characterization. Microstructure

Fig. 3, cracking has been observed duringrepair welding. One of the objectives ofthis work was to develop a fundamentalunderstanding of the repair weldability ofthe three materials provided.

Microstructure evolution during serv-ice exposure and simulated thermal expo-sure during repair are critical to under-standing the hot-ductility behaviordiscrepancies and resultant repair weld-ability between the two types of alloys. Mi-crostructure evolution in these materialsfrom the as-cast condition to the service-exposed condition is described elsewhere(Ref. 5). It was found that the atomic ratioof Nb to C is a key factor in determiningthe type of carbides formed during serviceexposure. The microstructure evolutionprocess is summarized in Table 3 of Ref. 5.

In this investigation, the Gleeble@, aprogrammable thermal-mechanical simu-lator, was used to study the hot ductilitybehavior of service-exposed HP-Nb modi-fied and 20-32Nb alloys during simulatedrepair welding thermal cycles. The impli-cations on repair weldabiIity arediscussed.

-_/

Hot Ductility TestCast Stainless SteelsHeat-Resistant Stainless SteelsWeldabilityNb-Bearing Stainless SteelsGleeble@Repair Weldability

KEYWORDS

brittlement of these heat-resistant cast-ings is of great practical concern in thepower-generation, refinery, and petro-chemical industries. The materials studiedwere provided by two independent petro-chemical companies. The as-received HP-45Nb has experienced fracture duringservice - Fig. 1. For the 20-32Nb alloy, Ii-quation cracking was immediately ob-served in the heat-affected zone (HAZ)during repair welding - Fig. 2. Difficul-ties in repair of service-exposed 20-32Nb alloy have been re-ported in other refineries. As shown in

S. SHI (shu.shi@shell.com) is welding and ma-terials specialist, Shell Global Solutions, Hous-ton, Tex.J C. LIPPOLD (lippold.l@osu.edu)isprofessOl; Welding Engineering Program, TheOhio State University, Columbus, Ohio. J.RAMIREZ is principal engineer,Edison WeldingInstitute, Columbus, Ohio.

Heat-resistant, cast stainless steelssuch as Alloy HP-Nb modified (ASTMA297) and Alloy 20-32Nb (ASTM A351/A351M-05) are used in applications re-quiring good corrosion resistance andmoderate strength at temperatures up to1l00°C (2012°F) (Ref. 1). At these tem-peratures the cast microstructure willtransform with time leading to the forma-tion of carbides and other compounds,

Isuch as nickel silicide. The loss of repairweldability after service exposure in these, heat-resistant alloys, including HP-Nb

/

' modified and 20-32Nb, has been related to. the loss of ductility due to the formation ofM23C6 and nickel silicide (Refs. 2-4).Cracking during shut-down or repairwelding due to the service-induced em-

ABSTRACT

The loss of repair weldability after service exposure in heat-resistant alloys hasbeen related to the loss of ductility due to the formation of carbides and other com-

~. J I". pounds, such as nickel silicide. The hot ductility behavior of three service-aged,

Fig. I -As-received service-exposed HP45Nb pipe section that mptured.

Fig. 3 - Cracking observed dl/ring repair welding of the service.exposed 20-32Nb header assembly. (Photo COl/rtesy of Jorge Penso &Hearl Mead.)

tests may be any temperature below thebulk melting temperature of the materialbeing studied. However, T isnormally cho-sen based on the tested NST. In this investi-gation, a temperature between NST andNST-1DoCwas used as Tp for on-coolingtests.

Standard hot ductility samples wereused in this investigation, as shown in Fig.

heating to a peak tempera-ture (Tp) between the nilductility temperature(NDT) and nil strengthtemperature (NST). Duc-tility is normally measuredas the percentage of reduc-tion of area (ROA) at aspecific temperature. NDTand NST are two importantpoints on the on-heatingductility curve. The ductil-ity recovery temperature(DRT), the temperature atwhich some ductility ismeasured during coolingfrom TP' is an importantparameter during the on-cooling cycle.

