microstructure model for the heat-affected zone of x80...

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=ymst20

Download by: [The University of British Columbia] Date: 04 March 2016, At: 09:25

Materials Science and Technology

ISSN: 0267-0836 (Print) 1743-2847 (Online) Journal homepage: http://www.tandfonline.com/loi/ymst20

Microstructure model for the heat-affected zoneof X80 linepipe steel

T. Garcin, M. Militzer, W. J. Poole & L. Collins

To cite this article: T. Garcin, M. Militzer, W. J. Poole & L. Collins (2016): Microstructure modelfor the heat-affected zone of X80 linepipe steel, Materials Science and Technology

To link to this article: http://dx.doi.org/10.1080/02670836.2016.1142705

Published online: 04 Mar 2016.

Submit your article to this journal

View related articles

View Crossmark data

http://www.tandfonline.com/action/journalInformation?journalCode=ymst20

http://www.tandfonline.com/loi/ymst20

http://dx.doi.org/10.1080/02670836.2016.1142705

http://www.tandfonline.com/action/authorSubmission?journalCode=ymst20&page=instructions

http://www.tandfonline.com/action/authorSubmission?journalCode=ymst20&page=instructions

http://www.tandfonline.com/doi/mlt/10.1080/02670836.2016.1142705

http://www.tandfonline.com/doi/mlt/10.1080/02670836.2016.1142705

http://crossmark.crossref.org/dialog/?doi=10.1080/02670836.2016.1142705&domain=pdf&date_stamp=2016-03-04

http://crossmark.crossref.org/dialog/?doi=10.1080/02670836.2016.1142705&domain=pdf&date_stamp=2016-03-04

Microstructure model for the heat-affectedzone of X80 linepipe steelT. Garcin1, M. Militzer∗1, W. J. Poole1 and L. Collins2

An important aspect of the integrity of oil and gas pipelines is the heat-affected zone (HAZ) of girthwelds where the microstructure of the as-hot rolled steel is altered with potentially adverse effectson the HAZ properties. Therefore, it is critical to evaluate the HAZ microstructure for differentwelding scenarios. Here, an integrated microstructure evolution model is proposed and appliedto the HAZ of an X80 linepipe steel. The model considers dissolution of Nb-rich precipitates,austenite grain growth and austenite decomposition into ferrite and bainite. Microstructure mapsshowing the fraction of transformation products as a function of distance from the fusion line areobtained and used to compare the effect of different welding procedures on the HAZmicrostructure.Keywords: Linepipe steel, Heat-affected zone, Microstructure modelling, Precipitate dissolution, Austenite grain growth, Austenite decomposition, Ferrite,Bainite, Martensite/austenite

Nomenclaturea thermal diffusivitybi parameters for bainite start modelCv volumetric heat capacity for ironDC

0 pre-exponential factor for the carbon diffusiv-ity in austenite

DNb0 pre-exponential factor for the niobium diffusiv-

ity in austeniteDC carbon diffusivity in austeniteDNb niobium diffusivity in austenitedg mean austenite grain diameterfNbCN(L) volume fraction of large Nb(CN) precipitatesfNbCN(S) volume fraction of small Nb(CN) precipitatesf 0NbCN initial volume of Nb(CN) precipitates in the as-

received material (small and large precipitates)f 0TiN initial volume fraction of TiN precipitatesFF,B true fraction of ferrite (F) and bainite (B)FM/A fraction of martensite/austenite constituentFUB fraction of upper bainiteFLB fraction of lower bainiteFF,BN normalised fraction of ferrite (F) and bainite

(B)FUB,LBN normalised fraction of upper bainite (UB) and

lower bainite (LB)kB Boltzmann constantKNbCN temperature-dependent solubility product for

Nb(CN) precipitatesMg mobility of the austenite grain boundaryMg

0 pre-exponential factor for the mobility of aus-tenite grain boundary

n JMAK exponentq heat inputQg activation energy for the austenite grain

boundary mobility

QCD activation energy for the carbon diffusivity in

austeniteQNb

D activation energy for the niobium diffusivity inaustenite

r distance from centre linerNbCN(L) mean radius of large (L) Nb(CN) precipitatesrNbCN(S) mean radius of small (S) Nb(CN) precipitatesrTiN mean radius of TiN precipitatesR gas constantRf radius of the growing ferrite nucleiT absolute temperatureT0 pre-heat temperatureTB0 nucleation temperature for bainite

TF0 nucleation temperature for ferrite

Tpeak peak temperatureTBS bainite start transformation temperature

TFS ferrite start transformation temperature

v weld speedvAtFe mean atomic volume for ironvAtNbCN mean atomic volume for Nb(CN) precipitatesX ∗

C critical atomic fraction of carbon at the auste-nite grain boundary at the ferrite start trans-formation temperature

XaC equilibrium atomic fraction of carbon in ferrite

XgC equilibrium atomic fraction of carbon in

austeniteXM

C+N solute atomic fraction of carbon and nitrogenin the matrix

XNC nominal atomic fraction of carbon in the steel

XNC+N nominal atomic fraction of carbon in the

matrix plus the excess nitrogen that is not pre-cipitated in TiN

X 0Nb initial atomic fraction of niobium in the matrix

X INb equilibrium atomic fraction of Nb at the pre-

cipitate/matrix interfaceXM

Nb solute atomic fraction of niobium in the matrixXN

Nb nominal atomic fraction of niobium in the steelXP

Nb atomic fraction of the niobium in the Nb(CN)precipitate

1Centre for Metallurgical Process Engineering, The University of BritishColumbia, 309-6350 Stores Rd., Vancouver, BC, Canada V6T 1Z42Evraz, Inc. NA, Regina, SK, Canada S4P 3C7

∗Corresponding author, email [email protected]

© 2016 Institute of Materials, Minerals and MiningReceived 15 October 2015; accepted 12 January 2016DOI 10.1080/02670836.2016.1142705 Materials Science and Technology 2016 1

Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

mailto:[email protected]

Greek symbols

a geometrical constant for the grain curvatureb geometrical constant for particle pinning of

the grain boundarybij JMAK rate parameters for austenite

decomposition1 intensity of the interaction of solute Nb with

the austenite/ferrite interfacew instantaneous cooling ratek thermal conductivity for ironli parameters for ferrite start modelsB austenite grain boundary energysP matrix precipitate interfacial energy

IntroductionThere is currently a need to build new or replace existingnatural gas and oil pipelines in North America in order todistribute these resources in a safe and cost-effective man-ner from northern locations to market. The trend in theindustry is to use pipelines of larger diameter and/orincrease the operating pressure of the pipeline. Thisleads to using higher strength linepipe steel grades toavoid thicker wall dimensions, such as X80 or evenX100 linepipe grades with a minimum yield stress of550 and 690 MPa, respectively, for new pipeline construc-tion. In addition to the higher operating pressures, mini-misation of construction costs is a prime concern. Usinghigher strength steels translates into less material beingrequired thereby lowering material and transportationcost. Lighter wall thicknesses may also require fewerweld passes thereby improving construction efficiency.Further, pipeline integrity hinges to a significant degreeon having high-efficiency girth welds, that is, the in-fieldwelds of pipeline segments. Thus, advanced welding pro-cedures are being considered such as tandem-wire, dual-torch and laser-hybrid welding. An important concernhere is the heat-affected zone (HAZ) in which mechanicalproperties are modified compared to the base metal. Inparticular, the austenite (γ) – ferrite (α) transformationassumes a critical role in determining HAZ microstruc-tures and resulting mechanical properties. It is criticalfor future pipeline projects, which incorporate new designapproaches and welding procedures that fundamentalmicrostructure-property knowledge is advanced for theHAZ.In the HAZ, austenite formation and austenite grain

growth occur during rapid heating as the welding torchpasses. Experimental measurements indicate that theheating rate is >1000°C s–1 under typical welding con-ditions. An extensive review byMishra and DebRoy sum-marises non-isothermal austenite formation and graingrowth behaviour with significant spatial and temporalvariations that are relevant for the HAZ.1 Ashby andEasterling considered a complete welding cycle to pro-duce grain growth diagrams for the prediction of themean grain size in real or simulated welds,2 whereasAndersen and Grong developed an analytical approachfor grain growth in the presence of coarsening and dissol-ving particles.3 More recently, Banerjee et al. investigatedaustenite formation and grain growth during rapid heat-ing for the same X80 steel considered in this work anddeveloped a grain growth model that accounts for thepresence of precipitates and their potential dissolution.4

