98-s | APRIL 2001

A B S T R AC T. Nonmetallic inclusions intwo weld metals were characterized withrespect to variations in weld aluminumc o n c e n t ration. Two self-shielded fluxcored arc welding (FCAW-S) electrodeswere used to produce welds for optical,scanning and transmission electron mi-croscopy. The inclusions in the weld withhigh-aluminum concentration were pre-dominantly aluminum nitride. In con-trast, the inclusions in welds with low-aluminum and high-titanium concentra t i o n swere mostly aluminum oxide and titaniumcarbonitrides. The measurements werecompared with predictions from multi-phase, multicomponent thermodynamicequilibrium calculations. The calcula-tions agreed with the experimental mea-surements and predicted the formation ofaluminum nitride in high-aluminumwelds and also simultaneous formationof aluminum oxide and titanium car-bonitrides. However, the predicted vol-ume fractions were lower than experi-mental values.

Introduction

It is well known that nonmetallic in-clusions play an important role in theevolution of microstructures in steel weldmetals. They influence the partitioning ofalloying elements between solid solutionand second phases depending upon thetemperature of the formation. Also, theymay act as nucleation sites for solidifica-tion and solid-state transformations oncooling. Inclusions are also known tohave a direct effect on mechanical prop-

erties. If present in sufficient numbersand size, inclusions may also influencethe ductile-to-brittle Charpy transition byp r oviding initiation sites for cleava g ecracks and reduce upper-shelf energy.

Most of the documented work on in-clusions in weld metal has focused onwelding processes that shield the arc andmolten metal from atmospheric contam-ination [i . e . , gas metal arc welding( G M AW), gas-shielded flux cored arcwelding (FCAW-G), shielded metal arcwelding (SMAW) and submerged arcwelding (SAW)]. In conventional steelweld metals produced by theseprocesses, the oxidizing atmosphere pro-duced by the consumables and/or sup-plied by the shielding gas is accommo-dated by excess Mn and Si in theelectrode, wh i ch effectively deoxidizethe molten weld metal. Much of the de-oxidation product “floats out,” creatingsilicate islands on the surface of GMAwelds and contributing to the slag layerin FCAW-G, SMAW and SAW. Histori-cally, much experimental work was doneto ch a racterize the oxide inclusions inthese systems in relation to the resultingmicrostructures and properties (Refs.1–8). Recent advances in computationalmodels and analytical tools make it pos-sible, in some cases, to predict the deox-idation sequence and the oxide inclusionformation with reasonable accura cy

(Refs. 7–10). By contrast, relatively littleeffort has focused on weld metals inwhich nitride rather than oxide formationis dominant [i.e., self-shielded flux coreda rc welding (FCAW-S)] (Refs. 11–16).This work was undertaken to determinewhether the same analytical techniquescould be used to accurately predict in-clusion formation in FCAW-S deposits.

Weld metal produced by FCAW-S isunique in that the welding process andconsumables do not intentionally pro-tect the molten metal from atmosphericcontamination. Rather, such contamina-tion is anticipated and necessitates theuse of strong deoxidizers and denitridersto ensure deposition of sound weld de-posits. Inclusion of these elements (e . g. ,Al, Ti and Zr) results in weld metal ch e m-ical compositions that are significantlydifferent from other conventional arcweld deposits in the same strengthrange. As illustrated in Table 1, FCAW- Sdeposits have higher aluminum and ni-trogen in conjunction with lower oxygenthan other conventional arc weld de-posits. The alloy balance in terms of car-bon and manganese levels may also dif-fer in some cases, but the majordifferences are in the nitrogen and oxy-gen contents and the amount of excessdeoxidizer/denitrider remaining in theweld metal (Refs. 14, 15).

Experimental Approach

Two FCAW-S weld metal systemswere selected for investigation, E70T- 4and E71T-8 (Ref. 17), which represent sig-nificantly different Al, O and N levels aswell as alloy balance. Specifically, theseelectrodes represent the extremes of thetypical aluminum range for FCAW-S de-posits. Single V-groove welds were pro-duced over steel backing using the jointgeometry illustrated in Fig. 1.