For hot ductility testing,a test temperature andstroke (extension rate) arcpre-programmed as shown schematically inFig. 4. For on-heating tests, specimens arefirst heated up to the programmed temper-ature and held for 0.5 s to stabilize the tem-perature in the specimen. A stroke is thenapplied at the rate of 25 mm/s to pull thespecimen to failure. For on-cooling tests, apeak temperature (Tp) needs to be deter-mined. The Tp employed in hot ductility

Table 1 - Chemical Composition (wt-%) of Alloys Studied

ID HP45Nb HP50Nb 20.32Nb 20.32Nb 20-32Nbwt-% 9 year 12 year 3 year 7 year 15 year

C 0.41 0.38 0.09 0.1 0.1Mn 1.04 1.22 1.01 0.96 1.13P 0.022 0.031 0.013 0.014 0.013S 0.007 0.016 0.007 0.006 0.006Si 1.15 1.56 0.91 0.87 0.95Cu 0.026 0.1 0.05 0.05 0.05Mo 0.026 0.085 0.01 0.02 0.02Ni 33.55 33.67 33.6 33.4 33Cr 25.58 25.11 19.9 19.9 20.2Nb 0.9 1.37 1.36 1.33 1.19Fe Bal Bal Bal Bal BalI. Philips is a trademark of Royal Philips Elec-

tronics N V, The Netherlands. FEI and Sirion aretrademarks of FEI Co., Hillsboro-Oregon, U.S.A.

characterization was performed on bothservice-exposed (EX) and simulated as-cast alloys (CA). Due to the difficulty inobtaining unexposed, as-cast material ofsimilar composition, the EX alloys wereremelted using a button melting apparatusto simulate the original as-cast mi-crostructure. The button melting appara-tus uses a tungsten torch with argonshielding to produce small cast samples.

The methods used for microstructurecharacterization and fractographic analy-sis included optical microscopy (OM) andscanning electron microscopy (SEM).SEM analysis was conducted in both thesecondary electron (SE) and backscat-tered electron (BSE) modes. Compositionanalysis and line scan analysis were con-ducted using both Philips XL-3D ESEMFEG and FEI Sirion FEGI microscopesequipped with X-ray energy-dispersivespectroscopy (XEDS). With the combina-tion of SE, BSE, and EDS analysis, it ispossible to make reasonable estimates ofthe precipitates and intermetallic phasesthat are present.

Hot Ductility Testing. Hot ductility be-havior was studied using the Gleeble@3800, which has been used extensively forstudying the wcldability of a wide varietyof materials (Refs. 10-19). Hot ductilitytests arc essentially high-temperature ten-sile tests, but are conducted both "on-heating" and "on-cooling." The test tech-nique can provide the tensile properties ofthe HAZ microstructure during the weldthermal cycle. Therefore, it is consideredan effective method to study the nature ofHAZ cracking that occurs during weldingor subsequent postweld processing (Refs.19-22).

In this study, the hot ductility of each ofthe materials was determined by both on-heating and on-cooling tests, where theon-cooling behavior was determined after

WELDING JOURNAL fill

Programmedtest temperature

I'On-heating

II

tProgrammed 1(1)test temperaturel~

, On-cooling 1II>,I

Time

Fig. 4 - Schematic of on-heating and on-cooling hot ductility test program. Fig. 5 - Specimen dimensions of %-in. round bars with reduced gauge sectionused in hot ductility tests of HP alloys.

used to quantify the degree of liquationcracking susceptibility as this is the tem-perature range in which continuous liquidnetworks are present in the microstruc-ture (Refs. 23-26). The larger the NST-DRT temperature range, the .greater is theliquation cracking susceptibility. The im-portant temperatures determined by hotductility testing are summarized in Table3. The NDT for the solutionized 20-32Nballoy was not obtained due to the limitedsupply of material.As Table 3 implies, the service-exposed