The physically based model used in the work of Banerjee

et al. incorporates the work of Moon et al. who quantifiedthe effect of particle pinning on grain boundary migrationduring isothermal austenite grain growth associated withparticle coarsening and dissolution.5

The subsequent decomposition of austenite and itsrelation to the prior austenite grain structure has beenstudied for the HAZ. For example, Henwood et al.6

used a finite element heat transfer analysis in combinationwith the Ashby–Easterling grain growth model and thesolid-state Kirkaldy reaction approach for the austenitedecomposition kinetics to compute the microstructureas a function of space and time. Bhadeshia et al.7 calcu-lated the ferrite growth kinetics under paraequilibriumconditions using the principle of additivity. A review ofthese approaches can be found in the classic textbookby Grong.8 More recently, Zhang et al. described the aus-tenite decomposition in the HAZ of a low carbon steelusing a numerical model based on the Johnson–Mehl–Avrami–Kolmogorov (JMAK) analysis and Monte-Carlo simulations.9 A critical aspect of these variousapproaches is the need to explicitly link the various met-allurgical processes during a welding cycle, for example,grain growth, dissolution of microalloying precipitatesand decomposition of austenite to predict the gradedmicrostructure present in the HAZ and its associatedmechanical properties.The concept of microstructure engineering applied to

the development of linepipe steels has recently gainedincreasing attention to link the operational parametersof awelding operation to the microstructure and resultingproperties of the material.10 In this approach, the evol-ution of the microstructure is described by a set of linkeddifferential equations, which are integrated over the rel-evant thermal history aided, as required, by empiricalrelationships. For the HAZ in microalloyed steels, themodelling approach carefully considers austenite graingrowth and dissolution of precipitates4,11 since the auste-nite grain size and the solute content of Nb and Tistrongly affects the subsequent decomposition of auste-nite.12,13 In particular, a large austenite grain size and/ora high level of Nb in solid solution favours the formationof transformation products such as bainite and the associ-ated martensite/austenite (M/A) constituents, which mayhave detrimental effects on weld toughness, hydrogenresistance, cold and reheat cracking.2,14–19 A recentstudy by Gaudet20 has shown the variation of tensileproperties and tear resistance over a range of tempera-tures (20 to −60°C) for relevant microstructures foundin the HAZ of the current X80 steel (with the exceptionof the intercritical region, which is currently underinvestigation).As part of a larger project, a significant body of exper-

imental and theoretical work has recently been conductedfor a Ti–Nb microalloyed X80 linepipe steel.4,11,20–25 Theaim of this paper is to integrate the previous work on aus-tenite grain growth, dissolution of microalloyed precipi-tates and decomposition of austenite into a single modelto allow for the prediction of microstructure evolutionin the weld HAZ for this steel with an emphasis on rel-evant conditions of girth welds. Additional experimentshave been conducted to validate and, where necessary,to refine the model parameters previously reported inthe literature and examine the predictive capability ofthe model. A simplified temperature model (calibratedby experimental measurements in welds) has been

Garcin et al. Microstructure model for the heat-affected zone of X80 linepipe steel

2 Materials Science and Technology 2016

Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

implemented with the microstructure evolution model inorder to construct maps of the microstructure in theHAZ for different welding conditions.

Background and experimentsThe material used for the construction of the model is ahot-rolled X80 linepipe steel supplied by EVRAZ, Inc.NA (Regina, SK) with a nominal composition reportedin Table 1. The microstructure of the as-received steel con-sists mainly of irregular ferrite in which M/A particles arerandomly distributed. The experimental studies pre-viously reported for this steel are briefly summarised inthe following. Based on the review of these experimentalresults, a number of additional experiments are proposedto validate and/or refine parameters required for the inte-grated model.Gaudet conducted a series of weld trials on plates

instrumented with thermocouples spot welded in the bot-tom of small holes drilled at various positions from theweld fusion line.20 The temperature cycles, recorded insitu during the weld trials, provide critical informationon the relevant heating rates, peak temperatures and cool-ing rates at different positions in the HAZ for gas metalarc welding (GMAW) in single- and dual-torch configur-ations. More recently, similar experiments were alsoextended to submerged arc welding (SAW) conditions inorder to compare the two welding techniques and theirinfluence on the resulting microstructure.The state of precipitates present in the as-received hot

rolled steel was quantified by transmission electronmicroscopy (TEM) investigations coupled with energydispersive X-ray (EDX) and selected area diffraction pat-tern analysis.4 It was concluded that the initial structureafter hot rolling and coiling showed three families of pre-cipitates homogeneously distributed within the structure,that is, Ti-rich, Nb-rich and Mo-rich particles. Theirsizes range from 1 to 200 nm. A detailed analysis of theNb-rich precipitates revealed two populations, that is,large precipitates that presumably had formed in austeniteduring hot rolling with a radius >10 nm and small pre-cipitates formed in ferrite during coiling with a radius<5 nm. Thermodynamic calculations with Thermocalc(TCFE7 database) indicate that the dissolution tempera-ture of Mo-rich precipitates is 670°C, whereas the dissol-ution temperature for the Nb and Ti-rich precipitates are1080 and 1445°C, respectively. Thus, it is assumed that theMo-rich precipitates are fully dissolved when the steelreaches the austenite region during weld heat treatmentcycles, that is, all Mo is in solid solution. Second, accord-ing to the high dissolution temperature for the TiN pre-cipitates, it is assumed that the equilibrium fraction ofTiN precipitates is present in the as-received materialand will remain stable throughout the HAZ. Thus, onlythe Nb(CN)precipitates are considered to evolve (in sizeand volume fraction) during the HAZ heat treatmentcycles.

The kinetics of austenite grain growth were investigatedduring two main experimental test series.4,11 First, theaustenite grain size was examined during continuousheating and the influence of heating rate on the austenitegrain size was correlated to the peak temperature.4 Thesecond test series aimed at identifying in more detail therelative fraction of small and large precipitates. To thisaim, the limiting austenite grain size was measured duringisothermal holding at various temperatures by conductinga series of in situ grain size measurements with a laserultrasonics for metallurgy (LUMet) system.11 Therelationship between the limiting austenite grain sizeand the limiting pinning force associated with the pres-ence of Nb(CN) precipitates and their dissolution wasthoroughly examined using detailed analysis of the pre-cipitates solubility and considering a size distributionfor the precipitates using MatCalc. This detailed analysisled to a slightly modified model compared with that pro-posed by Banerjee et al. with both models leading to sat-isfactory descriptions of grain growth during continuousheating scenarios relevant for HAZ thermal cycles.Here, we will adopt the model proposed by Banerjeeet al. because of its simplicity and ease to incorporateinto an integrated model, for example, no coupling toMatCalc software is required.The subsequent austenite decomposition occurring

upon cooling was investigated by conducting a numberof continuous cooling transformation (CCT) exper-iments.21,23,24 Test samples were heat treated to bring allNb in solid solution and quenched before conductingthe austenitisation treatment for the CCT studies. Theaustenitisation treatments were designed to obtain a pre-determined austenite grain size and then the sampleswere either directly continuously cooled such that all Nbremain in solution or subjected to a holding of 20 minutesat 900°C (where Nb precipitates and only a small fractionremains in solid solution) before continuous cooling. Thefraction of Nb remaining in solution as a function ofholding time at 900°C was estimated from ageing testsat 570°C based on the magnitude of the age-hardeningpeak.26