The welding conditions summarizedin Table 2 are within the manufacturer’srecommended operating ranges for each

Inclusion Formation in Self-Shielded Flux Cored Arc Welds

BY M. A. QUINTANA, J. McLANE, S. S. BABU AND S. A. DAVID

An investigation was made to see whether the same analytical models used topredict oxide inclusion in weld metal can also be used to predict nitride formations

KEY WORDS

Self-Shielded Flux CoredOxide InclusionsNitride InclusionsNonmetallic InclusionsAluminum OxideAluminum Nitride

M. A. QUINTANA is with The Lincoln ElectricCo., Cleveland, Ohio. J. McLANE is withE ve r e a dy Battery Co., Cleveland, Ohio (for-merly with The Lincoln Electric Co.). S. S.BABU and S. A. DAVID are with Oak RidgeNational Laboratory, Oak Ridge, Tenn.

Paper presented at the AWS 80th AnnualMeeting, April 12–15, 1999, St. Louis, Mo.

of the electrodes. All welding was ac-complished in the flat (1G) position. Thetwo welds were made with significantlydifferent welding heat inputs necessi-tated by the respective electrode diame-ters and are representative of actualusage. It was not possible to obtain elec-trodes representing extremely high andextremely low aluminum levels in sizesthat would permit welding with the sameheat input. Although differences in thethermal cycle due to heat input influenceinclusion formation (Ref. 18), the largedifferences in chemical composition areexpected to overshadow the effect of dif-ferent heat inputs.

Transverse macrosections were takenfrom each weld for the experimentalwork. Bulk weld metal chemical compo-sitions were determined using a BAIRDModel DV 4 emission spectrometer andLECO analysis equipment. Samples forcarbon, sulfur and aluminum analyseswere taken by collecting chips afterdrilling at the same locations. Total alu-minum content was determined byatomic absorption spectroscopy follow-ing dissolution in aqua regia/hy d r o g e nfluoride and fuming in perchloric acid.Solid cylinders were removed fromequivalent locations in adjacent sectionsfor oxygen and nitrogen determinations.

M e t a l l o g raphic specimens were pol-ished with the final step with a 1-µm dia-mond and were examined withoute t ching in a light microscope at magnifi-cations up to 1000X. Subsequently,c a r b o n - e x t raction replicas and thin foilswere prepared for examination at highermagnifications. All metallographic sam-ple preparation utilized standard tech-niques with one exception. Because thelikelihood of aluminum nitride formationwas considered high and aluminum ni-tride is known to be soluble in water andmild alkali, it was necessary to eliminatethe use of all soap and water in the sam-ple preparation. Accordingly, polishedsamples were rinsed in either reagent-g rade methanol or toluene, and carbon-e x t raction replicas were rinsed in alcoholrather than wa t e r. Transmission electronm i c r o s c o py (TEM) was accomplishedusing Philips CM-12 and Philips CM200T

equipped with light element energy-d i s p e r s ive X-ray (EDS) analysis capability.Scanning electron microscopy was ac-complished using a JEOL 5800 and anA m ray 1645 with light element EDS.

Chemical compositions of inclusionswere determined semiquantitative l yusing EDS in both SEM and TEM. Inclu-sion number density, size distributionand volume fractions were determinedusing SEM images from the Amray at5000X. Data from ten randomly selectedframes for each weld deposit were col-lected, resulting in a sample size of about4300 square µm, 265 particles for thehigh-aluminum case and 492 particlesfor the low-aluminum case. The SEM im-ages from extraction replicas made itpossible to determine the inclusion sizeand shape more accurately than wouldhave been possible with just the polishedcross sections. Image analysis softwa r ewas used to collect statistics on wh a tamounted to two-dimensional projec-tions of three-dimensional particles withrelatively complex geometric configura-tions. In conventional C-Mn weld metalsystems that produce generally sphericalparticles, it is customary to compile in-clusion statistics based on diameter. Be-cause of the more complex shapes en-

countered in the two FCAW-S systemsconsidered here, the maximum dimen-sions and equivalent diameters wereconsidered more relevant. “Equiva l e n tdiameter” is the diameter of a circle of asize equivalent to the area of complexshape. Consequently, inclusion vo l u m ewas determined by calculating the vol-ume of an equivalent sphere. The totalarea sampled for each weld multiplied bythe respective mean equivalent inclusiondiameter was considered a reasonableestimate of the volume of material sam-pled. Volume fractions were estimated bydividing total inclusion volume by sam-ple volume.