HP alloys are relatively resistant to HAZliquation cracking as the (NST-DRT)range is less than 20°e. In contrast, theservice-exposed 20-32Nb alloys had a(NST-DRT) ranging from 86° to 209°C, in-dicating a greater susceptibility to HAZ li-quation cracking. Solution heat treatmentat I 150°Cfor 6 h improved the overall hotductility (Fig. 6), but increased suscepti-bility to HAZ liquation cracking as theNST-DRT increased from 86° to 260°C forthe IS-year service-exposed 20-32Nballoy.The 20-32Nb alloys have lower carbon

content (-0.1 wt-%), and a lower fractionof embrittling phases, than the higher car-bon HP alloys (-0.4 wt-% C). For theservice-exposed condition, the embrittlingphases have been reported as Ni-Nb sili-cide and Cr-rich, M23C6 for both types ofalloys (Refs. 3, 5). Based on carbon con-tent, one might expect that the 20-32Nb al-loys would have better ductility than theHP alloys. The hot ductility tests, however,showed results to the contrary. As shownin Fig. 7, the on-cooling ductility for the20-32Nb alloy from 1000° down to 600°Cis less than 10%, while the ductility of theHP-Nb alloys in this same temperaturerange is between 10 and 20%.The low on-cooling ductility observed

in service-aged 20-32Nb alloys is related tothe microstructure evolution during theservice exposure and subsequent simu-lated on-heating and on-cooling thermalcycles. The microstructure during serviceexposure is discussed in detail elsewhere(Ref. 5). The microstructure changes dur-

ture oxidation, testswere run in an argonatmosphere. Thechamber was evacu-ated and backpurgedtwice with argon toprevent sample oxi-dation during testing.

To study materialsusceptibility to bothductility dip crackingand liquation crack-ing, standard hotductility tests wereperformed over thetemperature rangefrom 500°C to theNST.The testing con-ditions are listed inTable 2.

1300

Value

100°C/s0.5 s0.05 s25 mm/sFree cooling (50 to 25°C/s)

1200

In this study, the temperature differen-tial between NST and DRT (NST-DRT) is

Gleeble Hot Ductility Test Results

Results and Discussion

1100800 900 1000Temperature [.C]

700600

• SA- heating• SA-cooling

70

~60

:50

~40o5301;20::J

"tJ

~ 10

o500

Fig. 6 - Hot ductility clllves of soilitionized 20-32Nb alloy.

5. Since some of the single thermal cyclehot ductility tests were conducted at rela-tively low temperatures, specimens with areduced section at the mid-span were usedto ensure that the sample would fail at theprogrammed temperature and at the pre-ferred location. To prevent contaminationof the fracture surface by high-tempera-

Table 2 - Parameters Used in Hot Ductility Tests

Heating rateHolding time at test temperatureHolding time at peak temperatureApplied stroke rateCooling rate for peak temperature

Parameters

Service-exposed HP alloys 20-32Nb alloys

HP45Nb HP50-Nb 3 year 7 year 15year Solutionized

1240 1170 1200 1210 1190 12501250 1177 1209 1218 1196 12601236 1171 1159 1162 11511232 1163 1088 1009 1110 100018 14 121 209 86 260

• OCTOBER 2010, VOL. 89

able 3 - Important Temperatures eC) Obtained from the Gleeble Hot Ductility Tests

:em••••c-zC):Dmenm>:Do:I:

B

1100"'.11001000.50.

HOIDuctility CUrve. of Servlce-£xposed 2G-32Nb Alloy.

" - .3yOn-H •• tf,.so _.]v On-coollrw

____ .•. 7yOtWt •• tlrc ti:: =-~~:~:~:::. ::::::::::::::::::::~--~Z~~:~:-::::::::::::::li - - - • 15yOn~ooUrc ~ I I'

I:=:-~~.;--::i~~;-:;~~::=L:::

BHot Ductlvlty Curves of Service-Exposed HP.NbAlbys

Temp.ratur. rCl

.... ""

-. HP45NbOn--H@atirc

so -==~!1.~~~~.~~~.C:~l!l__ _ _......... ..~•••• __.•. HPSONb On-Heati •.••.•.•• -HPSONbOn-Coollrc

..A

Fig. 7 - Hot ductility cUlVes of selVice-e.xposed alloys: A - HP-Nb; B - 20-32Nb.