The austenite decomposition kinetics were measuredwith a dilatometer for different initial austenite conditions(grain size, Nb in solution) and a range of cooling rates(1–40°C s–1) that are of relevance for the HAZ and differ-ent welding scenarios.21,24 Conventional metallographywith Nital and Le Pera etches was employed to quantifythe as-transformed microstructures in terms of the areafraction of each microstructure constituent, that is, fer-rite/bainite and M/A, respectively.21–25 Based on theresults of these CCT tests, a phenomenological modelfor the austenite decomposition into ferrite and bainitewas developed. The model consists of submodules for fer-rite start, ferrite growth, bainite start and bainitegrowth.22,27–29 Furthermore, the fraction of M/A con-stituents in the different samples generated from CCTtests was related to the transformation start temperaturemeasured upon cooling.23,25 The proposed overall auste-nite decomposition model replicates the CCT obser-vations with sufficient accuracy.The initial fraction of Nb present in solid solution in

the base metal is an important parameter in the modelbut is difficult to quantify accurately. Initially, it was con-sidered in the work of Banerjee et al. that the equilibriumfraction of Nb is reached after hot rolling and coiling of

Table 1 Chemical composition of X80 linepipe steel (keyalloying elements)

Elements C Mn Mo Nb Ti N

wt-% 0.060 1.65 0.24 0.0340 0.0120 0.005at.-% 0.278 1.67 0.14 0.0204 0.0139 0.019


Materials Science and Technology 2016 3

Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

the base plate. However, the analysis of Lu on a range ofas-hot rolled linepipe steels suggests that while themajority of Nb is precipitated, equilibrium is not typicallyreached.30 Based on the low coiling temperature of thepresent steel it is, thus, concluded that at least 50 ppmNb remains in solid solution after coiling. Consideringthe challenges associated with an accurate measurementof such a small fraction of microalloying elements insolid solution, the amount of Nb in solution in the basemetal is kept as an adjustable parameter to benchmarkthe model predictions with the transformation behaviourobserved during weld heat treatment cycle simulationsconducted in the present work.In this work, a number of additional tests were con-

ducted for conditions that are of particular relevance forthe HAZ. All tests were carried out using a Gleeble3500 thermomechanical simulator and employing speci-mens from the as-hot rolled steel with dimensions of10 × 60 × 1.5 mm. First, laser ultrasonic grain sizemeasurements were conducted during continuous heatingat rates of 10 and 100°C s–1 as well as during a dual passscenario with heating rates of 100°C s–1 and peak temp-eratures of 1225°C in both passes. Furthermore, dilato-metry tests were conducted to quantify the austenitedecomposition kinetics during simplified heat treatmentsof single and dual pass welding. The cooling portions ofeach cycle are characterised by the time required for thetemperature to drop from 800 to 500°C (t8–5). Threepeak temperatures, that is, 1000, 1200 and 1350°C, andtwo different cooling conditions with t8–5 = 7 s and t8–5= 15 s were selected for the single pass scenario as illus-trated in Fig. 1a. The validation tests were furtherextended to a heat treatment involving two successivewelding cycles as shown in Fig. 1b. The first pass is definedby a peak temperature of 1200°C with t8–5 = 7 s, and thesecond pass by a peak temperature of 1200°C with t8–5= 15 s, that is similar to what was measured in the exper-imental weld trials.20 In all cases, the heating step is con-ducted at a rate of 100°C s–1 up to the peak temperaturewhere the holding time is 0.5 s before the onset of cooling.Finally, the integrated model is applied to the predic-

tion of the microstructure gradient from the fusion lineto the intercritical annealing zone in four different weld-ing conditions for single torch GMAW with no pre-heat(GMAW-ST), with pre-heat (GMAW-ST-P), single-torch SAW with no pre-heat (SAW-ST) and a dual-torchGMAW with no initial pre-heat (GMAW-DT). The

influence of different welding techniques on the resultingmicrostructure in the HAZ can thereby be directly com-pared using the integrated model.

Modelling strategyInitial state of precipitatesThe initial volume fraction of TiN precipitates f 0TiN is esti-mated to be 2.33×10–4 assuming that all Ti is precipitated.The initial volume fraction of Nb(CN) precipitates f 0NbCNis assumed to be lower than the equilibrium fraction andcan be related to the atomic fraction of niobium X 0

Nb inthe matrix of the as-hot rolled steel by

f 0NbCN = (XNNb − X 0

Nb)2vAt

NbCN

vAtFe

(1)

where XNNb is the nominal atomic fraction of niobium in

the steel, vAtFe is the atomic volume for iron and vAt

NbCN isthe atomic volume for Nb(CN). As mentioned earlier,the initial atomic fraction of niobium in the matrix isnot known accurately and is thus considered as an adjus-table parameter in the present approach.In addition to the initial volume fractions precipitated,

the initial mean particle radii are required in the model.Although well aware of the uncertainties associated withthe measurement of precipitate sizes using TEM, Bane-rjee et al. had estimated the mean radii for the various pre-cipitate families. In particular, they found two families ofNb(CN) precipitates, that is, larger precipitates (meanradius of r0NbCN(L) = 69 nm) and smaller precipitates

(mean radius of r0NbCN(S) = 2 nm), and a mean radius

r0TiN = 61 nm for TiN precipitates.As proposed by Banerjee et al.,4 the initial volume frac-

tion of large Nb(CN) precipitates is estimated from theratio of the overall precipitate volumes in the as-receivedmaterial for large Nb(CN) (23 pct.) and TiN (61 pct.)such that

f 0NbCN(L) =2361

f 0TiN (2)

The initial volume fraction of small Nb(CN)precipitates is then deduced from the total initialvolume fraction of Nb(CN) precipitates f 0NbCN byf 0NbCN(S) = f 0NbCN − f 0NbCN(L).

1 Simplified heat treatment cycles to simulate a single torch and b dual torch welding



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

This initial precipitation state is only applicable for theas-hot rolled state of the investigated steel and would haveto be reconsidered for a different steel chemistry or even asignificantly different rolling schedule for the present steel.

Dissolution of precipitates and austenite graingrowthDuring welding, the initial population of precipitatesevolves with respect to the temperature history at agiven distance from the fusion line. For the rapid heattreatment conditions in the HAZ, only dissolution of pre-cipitates will be considered, whereas potential growth andcoarsening that are slower processes will not be taken intoaccount. Further, the austenite grain size is described asthe mean value of the size distribution, and potentialchanges in the width of the size distribution are not con-sidered in this approach. The evolution of the mean aus-tenite grain diameter dg is affected by the evolution of theprecipitate population via a Zener-type pinning pressure.In such a situation, the evolution of the mean austenitegrain diameter can be expressed as4

ddt

dg = MgsB a

dg− b

∑i

firi

( )(3)

where the summation index i represents the types of pre-cipitates present, TiN, small Nb(CN) and large Nb(CN)precipitates. The parameters a and b are geometrical con-stants, sB is the grain boundary energy and Mg is thegrain boundary mobility. Mg can be expressed by anArrhenius relationship, that is,Mg = Mg

0 exp (−(Qg/RT)), where Mg0 is the pre-expo-

nential factor and Qg is the activation energy, R is thegas constant and T is the absolute temperature.In this approach, the dissolution of only two families of

precipitates is considered, that is, large and small Nb(CN). Due to their high dissolution temperature, the vari-ation in the fraction of TiN precipitates is neglected in thepresent model. The diffusion-controlled dissolution rateof spherical Nb(CN) precipitates embedded in a solid sol-ution is described by31,32

ddt

rNbCN(L,S) = DNb

rNbCN(L,S)

XMNb − X I

Nb

vAtFe

vAtNbCN

XPNb

( )− X I

Nb

(4)

here, DNb is the diffusivity of niobium in the matrix ofaustenite and is given by an Arrhenius relationship,DNb = DNb

0 exp (QNbD /RT). The variable XP

Nb is theatomic fraction of niobium in the precipitate, that is, 0.5for Nb(CN). XM

Nb is the solute atomic fraction of niobiumin the matrix and X I

Nb is its equilibrium atomic fraction atthe precipitate/matrix interface. Due to the small sizes ofthe precipitates, the interface curvature plays an impor-tant role in the equilibrium atomic fraction at the inter-face, that is, the so-called Gibbs–Thomson effect, suchthat the equilibrium solubility depends on the precipitatesradius and the matrix precipitate interfacial energy sPas