Thermodynamic Calculations

Th e r m o dynamic equilibrium amongvarious oxides, nitrides and liquid steelwas calculated using version L of Th e r-moCalc™ software (Ref. 19). The calcu-lations considered the elements Fe, Al, C,Mn, Si, Al, Ti, O and N and the follow i n gphases: liquid, δ-ferrite, austenite, ce-mentite, Al2O3, MnOA l2O3, Ti O, Ti O2,Ti2O3, Ti3O5, MnOTi O2, SiO2, MnO-S i O2, MnO, AlN and Ti(CN). In order toestimate chemical compositions, vo l u m ef ractions and initial temperatures of for-

WELDING RESEARCH SUPPLEMENT | 99-s

Fig. 1 — Weld joint geometry.

Table 1 — Weld Metal Chemical CompositionComparison (wt-%)

Element SMAW FCAW-G FCAW-SE7018 E70T-1 E7XT-X

C <0.08 0.03–0.08 <0.4Mn 1.2–1.5 1.3–1.7 0.5–1.2Si 0.2–0.5 0.6–0.9 0.2–0.5Al 0.01 <0.2 0.5–1.8N <0.01 <0.01 ~0.05O ~0.040 ~0.070 0.005–0.040

Table 2 — Welding Conditions Used in This Investigation

High-Aluminum Low-AluminumE70T-4 E71T-8

Electrode diameter 0.120 (3) 0.078 (2.0)[in. (mm)]Tip-to work distance 2.75 (70) 0.75 (19)[in (mm)]Voltage 31–36 19–20Amperage 525–590 270–280Wire feed rate 225–250 (95–106) 120 (51)[in./min (mm/s)]Heat input 86–98 (3.4–3.8) 54–58 (2.1–2.3)[kJ/in. (kJ/mm)]Preheat ambient ambientInterpass (see note) 163 (325) max. 163 (325) max.[°C (°F)]

Note: Test welds heat quickly, reaching maximum interpass temperature in 1–2 passes. Thereafter, interpass temperature wasmaintained at 150–163 °C (300–325°F) through completion of the test welds.

mation of the nonmetallic phases in eachof the two weld metal systems were cal-culated at a temperature just before solid-ification. It was assumed the ch e m i c a lcomposition of the molten pool at thist e m p e rature could be approximated bythe final weld deposit composition.While calculating the volume fraction ofoxides and nitrides, the compounds thatformed during cooling from initial tem-p e rature of formation to 1527°C (~ melt-ing point of the welds) were added to-g e t h e r. Moreove r, the changes in liquidcomposition due to the formation of thesecompounds were also considered usingS cheil assumptions. Further details abouts u ch calculations can be found in Ref. 20.

Results and Discussion

The chemical test results for the twoF CAW-S deposits are summarized inTable 3. In the high-aluminum case, theE 7 0 T-4 electrode produced weld metalaluminum at 1.70 wt-% with oxygen andnitrogen levels of 60- and 640-wt ppm,r e s p e c t ive l y. The low-aluminum weldmetal produced with the E71T-8 elec-trode had far less aluminum at 0.53% byweight with oxygen and nitrogen leve l sof 300- and 330-wt ppm, respective l y.Also, the presence of 0.058 wt-% tita-nium in the low-aluminum depositindicates the use of a second deoxi-

dizer/denitrider in the E71T-8 electrode.While the evolution of microstructures

in these two weld metals is discussed indetail elsewhere (Ref. 21), the opticalm i c r o g raphs of as-welded microstruc-ture from both weld deposits are pre-sented in Fig. 2 to supplement the inclu-sion comparisons. It is apparent that thedifferences in chemical composition af-fect more than the inclusion formation.The micrographs show a large fraction ofcolumnar skeletal δ-ferrite morphologyin high-aluminum welds. The presenceof δ-ferrite is attributed to the sluggishaustenite formation in these welds dur-ing weld cooling. The sluggish austeniteformation is related to the increase in thestability of δ-ferrite, owing to a large con-c e n t ration of aluminum in solid solution.In contrast, the low-aluminum welds h owed classic α-ferrite morphology,wh i ch formed from decomposition ofaustenite. These observations are consis-tent with earlier work by Ko t e cki (Refs.15, 16).