Fig. 9 - Fracture profile of HP45Nb Gleeble sample tested at 900°C on heat-ing:A - Overall view; B - microcracking in Cr-rich particles; C - EDS linescan at dark phase showed a high Cr concentration; D - EDS line scan atbright phase showed a high Nb concentration.

J:o[[

Fig. II - Fracture profile of solutionized 20-32Nb sample tested at 900°C on heating: A - De-fonned dendritic stmcture; B - EDS line scan result.

Fig. 10 - Fractllre surface of service-exposed 20-32Nb sam-ple tested at 90(J'C on heating: A - Fractllre mode viewedat low magnification; B - EDS line scan results of theMZ3C6 and Ni-Nb silicide.

=Emr-C-ZG)JJmenm»JJo:I:

cause of variations in ductility and fracturestress in these materials, hot ductility sam-ples tested at 900° and 1l00°C on both on-heating and on-cooling were subjected tofractographic analysis in the SEM. EDSspot and line scan techniques were used ex-tensively to analyze the microconstituentson the fracture surface or in the mi-crostructure. All the metallographicallypolished samples were examined in the un-etched condition. The backscattered elec-tron (BSE) detector in the SEM was usedto provide phase contrast.The two service-exposed HP alloys

under investigation showed similar resultsregarding the ductility and fracture stress.The HP45Nb alloy was chosen for thefractographic analysis. The hot ductilityresults for the service-exposed 20-32Nb al-loys arc essentially the same regardless ofthe service exposure times.On-heating at 900°C. When tested at

900°C on-heating, both Cr-rich and Nb-rich particles were found in the service-exposed HP-Nb and 20-32Nb alloys (Figs.9, 10). In the HP alloy, evidence of trans-formation from Ni-Nb silicide to NbC wasalso observed. In contrast, no such evi-dence was found in the service-exposed20-32Nb alloys. This may be due to thehigher carbon concentration in the HP-Nballoys, which promoted a higher fractionof M23C6 formation during service. Whenthe alloy was heated to 900°C, M23C6started to dissolve, putting carbon back

.OCTOBER 2010, VOL. 89

into solution. The free carbon in solutionpromoted the decomposition of Ni-Nbsilicide due to a high affinity between nio-bium and carbon. Because the transfor-mation requires carbon diffusion from dis-solved M23C6 to Ni-Nb silicide, thetransformation is limited under fast heat-ing conditions. In the 20-32Nb alloy, thistype of reaction is further retarded be-cause of a relatively small amount ofM23C6 and a large dendrite size. The re-sult is that the carbon content is not suffi-cient and/or the diffusion distances are notshort enough to trigger the transformationfrom Ni-Nb silicide to NbC at a tempera-ture of 900°C.Only Nb-carbides existed in the solu-

tionized 20-32Nb alloy - Fig. 11. Smallcracks were associated with particles of theservice-exposed alloys, and large cavitiesformed at the triple-point grain boundaryjunctions in the solutionized alloy.On-heating at llOO°C. Nb-carbides

were observed on the fracture surface ofall samples when tested at 1100°Con heat-ing. In contrast, no Nb-carbides werefound in the service-exposed 20-32Nballoy when tested at 900°C on heating.This indicated that the transformationfrom Ni-Nb silicide to NbC occurred at ahigher temperature in the 20-32Nb alloyas compared to the HP-Nb alloy. In theservice-exposed HP alloy, the M23C6 car-bide and Ni-Nb silicide no longer existedat the fracture area. In contrast, M23C6

and Ni-Nb silicide were found in the serv-ice-exposed 20-32Nb alloy. NbC is theonly phase that was present in the solu-tionized 20-32Nb alloy.On-cooling at 1100°C.As compared to