X INbX

MC+N = KNbCN exp

4sPvatNbCN

rNbCNkBT

( )(5)

where XMC+N is the solute atomic fraction of carbon and

nitrogen in the matrix, KNbCN is the temperature-

dependent solubility product for Nb(CN) precipitatesand kB is Boltzmann’s constant.In this approach, the number density for each precipi-

tate family is considered constant such that the volumefraction for each class can be updated according to theevolution of their mean radii. The volume fraction isthen used to evaluate the mean solute atomic fraction ofniobium as well as carbon and nitrogen in the matrixaccording to mass balance, that is,

XMNb = XN

Nb −12vAtNbCN

vAtFe

( fNbCN(L) + fNbCN(S)) (6)

XMC+N = XN

C+N − 12vAtNbCN

vAtFe

(fNbCN(L) + fNbCN(S)) (7)

where XNC+N = XN

C + (XNN − XN

Ti) is the nominal carbonconcentration plus the excess nitrogen that is not precipi-tated in TiN. Following this approach, the evolution ofthe mean austenite grain diameter as well as mean radiusand volume fraction for each precipitate family is calcu-lated along the thermal path.Table 2 summarises the parameters used in the com-

bined precipitate dissolution and austenite grain coarsen-ing model. This model provides two importantparameters for the subsequent austenite decompositionmodel, that is, (1) the prior austenite grain diameter and(2) the mean atomic fraction of niobium in solution inthe matrix. In addition to cooling rate, these two par-ameters affect the onset of the austenite decompositionduring subsequent cooling.

Austenite decompositionIn this approach, two separate models are utilised for theprediction of the transformation start temperature sinceaustenite decomposition may start with either ferrite orbainite in the present steel depending on the HAZ coolingconditions.The model constructed to predict the onset of ferrite

formation was originally developed for plain carbonsteels.27 It considers the carbon diffusion-controlledearly growth of ferrite nuclei formed at austenite grainboundary corners at a temperature TF

0 . The model wasextended to include the solute drag effect of niobium onthe moving austenite/ferrite interface.22,40 The evolutionof the mean radius of the corner-nucleated ferrite graincan be written such that

Table 2 Parameters for the austenite grain growth andprecipitate dissolution model

Parameters Value References

DNb0 /m2 s−1 5.3×10−2 33

QNbD /kJ mol−1 344 33

Qg/kJ mol−1 350 34a 4 35b 12 36sB/J m−2 0.5 37sP/J m−2 0.5 33log10(KNbCN) −1.32− 6670/T 38DC

0 /m2 s−1 1.5×10−5 39

QCD /kJ mol−1 142.1 39



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

Rf =∫TTF0

DCXg

C − XNC

XgC − Xa

C

1wRf

1+DC1XMNb

Rf

( )−1[ ]

dT (8)

where Rf is the radius of the growing ferrite grain, DC isthe carbon diffusivity in austenite described by an Arrhe-nius relationship DC = DC

0 exp(QCD/RT) and w is the

instantaneous cooling rate. XgC and Xa

C are the equili-brium atomic fraction of carbon in austenite and ferrite,respectively. The coefficient 1 is an adjustable parameterdescribing the intensity of the interaction of solute Nbwith the austenite/ferrite interface following the approachdescribed by Fazeli and Militzer.22 The temperaturedependence of X g

C and XaC are calculated with Thermo-

calc using the TCFE7 database and assuming full equili-brium for all alloying elements, that is, orthoequilibrium.The condition for measurable start temperature of ferriteformation TF

S is given by attaining a critical carbon levelX ∗

C at the austenite grain boundary such that

Rf .12

√ X ∗C − XN

C

X gC − Xa

Cdg (9)

According to the analysis by Militzer et al.,28 the par-ameter X ∗

C is related to dg with an empirical expressionof the form X ∗

C/XNC = l1 + l2/dg. Further, it is con-

sidered that the transformation start condition coincideswith reaching a measurable fraction of ferrite, that is,5% transformed.For sufficiently fast cooling, ferrite does not form and

austenite decomposition starts with the formation of bai-nite. The bainite start temperature can be attributed to acritical driving pressure,41 but measurable bainite starttemperature is similar to the ferrite start temperatureassociated with 5% transformed. Based on the CCTstudies, the following expression had been proposed tofind the bainite start temperature TB

S :42

0.05 =∫TBS

TB0

b1 + b2(T − 273.15)w(1+ b3XM

Nb)

[ ]dT (10)

Here, the temperature TB0 is the temperature where

the critical driving pressure is reached for the givensteel composition and the bi are empirical parameters.Similar to the ferrite start temperature, it was foundthat an Nb solute drag effect is operational in theearly stages of bainite formation as described by theparameter b3.In the intermediate cases where both ferrite and bai-

nite form, the temperature TB0 is taken as the onset of

bainite formation. Since the critical driving pressure isnow dependent on the carbon enrichment of austeniteduring ferrite formation, this temperature decreaseswith ferrite fraction from its value for the nominal steelcomposition.Subsequent growth of ferrite and bainite is described

using the JMAK model and adopting additivity suchthat the normalised fraction transformed FF,B

N along acooling path is given by

FF,BN = 1− exp −

∫TF,BS

T

b(T , XMNb, dg)w

dT

⎛⎜⎝

⎞⎟⎠

n

(11)

Here, n is the JMAK exponent and the subscript N isintroduced to refer to the normalised fraction of ferrite(F) or bainite (B). The rate parameter β can be written as

b = exp b1T + b2

db3g

( )(12)

where the bi are taken to be linear functions of theamount of niobium in solution, that is,

bi = bi1XMNb + bi2 (13)

A linear dependence is adopted for simplicity as a rela-tively narrow range of Nb levels is considered in the pre-sent work. Further investigations would be necessary toobtain a more accurate description of the phenomenon.When no bainite forms, the true ferrite fraction after

completion of the austenite decomposition FF is givenby 1− FM/A where FM/A is the true M/A fraction thatcan be related to the transformation start temperatureTF,BS by23

FM/A = 0.02+ 0.107 exp − (TF,BS − 859)

2

648

( )

+ 0.03

1− exp (−014TF,BS + 124)

(14)

Here, the M/A fraction is about 0.05 for a mostly ferri-tic microstructure leading to a true ferrite fraction of 0.95which is consistent with the paraequilibrium ferrite frac-tion at temperatures of 650°C and below where ferrite for-mation would typically cease. When both ferrite andbainite form, the true ferrite fraction is obtained by divid-ing the normalised ferrite fraction, FF

N, with the paraequi-librium ferrite fraction at the bainite start temperatureTB0 . The paraequilibrium ferrite fraction is obtained

with Thermocalc (TCFE7 database). The total bainitefraction at the end of the transformation, FB, is thendeduced from the true fractions of ferrite and M/A as1− FF − FM/A. Further, the relative fraction of upperand lower bainite can be estimated from the total fractionof bainite based on the transformation temperature. Thedistinction between upper bainite and lower bainite hasbeen clarified in the work of Takayama et al.43 and Reich-ert et al.25 by detailed microstructure characterisationusing electron backscatter diffraction pattern (EBSD)mapping. Each austenite grain transforms to primarilyone Bain group variant in the case of upper bainite,while lower bainite is characterised by the occurrence ofall three Bain groups in each parent austenite grain.Further, the determination of the Kernel Average Misor-ientation (KAM) from the EBSD maps indicates increas-ing KAM values and, thus, increasing dislocation density,with lowering the transformation temperatures. Inaddition, the M/A fraction study reported by Reichertet al.23 for this steel shows that 12.8% of M/A is presentin the upper bainite microstructure with a transformationstart temperature of 586°C, whereas only 2% of M/A ispresent in the lower bainite microstructure with a trans-formation start temperature of 540°C. Thus, to rationalisethe gradual change from upper-to-lower bainite betweenthese two transformation start temperatures, the relativenormalised volume fractions of upper and lower bainite(FUB

N , FLBN ) can be deduced from



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

FUBN = (1− FLB

N ) = FM/A − 0.020.108

(15)

The true fraction of upper bainite FUB and lower bai-nite FLB are then evaluated proportionally to the totalfraction of bainite FB. Table 3 lists the parameters usedin the austenite decomposition model for ferrite and bai-nite formation, respectively.