Thermodynamic Predictions

High-Aluminum Deposit (E70T-4)

In the case of the E70T-4 deposit, theThermoCalc™ predictions favored theformation of AlN over Al2O3 or Ti(CN).Fig. 3A shows the volume fraction of AlNas a function of temperature while cool-ing. The initiation of AlN formation fromliquid steel was predicted to occur at1666°C. No other nonmetallic phases

were predicted to form. To evaluate the relative stability of ox-

ides and nitrides in this weld metal sys-tem, the stability diagrams for liquid steelin equilibrium with Al2O3, AlN or Ti(CN)were calculated at 1527°C and areshown in Fig. 3B. This calculation is con-sistent with multicomponent, but wa slimited to two phases at a time. The nom-inal concentrations of Al, O, Ti and N ofthe weld metal are also shown in the di-a g ram. The calculations clearly showthat for this weld metal only AlN willform. However, the stability diagram alsoshows some interesting features. The for-mation of Al2O3 is predicted to occuronly below certain Al and O concentra-tions, and an inadvertent increase in Alwill not promote Al2O3 formation. Thiscomplex stability of Al2O3 is attributed tothe interaction energies of Al and O inliquid steel. The stability diagrams alsoshow minimum concentrations of Al, Tiand N are needed for the initiation of AlNand/or Ti(CN) phases in the liquid.

The above calculations consideredeach phase in isolation. However, it is de-s i rable to evaluate the competition be-tween each phase as a function of one ormore alloying element concentra t i o n s .Therefore, the calculations were ex-tended to predict the volume fraction ofvarious phases as a function of aluminumconcentration only in Fig. 3C at 1527°C.The calculations are consistent with mul-ticomponent, multiphase equilibrium.The calculations indicated that below~0.5 wt-% Al concentration, the forma-tion of AlN will cease and Al2O3 forma-tion will be favored while other elemen-tal concentrations remain constant.

Low-Aluminum Deposit (E71T-8)

Similar calculations were performed

100-s | APRIL 2001

Fig. 2 — Optical micrographs of the as-welded mirostructure. A — High-aluminum weld deposit (E70T-4); B — low-aluminum weld deposit( E 7 1 T- 8 ) .

A B

Table 3 — Chemical Test Summary

Weld C S P Mn Si Al Ni Ti O N

High-aluminum 0.234 <0.003 0.011 0.50 0.28 1.70 0.02 0.003 0.006 0.064E70T-4Low-aluminum 0.149 <0.003 0.005 0.64 0.30 0.53 0.01 0.058 0.030 0.033E71T-8

for low-aluminum weld deposits. In thecase of the E71T-8 deposit, Th e r m o C a l c ™f avored the formation of aluminum oxideand titanium carbonitride over the forma-tion of aluminum nitride. Figure 4As h ows the fraction of AlN and Ti(CN) as afunction of temperature. The Al2O3 f o r-mation was predicted to start at 1809°Cwith Ti(CN) beginning at 1654°C.

Stabilities of oxides and nitrides in thisweld metal system were evaluated at1527°C and are compared in Fig. 4B. Th enominal concentrations of Al, O, Ti and Nof the weld metal are also shown in thed i a g ram. The stability diagrams for Al2O3and AlN are more or less similar to that ofthe high-aluminum weld. The calcula-tions show that, for this weld metal, both

A l2O3 and Ti(CN) can form at 1527°C.H ow e ve r, the formation of AlN is not fa-vored due to low concentrations of alu-minum and nitrogen. Because both Al2O3and Ti(CN) form at 1527°C, the probabil-ity of each acting as a heterogeneous sitefor the other is high. The calculations alsos h owed the stability of Ti(CN) in this weldmetal is quite different from that shown in

WELDING RESEARCH SUPPLEMENT | 101-s

Fig. 3 — Equilibrium thermodynamic calculations for high-aluminumweld (E70T-4). A — Variation of volume fraction of AlN with temper-ature while weld cooling; B — stability diagrams for AlN, Al2O3 andTi(CN) phases as a function of aluminum, titanium, oxygen and nitro-gen concentrations at 1527°C; C — variation of volume fraction ofAl2O3 and AlN as a function of aluminum concentration at 1527°C.