samples tested at 1100°C on heating, nosignificant changes regarding the type andmorphology of constituents were found atthe fracture area of the service-exposedHP alloy and solutionized 20-32Nb alloywhen samples were tested at 1100°C oncooling. For the service-exposed 20-32Nballoy, apparent liquation was observedaround the Nb-rich particles (Fig. 12),which can be fully transformed NbC or apartially transformed component from Ni-Nb silicide. The high concentration of sil-icon around the particle (Fig. 13) in-creases the susceptibility to liquation.Once the liquation formed continuouslyaround the dendrite boundary, a signifi-cant loss in on-cooling ductility of the alloywould be expected.On-cooling at 900°C.At the on-cooling

test temperature of 900°C, transformationfrom Ni-Nb silicide to Nb-carbides was ob-served in both service-exposed HP-Nb alloyand 20-32Nb alloy. Chromium-rich parti-cles were observed in the HP-Nb alloy, butnot in any of the other materials. Recrystal-lization occurred in the service-exposed 20-32Nb alloy - Fig. 14. No liquation wasfound in the HP alloy but liquation was ev-ident in the service-exposed 20-32Nb alloy.Niobium-carbides were the only constituent

Fig. 13 - EDS results of the particles presellted at the fracllIre area of 20-32Nb sample tested at I IO(f'C all coolillg: A - EDS spot Scali result from a square-shaped particle; B - EDS lille scali results.

present in the solutionized alloy.

Transformation from Ni-NbSilicide to NbC

The transformation behavior was stud-ied using the SEMIEDS spot analysis andline scan technique. The spot analysis re-sults revealed the particle (Fig. lSA) isrich in Si, Nb, and Ni, and can be charac-terized as Ni-Nb silicide. The dark lines inthe pictures represent the position thathas been analyzed due to "charging" ofthesample by the electron beam. Nickel con-tent has dropped to approximately thenominal composition. No carbon was de-tected in this particle. This is because theCr-rich, M23C6 is of relatively low concen-tration in the alloy, and at this stage the

concentration of carbon from the dissolu-tion of the MZ3C6 is not high enough to bedetected by the SEM. This signified theinitial stage of the transformation fromNi-Nb silicide to Nb-carbide.

A further transformation can be illus-trated with Fig. lSB. Peaks in carbon andsilicon were observed at the edge of thewhite particle. Within the particle, the Niconcentration dropped well below that ofthe matrix, but the Nb concentration stillremained high. It is most likely that theformation of NbC would start at the edgeof the original Ni-Nb silicide due to a highconcentration of carbon. The observedtransformation in Fig. ISB presents thesecond stage of the transformation fromNi-Nb silicide to NbC. In Fig. ISC, parti-cles on the left side of the picture showedenrichment in carbon and niobium and

depletion in Ni, Cr, and Fe, representing acomplete transformation to NbC. Thenewly formed Nb-carbides are muchsmaller than the Ni-Nb silicide. It is likelythat a Ni-Nb silicide particle would breakdown into several NbC particles as a con-sequence of the transformation. A high Siconcentration was also expected aroundthe transformed NbC, and is confirmed inFig. lSC. As highlighted by a red circle, apeak in silicon was observed at the regionbetween two NbC particles.

The phenomenon involving diffusion ofSi and Ni to NbC during aging was previ-ously studied by Patchett (Ref. 3). Thetransformation from Ni-Nb silicide to NbCduring a simulated thermal cycle is the re-verse reaction from the aging process as thethermal-dynamic condition favors the for-mation of NbC at higher temperature.

___ W_EL~NG JOURNAL ill

Fig. 14 - Fracture profile of service-exposed 20-32Nb sample tested at 900°C on cooling: A -lnterdendritic fracture profile; B - recrystallization at dendritesalong the fracture path.

Fig. 15 - EDS result of the process of transformation from Ni-Nb silicide to NbC carbide: A-Ni-Nb silicide; B - the transient component from Ni-Nb silicide; C - NbC.