Simplified temperature modelBased on the data from the weld trial, a simplified temp-erature model (in °C) is proposed by employing theRosenthal equation for thick plates,16 that is,

T (t, r) = T0 + q2pkvt

exp−r2

(4at)

( )(16)

Here t is the time, r is the distance from the centre line, v isthe weld speed, q is the heat input, k is the thermal con-ductivity, a is the thermal diffusivity and T0 is the pre-heat temperature. The peak temperature is given by

Tpeak(r) = T0 + 2q/vpeCvr2

(17)

where Cv is the volumetric heat capacity for iron and e isEuler’s number. According to equation (16), the coolingtime from 800 to 500°C is given by

t8−5 = q/v2pk

1500− T0

− 1800− T0

( )(18)

To describe the measured time–temperature curves, q/vis employed as an effective heat input parameter while forsimplicity assuming k = 41 J m−1 s−1 K−1 and Cv = 600 Jkg−1 K−1 16 as representative values for low-carbon steel.The HAZ is defined with respect to the position of thefusion line where the peak temperature is the meltingtemperature of 1500°C. The width of the HAZ can thenbe determined from the locations of the peak tempera-tures of 900°C for the zone fully austenitised where theproposed microstructure evolution model is applied andthe width of the overall HAZ including the intercriticalregion is determined from the location of the peak temp-erature of 700°C where the austenite/ferrite two-phasefield is reached.

General structure of integrated modelThe integrated model is constructed from the individualsub-models calibrated using the aforementioned exper-imental studies. The flow chart of the model is presentedin Fig. 2. The temperature–time cycles calculated fromthe simplified temperature model for each position inthe HAZ are used as input for the calculation togetherwith the initial precipitate fractions and radii. The evol-ution of austenite grain size and precipitate sizes andvolume fractions are quantified for all positions alongthe HAZ thereby providing the starting conditions forthe austenite decomposition model during cooling. Boththe ferrite and bainite start transformation temperatureare evaluated simultaneously. If the ferrite start tempera-ture is higher than that for bainite, the ferrite growthmodel is called together with the model to determinethe transition from ferrite to bainite formation. If thetransition temperature to bainite is reached before thecompletion of the austenite decomposition, the bainitegrowth model is called. On the other hand, if the bainitestart formation temperature is above the ferrite starttemperature, only the bainite growth model is called.The fractions of ferrite, upper and lower bainite are renor-malised together with the fraction of M/A calculated for agiven heating cycle.

ResultsValidation testsFigure 3a compares the measured evolution of the auste-nite grain size during continuous heating with the predic-tions of the austenite grain growth model. The initialaustenite grain size of 5 µm resulting from the austeniteformation from the starting hot-rolled microstructure isindependent of heating rate. The onset of measurablegrain coarsening scales with heating rate from about1050°C for 10°C s–1 to 1100°C for 100°C s–1, respectively,as predicted by the model. Measurements are extended toabout 1250°C above which laser ultrasonic signalsbecome unreliable. As austenite grain sizes measuredwith LUMet are typically consistent with metallographicdata within a 20% margin,44,45 measured and predictedgrain sizes are in reasonable agreement in particular forthe faster heating rate of 100°C s–1. For the HAZ, aneven higher heating rate of at least 1000°C s–1 wasmeasured and the model predictions are included for

Table 3 Model parameters of the austenite decomposition model

Ferrite transformation start∗

T F0 (°C) X ∗

C /XNC 1(s μm−1 at. ppm−1)

700 1.74+ 6.8/dg 0.043Bainite transformation startb1 (s−1) b2 (s−1 C−1) b3 (at. ppm−1)86.95 −0.132 2.1Bainite nucleation temperature from ferrite (°C):640− 143F F0.5 + 288F F − 528F F2 + 646F F3 − 380F F4

JMAK parameters for ferrite transformation∗

n b11 b12 b21 b22 b31 b321.1 5.27 10−5 −3.59 10−2 −4.64 10−2 24.9 1.97 10−4 5.98 10−2

JMAK parameters for bainite transformation∗

F F n b11 b12 b21 b22 b31 b32. 0 1.1 6.88 10−6 −2.34 10−2 −2.47 10−2 18.0 2.40 10−4 1.17 10−1

0 1.1 7.64 10−6 −2.36 10−2 −1.22 10−2 12.5 0 0∗Where T is in °C, dg in μm and XM

Nb in at. ppm.



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

this scenario in Fig. 3a, indicating that negligible graingrowth takes place during heating up to 1250°C. Figure3b illustrates the case for the dual-torch scenario employ-ing a heating rate of 100°C s–1. In the first pass, themeasured austenite grain size is 20 µm when reachingthe peak temperature of 1225°C and increases during abrief hold of 1 second there to about 30 µm before theonset of cooling during which no further grain growthoccurs. In the second pass, the austenite grain sizeobtained is also 30 µm even though there are in detailimportant differences in the grain growth behaviour inboth passes. In the second pass, the pinning of Nb(CN)particles is reduced due to partial dissolution in the firstpass and more importantly the initial austenite grainsize resulting from austenite formation is 15 µm, that is,significantly larger than that observed in the first pass.Here, one has to consider that the microstructure of theas-hot rolled steel and that obtained after the single passheat treatment are different. As an illustration one may

consider the microstructures that result from differentsingle pass scenarios. For sufficiently low peak tempera-tures (e.g. 1000°C, Fig. 4a), irregular ferrite forms similarto that found in the as-hot rolled steel. For sufficientlyhigh peak temperatures, for example, 1200°C (Fig. 4d ),bainite rather than the irregular ferrite of the base metalforms. These different microstructures have also verydifferent populations of M/A constituents in terms oftheir fraction, size and morphology23 such that differentaustenite grain sizes can be expected to result from auste-nite formation starting from these varieties of ferritic and/or bainitic microstructures. This observation is similar tothat discussed in a recent austenite formation study for alow carbon steel.46 Assuming an initial austenite grainsize of 15 µm for the second pass, the predicted austenitegrain size evolution is in reasonable agreement with themeasurements. The predicted austenite grain size is inboth passes 25 µm, which is within the error of measure-ments, that is, 20%, consistent with the measured 30

2 Flow chart of integrated microstructure model for the HAZ

3 Laser ultrasonic grain size measurements and comparison with austenite grain growth model predictions: a for continuousheating at different rates, b for dual-pass welding scenario with heating rates of 100°C s–1 and peak temperatures of 1225°Cwith a holding time of 1 second



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

µm. Rather than tuning the model parameters in order toobtain better agreement between measured and predictedgrain sizes, all model parameters for austenite graingrowth and precipitate dissolution are kept identical asthose reported by Banerjee et al.4 The only difference inthe implementation of Banerjee’s model is the assumptionmade with respect to the Nb level in solution, whichaffects the initial fraction of small Nb precipitates. A sen-sitivity analysis shows, however, that this is of little conse-quence for the grain size prediction as substantialaustenite grain growth can only occur once all small pre-cipitates are dissolved. Further, for the HAZ portionswith larger austenite grain sizes, bainite forms and thebainite transformation kinetics is independent of theprior austenite grain size. Thus, the prediction of a 20%smaller austenite grain size has only a negligible effecton the prediction of the resulting transformation pro-ducts. Thus, the proposed austenite grain size model isapplicable to the HAZ, provided that for dual-torch weld-ing the effect of the HAZ microstructure obtained in thefirst pass on the initial austenite grain size is taken intoaccount. Here, for simplification 5 µm is assumed for apredominantly ferritic microstructure and 15 µm for apredominantly bainitic microstructure.Examples of the final microstructures resulting from

different single-torch scenarios (see Fig. 1a) are shownin Fig. 4 illustrating the role of peak temperature andcooling conditions. For a peak temperature of 1000°Cand a t8–5 of 15 seconds, 90% of irregular ferrite formswith 10% of M/A constituent that is similar to that inthe base metal. Optical metallography gives 10% of irre-gular ferrite, 82% of bainite and 8% of M/A constituentfor the sample cycled with a peak temperature of 1200°C (Fig. 4b). On the other hand, 5% of M/A constituentsand 95% of bainite are observed in the sample cycledwith a peak temperature of 1350°C (Fig. 4c). For higher