A

C

B

A

B

C

Fig. 4 — Equilibrium thermodynamic calculations for low-aluminumweld (E71T-8). A — Variation of volume fraction of Al2O3 and Ti(CN)with temperature while weld cooling; B — stability diagrams for AlN,Al2O3 and Ti(CN) phases as a function of aluminum, titanium, oxygenand nitrogen concentrations at 1527°C; C — variation of volume frac-tion of Al2O3, Ti(CN), AlN and Ti2O3 as a function of aluminum con-centration at 1527°C.

Fig. 3B. This is attributed to the intera c t i o nenergies between dissolved titanium, ni-trogen and oxygen. It is notewo r t hy thate ven the formation of Al2O3 is very closeto the stability line and that a smallchange in aluminum concentration caneliminate the Al2O3 formation, favo r i n gonly Ti(CN) formation.

Calculations were extended to pre-dict the volume fractions of va r i o u sphases as functions of aluminum con-c e n t ration, considering all the phases at1527°C — Fig. 4C. Some interesting fea-tures can be observed in this calculation.At very low levels of aluminum, the cal-culations predicted the formation ofTi2O3. With an increase in aluminum thevolume fraction of Al2O3 i n c r e a s e s .H ow e ve r, at 0.7 wt-% aluminum the for-mation of Al2O3 ceases. The formationof aluminum nitride starts only above 2wt-%. This interval of the absence of alu-minum reaction is an interesting obser-vation, and further work is needed toe valuate this composition experimen-t a l l y. Interestingly, the formation ofTi(CN) was found to occur in all alu-minum concentrations, wh i ch needs tobe evaluated with future experiments.

Microstructural Observations

High-Aluminum Deposit (E70T-4)

Initial microscopic examination re-vealed a relatively uniform distribution ofcoarse, faceted particles, many withcomplex shapes, as illustrated in Fig. 5.On closer examination at higher magni-fications, the complex shapes were actu-ally large agglomerations (Fig. 6) of

smaller particles (Fig. 7). EDS verified thepresence of Al and N in the inclusions(Fig. 7C), indicating that they are, indeed,aluminum nitride and supported the pre-dictions from thermodynamic calcula-tions. It is important to note that no ex-perimental evidence was found foraluminum oxide formation in this weld.The large agglomerations and the rela-tively uniform distribution of smaller par-ticles suggest initial formation of these in-clusions in the liquid prior to the start ofsolidification. This is consistent with apredicted inclusion-formation start tem-perature above the liquidus — Fig. 3A. Inaddition, the AlN inclusions were highlyfaceted — Figs. 6, 7.

It was apparent from detailed TEM ob-servations of the Fe-rich cores found insome of the smaller discrete particles thatthe aluminum nitrides had nucleated het-erogeneously — Fig. 8. Specific identifi-cation of these Fe-rich particles proved tobe rather difficult. They were first ob-s e r ved in the TEM images of the ve r ysmall inclusions in the carbon extractionreplicas, appearing as dark spots near the

centers of the hexagonal inclusions —Fig. 8A. EDS indicated strong Fe peaksassociated with these particles, wh i chwere absent at the inclusion perimeters— Fig. 8B. EDS in the SEM indicated thepresence only of aluminum and nitrogen.No Fe-rich phases were found on the ex-terior surfaces of the inclusions, suggest-ing the Fe-rich particles observed by theTEM were internal, serving as heteroge-neous nucleation sites. Although thepresence of Fe in the nitride inclusionswas not predicted, it is consistent withother reported results (Ref. 11). Becausethe formation of iron compounds was notanticipated for either of the two weldmetal systems under consideration, itwas not included in the thermodynamicanalysis, which considered only the ox-ides and nitrides of aluminum and tita-nium. Therefore, these Fe - r i ch particlesare assumed to be the unmelted Fe-richcompounds. Further work is needed tounderstand the formation of these com-pounds during self-shielded flux-coredarc welding.

Simple statistical analysis of the in-

102-s | APRIL 2001

Fig. 5 — Optical micrograph of unetched high-aluminum weld( E 7 0 T- 4 ) .

Fig. 6 — SEM micrograph of high-aluminum weld (E70T-4).