A

:Emr-C-ZG):IImenm Element Wt% At%» Si 7.8 16.06:II Nb

24.93 15.51Cr 9.65 10.73

0 Fe 19.17 19.85::I: Ni 38.44 37.85Total 100 100

Repair Weldability

The service-exposed HP-Nb alloys ex-hibited good metallurgical stability, frac-ture stress, and on-cooling ductility, whichcould be related to acceptable repair weld-ability of this alloy. Reasonable repairweldability of service-aged HP-Nb alloy isalso reported elsewhere (Ref. 27). In con-trast, service-exposed 20-32Nb alloysshowed severe susceptibility to liquation

cracking and significant loss in on-coolingductility and fracture stress. The suscepti-bility to Iiquation cracking was related tothe combined effect of a high silicon con-centration at the dendrite boundaries re-sulting from the dissolution of Ni-Nb sili-cide, and a NbC constitutional liquationmechanism. The loss in on-cooling ductil-ity that resulted from these two mecha-nisms persisted from the peak HAZ tem-peratures of nearly l300°C to below

900°e. After solution annealing at 1ISO°Cfor 6 h, the on-cooling hot ductility and thefracture stress of the 20-32Nb alloywas re-stored, but with an increase in liquationcracking susceptibility at high tempera-tures as the (NST-DRT) range increasedto 260°e. This indicates that the extent ofthe HAZ in the 20-32Nb alloy exposed tothe (NST-DRT) range must be minimizedby increasing the temperature gradient inthe HAZ during repair welding. This is

• OCTOBER 2010, VOL. 89

probably best achieved with low heat inputpractice (Ref. 28).

Summary

This paper presents the results of hotductility testing of heat-resistant austeniticstainless steels, HP-Nb and 20-32Nb, thathad experienced extended service exposureat 815°C(15000P). Detailed metallographicand fractographic analyseswere performedon hot ductility samples tested at 900° and1100°C,both on-heating to the NDT tem-perature and on-cooling from a tempera-ture slightlybelow the NST.

The hot ductility tests revealed that theHP-Nb alloyshad much better ductility andhigher fracture stress than the 20-32Nb al-loys over the entire range of test tempera-tures. No significant loss in either ductilityor fracture stress occurred in the HP-Nb al-loysduring the on-cooling tests. The HP-Nballoys also exhibited a narrow range be-tween NST and DRT (less than 20°C), indi-cating good resistance to liquation cracking.In contrast, a significant degradation inboth ductility and fracture stress was ob-served in the service-exposed 20-32Nb al-loys during the on-cooling tests. The serv-ice-exposed 20-32Nb alloys are expected tohave a high susceptibility to liquation crack-ing resulting from a large range betwe~nNST and DRT (up to 200°C). Recovery IIIoverall ductility and fracture stress was ob-served only when the service-exposed 20-32Nb alloywas resolutionized, but this wasaccompanied by increased susceptibility toliquation cracking.

Practographic analysis revealed thatCr-rich and Nb-rich particles coexisted inthe fracture region of all the service-exposed HP45-Nb alloy specimens. Evi-dence of transformation from Ni-Nb sili-cide to NbC was also observed. No evi-dence of liquation was observed in theHP45-Nb samples. In contrast, Cr-richand Nb-rich phases coexisted only in theon-heating specimens of the service-exposed 20-32Nb alloywith only a Nb-richphase present in the on-co~ling sampl~s.This indicated that the CHiCh phase diS-solved completely in the service-exposed20-32Nb alloy at on-heating test tempera-tures above 1100°C.A significant amountof liquation was observed in the on-cool-ing 20-32Nb samples, which was a directresult of the high concentration of siliconin the microstructure. The high concen-tration of silicon was the by-product of thetransformation from Ni-Nb silicide toNbC, which decreased the local meltingtemperature significantly, leading to lowon-cooling ductility by liquation at tem-peratures down to 1100°