peak temperatures (1200 and 1350°C) and a t8–5 of 15seconds, the quantitative optical metallography resultsare consistent with primarily bainitic microstructures con-taining mainly upper bainite forming for lower coolingrates (e.g. for t8–5 = 15 second, see Figs. 4b and c). Increas-ing the cooling rate (i.e. decreasing t8–5 to 7 seconds) and/or peak temperature promotes the formation of fine baini-tic ferrite laths characteristic of lower bainite rather thanupper bainite, as illustrated for a peak temperature of1350°C (see Figs. 4c and d ).Figure 5 shows the evolution of the bainite microstruc-

ture resulting from a dual-torch scenario where pass 1 hasa peak temperature of 1200°C and t8–5 = 7 second and apass 2 has a peak temperature of 1200°C and a t8–5 of15 seconds. After the first pass, a mixed upper/lower bai-nite structure forms (see Fig. 5a), which appears afterNital etching as very fine and numerous bainite plateletsformed within the prior austenite grain. On the otherhand, after the second pass upper bainite forms as coarsebainitic ferrite islands with dispersed carbide precipitateslocated mainly at the grain boundaries (see Fig. 5b) due tothe slower cooling from the peak temperature.The transformation start temperatures can be taken as

a representative transformation temperature for continu-ous cooling scenarios such as those observed in theHAZ. Figure 6 compares measured and predicted trans-formation start temperatures for the simplified weld heattreatment simulations. The transformation start tempera-ture is measured with an accuracy of ±5°C by the appli-cation of the lever rule on dilatometry data. Reasonablygood agreement is obtained provided the initial amountof Nb in solution is assumed to be 60 ppm. The austenitedecomposition model is rather sensitive to the amount ofNb in solution. For example, bringing all Nb in solutioncan, depending on cooling rate and austenite grain size,decrease transformation temperatures by 50–100°C

4 Final microstructures obtained during single pass simulations a Tpeak = 1000°C, t8–5 = 15 seconds, b Tpeak = 1200°C, t8–5 = 15seconds, c Tpeak = 1350°C, t8–5 = 15 seconds, d Tpeak = 1350°C, t8–5 = 7 seconds



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

compared to the case of having most Nb precipitated.22

Having an accurate prediction of the transformationtemperature ensures also that the overall fraction of trans-formation products obtained from the austenitedecomposition model is consistent with the microstruc-ture observations shown in Figs. 4 and 5. An initialamount of Nb in solution of 60 ppm in the as-hot rolledsteel seems reasonable based on the observed austenitedecomposition kinetics and the abovementioned TEMstudies of Lu.30 A sensitivity analysis conducted byadjusting the initial amount of Nb in solution from 50to 70 ppm shows that this change modifies the predictionof the transformation start temperature by 10°C.

Application of integrated modelFour cases of welding conditions are considered inthe present study: single-torch GMAW with no pre-heat (GMAW-ST), single-torch GMAW with pre-heat (GMAW-ST-P), single-torch SAW with no pre-heat(SAW-ST) and dual-torch GMAW with no pre-heat(GMAW-DT). The three scenarios without pre-heat, thatis, GMAW-ST, SAW-ST and GMAW-DT, replicate con-ditions of the previously conducted weld trials. The dual-torch scenario is a combinationofGMAW-STwitha secondpass that can effectively be described by starting with a pre-heat. The peak temperatures in both welding passes areassumed to be the same for a given position in the HAZ.Table 4 lists these four cases with the t8–5 times measuredin the welding trials, except for GMAW-ST-P, where thistime was obtained using the same effective heat input,ratio of heat input to weld speed, as for GMAW-ST. TheRosenthal fit emphasises to replicate the t8–5 times by deter-mining the required effective heat input as shown in Table 4together with the resulting width of the HAZ, which is con-sistent with the estimates from the metallographic obser-vations made on the weld trial samples. The time–distancetemperature profiles in the HAZ for these four cases areillustrated with contour plots in Fig. 7.Based on these thermal profiles, the integrated model

predicts the dissolution of Nb(CN) precipitates, austenitegrain growth and the resulting austenite decomposition asa function of time and position in the HAZ. Figure 8 sum-marises the prediction of three important microstructureparameters, that is, Nb in solution, austenite grain sizeat the onset of austenite decomposition and the trans-formation start temperature as a function of distancefrom the fusion line for the four investigated welding scen-arios. For reference, the peak temperatures are shown aswell. In all scenarios, Nb is completely in solution near

5 Microstructures obtained during the two-pass simulation: a after the first pass with Tpeak = 1200°C, t8–5 = 7 second, b after thesecond pass with Tpeak = 1200°C, t8–5 = 15 seconds

6 Comparison of calculated and measured transformationstart temperature for Gleeble-simulated simplified HAZthermal paths

Table 4 Data for the temperature profiles based on the welding trials

CaseEffective heat input/kJ.

mm−1 Pre-heat/°C t8–5/sPosition of

Tpeak = 900°C/mmPosition of

Tpeak = 700°C/mm

GMAW-ST 2.02 27 6.5 2.4 3.9GMAW-ST-P 2.02 200 13.0∗ 3.1 5.3SAW-ST 5.76 27 18.4 4.1 6.5GMAW-DT (second pass) 2.02 200∗∗ 13.3 3.2 5.3∗Obtained assuming effective heat input of GMAW-ST.∗∗Effective pre-heat temperature resulting from the first pass.



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

the fusion line within a layer that increases from 0.4 to 0.8mm with the relative heat input (including pre-heat), thatis, the width of this zone is largest for the SAW scenario,which has the highest effective heat input for the con-sidered welding cases. Then, the Nb solute level dropsgradually to an intermediate value of 165 ppm that is con-sistent with the dissolution of the small Nb(CN) precipi-tates. A further decrease from this intermediate Nbsolute level occurs for a sufficiently large distance fromthe fusion line that coincides with the position where thepeak temperatures reaches about 1080°C that is expectedas the Nb(CN) dissolution temperature for the presentsteel composition. The austenite grain size at the fusionline is for all four welding scenarios in the range of 80–90 µm. The magnitude of this grain size is similar tothat observed in welding trials by Hamad et al.47 They,however, measured typically somewhat smaller grainsizes. In single-torch welding most grains were below 70µm and above 70 µm for dual-torch welding. These appar-ent discrepancies can be mitigated when considering thepredicted rapid decrease of grain size with distance fromthe fusion line, for example, for the GMAW-ST scenario,the grain size is about 70 µm at a distance of 200 µm fromthe fusion line and this decrease is less pronouncedimmediately near the fusion line for higher heat inputs(e.g. for SAW the grain size is still 80 µm at a distance200 µm away from the fusion line). Further, the presentmodel does not consider minor dissolution and/or coar-sening of TiN precipitates which in particular for dual-torch scenarios may lead to a further minor reduction

in pinning forces resulting in somewhat larger grainsizes when comparing single- and dual-torch welding.The rapid decrease of austenite grain size with distancefrom the fusion line is consistent with typical observationsmade on actual welds. The rate of the decrease in grainsizes mirrors the drop in peak temperatures with distance.In all cases, the position of onset of grain coarsening cor-relates well with the peak temperature being the Nb(CN)dissolution temperature of 1080°C and can, thus, be takenas the separation of the coarse-grained HAZ (CGHAZ)from the fine-grained HAZ (FGHAZ). The resultingtransformation start temperatures reflect the state of aus-tenite in terms of grain size and Nb in solution for eachindividual welding scenario. In a first approximationtransformation, start temperatures correlate well withthe amount of Nb in solution, that is, decreasing Nb insolution leads to an increase in the transformation temp-erature. When comparing different welding scenarios,however, the t8–5 times play a significant, if not dominantrole. Increasing t8–5, that is, decreasing the cooling rate,leads to an increase in transformation start temperatures.GMAW-ST with the lowest t8–5 of 6.5 seconds shows thelowest transformation temperatures, GMAW-ST-P andGMAW-DT have both a t8–5 of about 13 seconds andthus similar transformation temperatures, whereasSAW-ST with the largest t8–5 of about 22 seconds hasthe highest transformation temperatures.The transformation start temperatures shown in Fig. 8d

provide a critical indicator for the final microstructurethat forms as a function of distance in the HAZ. The