Table 4 — Inclusion Summary Statistics

Weld Mean Maximum Max. Number Volume ElementalEquivalent Equivalent Dimension, Density Fraction(a) CompositionDiameter, Diameter µm (N), m–3

dmean, µm dmax, µm

High-aluminum 0.93 5.0 8.5 6.627 x 1016 0.0279 Al, NE70T-4 some FeLow-aluminum 0.37 2.5 4.0 3.092 x 1017 0.0080 Al, OE71T-8 Ti, N

(a) Estimated volume fraction = N*(4/3)*∏*(dmean/2)3

clusions from both welds are summa-rized in Table 4, and the size distributionsare plotted in Fig. 9. In the case of high-aluminum deposit E70T-4, average inclu-sion size is on the order of 1 µm, with thegreatest number of inclusions falling inthe 1- to 2-µm range. The relatively smallf r e q u e n cy of very small inclusions (<1µm) is consistent with the clustering andagglomeration seen experimentally, giv-ing the impression of larger indiv i d u a lparticle size. The maximum equiva l e n tdiameter is 5 µm. However, because ofthe complex shapes caused by the ag-glomeration of large numbers of individ-ual inclusions, the maximum dimensionis greater than 8.5 µm. The volume frac-tion of inclusions in this high-aluminumdeposit was estimated experimentally tobe on the order of 0.0279, an order ofmagnitude higher than the thermody-namic prediction of 0.0037 — Fig. 3A. Inthis case, the ThermoCalc™ prediction isnot a reasonable estimate of the experi-mental result, perhaps because theanalysis did not consider other com-pounds or reactions during solidification.For instance, the residual concentrationsof dissolved Al and N at 1527°C after theScheil calculations were found to be 1.61and 0.016 wt-%, respective l y. This dis-solved aluminum and nitrogen can reactto form AlN, even after the completion ofsolidification. However, further implica-tions of slow diffusivity of elements insolid need to be evaluated further.

Low-Aluminum Deposit (E71T-8)

Initial microscopy revealed a rela-tively uniform distribution of finer spher-ical particles — Fig. 10. Subsequent ex-amination of the extraction replicas athigher magnifications in the SEM re-vealed the presence of more complexshapes, although not to the same level ofcomplexity observed previously for thehigh-aluminum (E70T-4) deposit. Fr o m

EDS analysis, Ti, Al, N, C and O were ver-ified to be present. It was not possible todetermine whether the C peaks resultede x c l u s ively from the carbon extra c t i o nfilm or if the inclusions were responsiblefor some of the variations in the peak in-tensity observed. No other chemical con-stituents were associated with the inclu-sions. The spherical inclusions appearedto be rich in Al and O at the surface withisolated areas of Ti and N — Fig. 11A.The more complex shapes tended to berich in Ti and N with cubic facets and agrowth habit in three orthogonal direc-tions — Fig. 11B. The experimental evi-dence demonstrates heterogeneous nu-cleation of one phase on another. Th eapparent precipitation of aluminumoxide on the titanium nitride is not con-sistent with the prediction of Ti(CN) for-mation at a temperature ~150°K low e rthan Al2O3 formation. However, the factthat the stable temperature ranges fore a ch phase overlap considerably (Fi g .4A) suggests each phase could precipi-

tate on the other during cooling. The statistics summarized in Table 4

and Fig. 12 are consistent with the ob-servation that the low-aluminum (E71T-8) deposit exhibits a finer inclusion dis-tribution than the high-aluminum( E 7 0 T-4) deposit — Fig. 9. The ave ra g eequivalent diameter, maximum diameterand maximum dimension are approxi-mately a factor of two lower than similarstatistics for the high-aluminum deposit.The frequency plot shows the greatestnumber of inclusions are less than 1 µmcompared with the 1 to 2 µm for the high-aluminum case. The higher number den-sity in this case is consistent with a uni-form distribution of finer, more discreteparticles. The estimated volume fractionof 0.0082 is higher than the predictedvalue of 0.0026. This value was obtainedby adding the volume fractions of bothAl2O3 and Ti(CN). This underestimationis again related to the inability of ther-modynamic calculations to consider thereactions at low temperature in the solid

WELDING RESEARCH SUPPLEMENT | 103-s

Fig. 7 — High-aluminum weld (E70T-4): individual inclusion in carbon extraction replicas. A —SEM image showing faceted nature of AlN; B — morphology observed in TEM micrographs; C— EDS spectrum obtained from the AlN inclusion.

A B

C

state. However, thermodynamic predic-tions agree with the trends observed inthe experimental data.

From the results presented above, it isapparent thermodynamic calculationsagree reasonably well with the experi-mentally observed deoxidation and den-itriding conditions in self-shielded fluxcored arc welds. The calculations pre-dict the formation of AlN in high-aluminum welds and the formation ofA l2O3 and Ti(CN) in low - a l u m i n u mwelds. The calculations predicted theexpected trend of decreasing inclusionvolume fraction in the low - a l u m i n u mweld compared with that of the high-alu-minum weld. How e ve r, the calculationsdo not agree with the magnitudes of vo l-ume fra c t i o n s .

The decision to neglect the influenceof heat input and reaction kinetics doesnot explain the consistent underpredic-tion of inclusion volume fraction. If any-thing, the nonequilibruim conditionsprevalent during weld solidification areexpected to lead to over prediction of in-clusion volume fraction when predic-tions are based exclusively on thermody-namic equilibria. It is more likely the

errors in prediction arisebecause inclusion forma-tion after solidification wasnot considered in theanalysis. Further, any com-parison of heat input va r i-ance between the two weldtypes studied would havelittle relevance consideringthe inclusions are of totally differentchemical compositions.

Summary and Conclusions

The chemical constituents identifiedin the inclusions from both the E70T- 4and E71T-8 weld metals are consistentwith the thermodynamic calculations.Although it was not possible to experi-mentally verify temperatures at wh i chprecipitation occurred, the agglomera-tion of inclusions in the E70T-4 weldmetal was consistent with the predictionof initial formation in the liquid.

The calculations predicted ch e m i c a lconstituents with reasonable accura cy ;h ow e ve r, the comparisons with experi-mental measurements of volume fractionwere only fair. The iron-rich phases at the

centers of the aluminum nitride inclu-sions were not predicted because theirformation was not anticipated and there-fore were not included in the analysis.This oversight and inability to considerthe inclusion formation at low tempera-ture may have contributed to the fact thataccurate thermodynamic estimates of in-clusion volume fraction were not consis-tently obtained. Although the predictionfor the low-aluminum weld is in fairagreement with experimental results, theprediction for the high-aluminum weldwas not; volume fraction was underesti-mated by an order of magnitude.

These initial results are by no meanscomplete; however, they suggest thermo-dynamic predictions can provide reason-able estimates of inclusion composition,if not volume fraction, in these FCAW-S

104-s | APRIL 2001

Fig. 8 — A — High-aluminum weld (E70T-4), AlN with Fe-rich areas; B — EDS spectrum obtained from Fe-rich areas.

Fig. 9 — Inclusion size distributions from high-aluminum weld.

Fig. 10 — Typical inclusions of low-aluminum weld (E71T-8).

A B

weld metals. The thermodynamic calcu-lations can be a useful tool in under-standing inclusion development in weldmetals, particularly in assessing the rela-t ive stability of different phases duringcooling and solidification.

Acknowledgments

R e s e a rch sponsored by The LincolnElectric Co. and the U.S. Department ofE n e r g y, Division of Materials Sciences, andAssistant Secretary for Energy Efficiencyand Renewable Energy, Office of IndustrialTe chnologies, Metals Processing Labora-tory User Center (MPLUS), Advanced In-dustrial Materials Program under contra c tD E - AC05-96OR22464 with Lock h e e dMartin Energy Research Corp.

References

1. Hill, D. C., and Passoja, D. E. 1974. Un-derstanding role of inclusions and microstruc-ture in ductile fracture. Welding Jo u r n a l53(11): 481-s to 485-s.

2. Widgery, D. J. 1976. Deoxidation prac-tice for mild-steel weld metal. Welding Jour-nal 55(3): 57-s to 67-s.

3. Abson, D. J., Dolby, R. E., and Hart, P.H. M. 1978. Investigation into the role of non-metallic inclusions on ferrite nucleation incarbon steel weld metals. TWI Res. Rep.67/1978/M.