7 Temperature map resulting from the various simulations conducted: aGMAW-ST, bGMAW-ST-P, c SAW-ST and dGMAW-DT



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

8 Important temperature and microstructure indicators as a function of distance from the fusion line for the four weld scen-arios: a peak temperature, b Nb in solution at onset of austenite decomposition, c austenite grain size at onset of austenitedecomposition and d transformation start temperature

9 HAZ microstructure maps for the simulated welding scenarios: a GMAW-ST, b GMAW-ST-P, c SAW-ST and d GMAW-DT



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

HAZ microstructure maps obtained with the integratedmodel are shown for the four welding scenarios in Fig.9. In the GMAW single-torch scenario with no pre-heat(Fig. 9a), the lowest transformation temperatures areobserved and lower bainite forms near the fusion line inthe CGHAZ and upper bainite in the remaining portionof the HAZ farther away from the fusion line. For thesingle-torch GMAW with pre-heat, the HAZ is widerand the t8–5 is larger such that upper bainite forms inthe CGHAZ and ferrite in the remaining HAZ (Fig.9b). This observation is similar to that made for the twoother cases, that is, upper bainite in the CGHAZ and fer-rite otherwise (see Figs. 9c and d ). In the SAW scenario,the HAZ is significantly wider (Fig. 9c), whereas theHAZ microstructure and width for the dual-torchGMAW scenario (Fig. 9d ) is essentially identical to thatobtained for the single-torch GMAW with pre-heat(Fig. 9b) since both cases have the same peak tempera-tures and t8–5 times. When comparing single- and dual-torch GMAW without pre-heat (see Fig. 9a and d ), theprediction of lower bainite forming near the fusion linein single-torch welding and upper bainite for dual-torchwelding is consistent with the observations of Hamadet al. in weld trials.47 There are, however, some discrepan-cies between the calculated and observed M/A fractionsfor these two welding scenarios. The current modelsuggests an M/A fraction near the fusion line of 0.02 forsingle-torch welding but 0.12 for dual-torch welding,whereas Hamad et al. observed an M/A fraction of0.01–0.02 for both welding scenarios. Further studiesare required to clarify the differences in calculated andobserved M/A fractions for dual-torch welding.To further evaluate the model, it is useful to determine

the sensitivity of the predictions with respect to a numberof fit parameters in particular those that were used to tunethe model, that is, the initial level of Nb in solution andthe initial austenite grain size resulting when austeniteforms from bainite in the second pass during dual-torchwelding. Changing the Nb level in solution by ±20 ppm,that is, considering 40 and 80 ppm instead of 60 ppm,has a marginal effect on the obtained microstructuremaps. The change in the initial Nb level does primarilyaffect the transformation temperature in the FGHAZand the ferrite fraction in this zone away from the fusionline increases with decreasing Nb content in solutionwhen a mixture of ferrite and bainite is present, whichhere applies only to the GMAW-ST scenario (Fig. 9a).In all other cases, a rather sharp transition from a primar-ily ferritic to a primarily bainitic microstructure occurs ata position, which is unaffected by the initial amount of Nbin solution. Similarly, changing the initial austenite grainsize by ±10 µm, that is, considering 5 and 25 µm insteadof 15 µm in the bainitic portion of the HAZ after thefirst pass in dual-torch welding, has essentially no effecton the resulting microstructure maps as it only appliesto the CGHAZ where the austenite decomposition intobainite is independent of austenite grain size.

ConclusionThe case study presented here illustrates that the proposedintegrated model is a powerful tool to compare HAZmicrostructures for different welding scenarios. Themodel is based on the microstructure engineering conceptbut employs still a number of empirical relationships, in

particular for bainite formation and the M/A fraction.The transition between different transformation products,that is, irregular ferrite, upper and lower bainite, which isof relevance for the HAZ is gradual. The proposed inte-grated model provides a modelling strategy that couldbe adopted for different steel chemistries as well which,however, would require to re-adjust some of the modelparameters.EBSD studies and evaluation of mechanical properties

indicate that a unified characterisation of these transform-ation products in terms of an effective grain size (e.g. linelength of high angle boundaries per unit area) and effec-tive dislocation density can be considered as a functionof transformation temperature. This approach may pro-vide an alternative avenue to formulate an austenitedecomposition model with a reduced number of tuningparameters that, moreover, would provide a natural linkto mechanical properties, which depend primarily ongrain size and dislocation density. Further, a more rigor-ous model approach is required for the M/A constituents.Here, it is required to also include their size and shape aswell as the actual fraction of retained austenite into animproved model.Connecting the predicted, graded microstructures with

a structure–property model will then permit to evaluatethe integrity of the HAZ as a function of welding pro-cedure. A direct linkage to the welding parameters (i.e.heat input and weld speed) will, however, require toreplace the simplified temperature model with a heattransfer model that accounts for the complexity of theweld geometry as well as the latent heat of both solidifica-tion of the weld metal and the austenite decomposition.The current model is limited to the portion of the HAZ

that is completely austenitized. Thus, it will be critical toextend the model to intercritical heat treatments with aprimary focus on an intercritical cycling of the CGHAZin a subsequent weld pass. There is increasing evidencethat martensite forming from intercritical austenite inthe CGHAZ can lead to a significant deterioration ofHAZ properties.48

AcknowledgementThe financial support of the Natural Sciences and Engin-eering Research Council (NSERC) of Canada, Evraz,Inc. NA and TransCanada Pipelines Ltd is acknowledgedwith gratitude. The contributions of Dave Taylor ofTransCanada and Michael Gaudet (UBC) in conductingand analysing weld trials are greatly appreciated. Further,we would like to thank Jennifer Reichert and MehranMaalekian for their many contributions to the presentwork.

References1. S. Mishra and T. DebRoy: ‘Non-isothermal grain growth in metals

and alloys’, Mater. Sci. Technol., 2006, 22, 253–278.2. M. F. Ashby and K. E. Easterling: ‘A first report on diagrams for

grain growth in welds’, Acta Metall., 1982, 30, 1969–1978.3. I. Andersen and O. Grong: ‘Analytical modelling of grain growth in

metals and alloys in the presence of growing and dissolving precipi-tates—I. Normal grain growth’, Acta Metall. Mater., 1995, 43,2673–2688.

4. K. Banerjee, M. Militzer, M. Perez and X. Wang: ‘Non isothermalaustenite grain growth kinetics in a microalloyed X80 linepipe steel’,Metall. Mater. Trans. A, 2010, 41, 3161–3172.



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

5. J. Moon, J. Lee and C. Lee: ‘Prediction for the austenite grain size inthe presence of growing particles in the weld HAZ of Ti-microal-loyed steel’, Mater. Sci. Eng. A, 2007, 459, 40–46.

6. C. Henwood, M. Bibby, J. Goldak and D. Watt: ‘Coupled transientheat transfer—microstructure weld computations (part B)’, ActaMetall., 1988, 36, 3037–3046.

7. H. K. D. H. Bhadeshia, L. E. Svensson and B. Gretoft: ‘Amodel forthe development of microstructure in low-alloy steel (Fe–Mn–Si–C)weld deposits’, Acta Metall., 1985, 33, 1271–1283.

8. O. Grong: ‘Metallurgical modeling of welding’, Materials modelingseries; 1997, London, Institute of Materials.

9. W. Zhang, J. W. Elmer and T. DebRoy: ‘Integrated modelling ofthermal cycles, austenite formation, grain growth and decompo-sition in the heat affected zone of carbon steel’, Sci. Technol.Weld. Join., 2005, 10, 574–582.

10. M. Militzer: ‘Computer simulation of microstructure evolution inlow carbon sheet steels’, ISIJ Int., 2007, 47, 1–15.

11. M. Maalekian, R. Radis, M. Militzer, A. Moreau and W. J. Poole:‘In situ measurement and modelling of austenite grain growth in aTi/Nb microalloyed steel’, Acta Mater., 2012, 60, 1015–1026.