4. Ito, Y., Nakanishi, M., and Komizo, Y.1982. Effect of oxygen on low carbon steelweld metal. Metal Construction (9): 472–478

5. Abson, D. J. 1989. The influence of alu-minum additions to tubular cored weldingelectrodes on the microstructure and tough-ness of as-deposited C-Mn-Ni weld metal.TWI Res. Rep. 407/1989.

6. Liao, F-C., and Liu, S. 1992. Effect of de-oxidation sequence on carbon m a n g a n e s esteel weld metal microstructures. We l d-ing Jo u r n a l 71(3): 94-s to 103-s.

7. Kluken, A. O., and Grøng, Ø. 1989.

M e chanisms of inclu-sion formation in Al-Ti-Si-Mn deoxidized steelweld metals. M e t a l l u r g i-cal Transactions A 20A(8):1335–1349.

8. Kluken, A. O. ,Grøng, Ø., and Rørvik,G. 1990. Solidificationmicrostructures andp h a s e - t ra n s f o r m a t i o n sin Al-Ti-Si-Mn deoxi-dized steel weld metals.Metallurgical Tra n s a c-t i o n s A 21A(7):2 0 4 7 – 2 0 5 8 .

9. Babu, S. S.,D avid, S. A., Vitek, J.M., Mundra, K., andD e b R oy, T. 1995. De-velopment of macroand microstructures of C–Mn low alloy steelwelds — inclusion formation. Materials Sci-ence and Technology 11(2):186–199.

10. Babu, S. S., David, S. A., and DebRoy,T. 1996. Coarsening of oxide inclusions in lowalloy steel welds. Science and Technology ofWelding and Joining 1(1): 17–27.

11. Krivenko, L. F. 1967. Research into thenitrides in the weld metal when steel is weldedby open arc process. Avto. Svarka (7):6–9.

12. Dorling, D. V., Rodriguez, P. E. L. B.and Rogerson, J. H. 1976. A comparison of thetoughness of self-shielded arc and submergedarc weld metals in C-Mn-Nb steels, part 1: ef-fect of consumables and process variables onweld metal toughness. Welding and MetalFabrication (7/8): 419–423.

13. Dorling, D. V., Rodriguez, P. E. L. B.,and Rogerson, J. H. 1976. A comparison of thetoughness of self-shielded arc and submergeda rc weld metals in C-Mn-Nb steels, part 2:through thickness toughness variation. Weld-ing and Metal Fabrication (9): 479–481.

14. Kotecki, D. J., and Moll, R. A. 1970. Atoughness study of steel weld metal from self-shielded flux cored electrodes — part 1. Weld-ing Journal 49(4): 157-s to 165-s.

15. Kotecki, D. J., and Moll, R. A. 1972. A

toughness study of steel weld metal from self-shielded flux cored electrodes — part II. Weld-ing Journal 51(3): 138-s to 155-s.

16. Es-Souni, M., Beaven, P. A., and Evans,G. M. 1992. Microstructure and AEM studiesof self-shielded flux cored arc weldments.Welding Journal 71(2): 35-s to 45-s.

17. American Welding Society. 1995.AWS A5.20, Specification for Carbon SteelElectrodes for Flux Cored Arc Welding.

18. Babu, S. S., Reidenbach, F., David, S.A., Böllinghaus, and Hoffmeister, H. 1999. Ef-fect of high energy density welding process oninclusion and microstructure formation insteel welds. Science and Technology of Weld-ing and Joining 4(2): 63–73.

19. Sundman, B., Jansson, B., and Anders-son, J. O. 1985. The thermo-calc databank sys-tem. Calphad 9(2): 1–153.

20. Hsieh, K. C., Babu, S. S., Vitek, J. M.,and David, S. A. 1996. Calculation of inclu-sion formation in low alloy steel welds. Mate-rials Science and Engineering, A215, 84–91.

21. Babu, S. S., David, S. A., and Quintana,M. A. 2001. Modeling microstructure evolu-tion in self-shielded flux-cored arc welds.Welding Journal 80(4): 91-s to 97-s.

WELDING RESEARCH SUPPLEMENT | 105-s

Fig. 11 — Inclusions of low-aluminum weld (E71T-8) observed with SEM.

Fig. 12 — Inclusion size distribution from low-aluminum weld.