12. M. Militzer, A. K. Giumelli, E. B. Hawbolt and T. R. Meadowcroft:‘Austenite grain growth kinetics in Al-killed plain carbon steels’,Metall. Mater. Trans. A, 1996, 27, 3399–3409.

13. L. P. Zhang, C. L. Davis and M. Strangwood: ‘Effect of TiN par-ticles and microstructure on fracture toughness in simulated heat-affected zones of a structural steel’, Metall. Mater. Trans. A, 1999,30, 2089–2096.

14. K. Easterling: ‘Introduction to the physical metallurgy of welding’;1992, Oxford, Butterworth-Heinemann Ltd.

15. M. Eroglu and M. Aksoy: ‘Effect of initial grain size on microstruc-ture and toughness of intercritical heat-affected zone of a low carbonsteel’, Mater. Sci. Eng. A, 2000, 286, 289–297.

16. S. Kou: ‘Welding metallurgy’; 2nd ed.; 2003, Hoboken, NJ, JohnWiley & Sons Inc.

17. J. C. Ion, K. E. Easterling andM. F. Ashby: ‘A second report on dia-grams of microstructure and hardness for heat-affected zones inwelds’, Acta Metall., 1984, 32, 1949–1962.

18. H. K. D. H. Bhadeshia and L. E. Svensson: ‘Modelling the evol-ution of microstructure in steel weld metal’, in ‘Mathematical mod-eling of weld phenomena’ (eds. K. E. Easterling and H. Cerjak),109–182; 1993, London, Institute of Materials.

19. R. G. Thiessen, I. M. Richardson and J. Sietsma: ‘Physically basedmodelling of phase transformations during welding of low-carbonsteel’, Mater. Sci. Eng. A, 2006, 427, 223–231.

20. M. Gaudet: ‘The tensile properties and toughness of microstructuresrelevant to the HAZ of X80 Line pipe Steel Girth Welds’, PhDthesis, The University of British Columbia, Vancouver, BC,Canada, 2015.

21. F. Fazeli, R. Tafteh, M. Militzer and W. J. Poole: ‘Modeling theeffect of Nb on austenite decomposition in advanced steels’, Proc.Int. Symp. on ‘The recent developments in plate steels’, 2011,AIST, Warrendale, PA, 343–350.

22. F. Fazeli and M. Militzer: ‘Effect of Nb on austenite formation anddecomposition in an X80 linepipe steel’, J. Iron Steel Res. Int., 2011,18(1–2), 658–663.

23. J. M. Reichert, T. Garcin, M. Militzer and W. J. Poole: ‘Formation ofmatensite/austenite (M/A) inX80 linepipe steel’, Proc. 9th Int. PipelineConf., 2012, Calgary, Alberta, Canada, IPC 2012–90465.

24. J. M. Reichert and M. Militzer: ‘Quantifying the effect of Nb andMo on transformation products using advanced EBSD analysis’,HSLA Steels 2015, Microalloying 2015 & Offshore EngineeringSteels 2015, CSM/CAE/TMS, 2016, 143–149.

25. J. M. Reichert, M. Militzer, W. J. Poole and L. Collins: ‘A newapproach using EBSD to quantitatively distinguish complex trans-formation products along the HAZ in X80 linepipe steel’, Proc.

11th Int. Pipeline Conf., 2014, Calgary, Alberta, Canada, IPC2014–33668.

26. J. M. Reichert, Personal communication, 2015.27. M. Militzer, R. Pandi and E. B. Hawbolt: ‘Ferrite nucleation and

growth during continuous cooling’, Metall. Mater. Trans. A, 1996,27, 1547–1556.

28. M. Militzer, E. B. Hawbolt and T. R. Meadowcroft:‘Microstructural model for hot strip rolling of high-strength low-alloy steels’, Metall. Mater. Trans. A, 2000, 31, 1247–1259.

29. D. Liu, F. Fazeli, M. Militzer and W. J. Poole: ‘A microstructureevolution model for hot rolling of a Mo-TRIP steel’, Metall.Mater. Trans. A, 2007, 38, 894–909.

30. J. Lu: ‘Quantitative microstructural characterization of microal-loyed steels’, PhD thesis, University of Alberta, Edmonton, AL,Canada, 2009, 165–177.

31. C. Zener: ‘Grains, phases, interfaces: an interpretation of micro-structure’, Trans. AIME, 1948, 175, 15–51.

32. M. Perez: ‘Gibbs–Thomson effects in phase transformations’, Scr.Mater., 2005, 52, 709–712.

33. T. Gladman: ‘The physical metallurgy of microalloyed steels’, TheInstitute ofMaterials; 2002, London,UK, IOMCommunications Ltd.

34. S. Uhm, J. Moon, C. Lee, Y. Yoon, and B. Lee: ‘Prediction model forthe austenite grain size in the coarse grained heat affected zone ofFe–C–Mn steels: considering the effect of initial grain size on iso-thermal growth behavior’, ISIJ Int., 2004, 44, 1230–1237.

35. M. Hillert: ‘On the theory of normal and abnormal grain growth’,Acta Metall., 1964, 13, 227–238.

36. P. R. Rios: ‘Overview no. 62: A theory for grain boundary pinningby particles’, Acta Metall., 1987, 35, 2805–2814.

37. F. J. Humphreys and M. Hatherly: ‘Recrystallization and relatedannealing phenomena’, 2nd ed., 2004; Oxford, UK, Elsevier Ltd.

38. K. J. Irvine, F. B. Pickering and T. Gladman: ‘Grain refined C–Mnsteels’, J. Iron Steel Inst., 1967, 205, 161–182.

39. J. Ågren: ‘A revised expression for the diffusivity of carbon in binaryFe–C austenite’, Scripta Metall., 1986, 20, 1507–1510.

40. M. Militzer, F. Fazeli and H. Azizi-Alizamini: ‘Modelling of micro-structure evolution in advanced high strength steels’, La Metall.Ital., 2011, 4, 35–41.

41. A. Ali and H. K. D. H. Bhadeshia: ‘Nucleation of Widmanstättenferrite’, Mater. Sci. Technol., 1990, 6, 781–784.

42. F. Fazeli, T. Jia and M. Militzer: ‘Critical assessment of bainitemodels for advanced high strength steels’, Sol. St. Phen., 2011,172–174, 1183–1188.

43. N. Takayama, G.Miyamoto and T. Furuhara: ‘Effects of transform-ation temperature on variant pairing of bainitic ferrite in low carbonsteel’, Acta Mater., 2012, 60, 2387–2396.

44. S. Sarkar, A. Moreau, M. Militzer and W. J. Poole: ‘Evolution ofaustenite recrystallization and grain growth using laser ultrasonics’,Metall. Mater. Trans. A, 2008, 39, 897–907.

45. S. Kruger, G. Lamouche, J. P. Monchalin, R. Kolark, G. Jeskey andM. Choquet: ‘On-line monitoring of wall thickness and austenitegrain size on a seamless tubing production line at the TimkenCompany’, Iron Steel Technol., 2005, 2, 25–31.

46. K. Ueda, T. Garcin and M. Militzer: ‘In-situ measurement of auste-nite grain size through reverse transformation of low carbon steel’,Proc. Int. Conf. on ‘Solid-solid phase transformation in inorganicmaterials’, 2015, Whistler, BC, Canada, 681–682.

47. F. Hamad, L. Collins and R. Volkers: ‘Effect of GMAW procedureon the heat-affected zone (HAZ) toughness of X80 (grade 550) line-pipe’, Proc. 7th Int. Pipeline Conf., 2008, Calgary, Alberta, Canada,IPC 2008–64097.

48. X. Li, X. Ma, S. V. Subramanian, R. D. K. Misra and C. Shang:‘Structure–property–fracture mechanism correlation in heat-affected zone of X100 ferrite-bainite pipeline steel’, Metall. Mat.Trans. E, 2015, 2, 1–11.



Dow

nloa

ded

by [

The

Uni

vers

ity o

f B

ritis

h C

olum

bia]

at 0

9:25

04

Mar

ch 2

016

microstructure model for the heat-affected zone of x80...

Documents