180-s | MAY 1999

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

A martensite boundary, based upon magnetic measurements andlongitudinal face bend tests, is proposed for the WRC-1992 diagram

ABSTRACT. The upper martensiteboundary line from the Schaeffler dia-gram for stainless steel weld metals canbe transposed to the WRC-1992 dia-gram. However, magnetic measurementsand longitudinal face bend tests ofweld metals do not show good corre-spondence to this boundary. Many de-posit compositions of lower chromiumand nickel equivalents than thosealong Schaeffler’s transposed martensiteboundary show no martensite as-deposited and pass a 2T longitudinal facebend test. Based upon magnetic mea-surements and bend tests of about 100weld metal compositions, obtained assingle-pass submerged arc deposits onASTM A36 mild steel, a new martensiteboundary is proposed for addition to theWRC-1992 diagram. This boundary sep-arates compositions that exhibit no mag-netic response attributable to martensiteand pass a 2T longitudinal face bend testfrom compositions that have magneticresponse that indicates the presence ofas-deposited martensite and fail the bendtest. Because manganese is part of nei-ther the chromium equivalent nor thenickel equivalent on the WRC-1992 dia-gram, the line is specific to the Mn levelconsidered in the tests — approximately1% — that is suitable for most stainlesssteel cladding and dissimilar metal join-ing situations. It is probably conservativefor deposits of much higher Mn content.

Introduction

The WRC-1992 diagram (Ref. 1) forstainless steel weld metals has been rec-ognized by the International Institute ofWelding (IIW) as the most accurate andpreferred constitution diagram for esti-mating or predicting ferrite in nominally

austenitic and duplex ferritic-austeniticstainless steel weld metals (Ref. 2). As aresult, it was incorporated into the ASMEBoiler and Pressure Vessel Code with itsWinter 1994 Addendum. In addition tobetter predicting accuracy (Ref. 3) thanthe DeLong diagram (Ref. 4), the WRC-1992 diagram expands the predictingrange to 100 FN maximum. However,the much older Schaeffler diagram(Ref. 5), as shown in Fig. 1, continues tobe used for predicting ferrite in claddingand dissimilar-metal joining (Ref. 6), inlarge part because it includes boundariesfor martensite appearance in stainlesssteel weld deposits. In significant quanti-ties, martensite is often undesirable in astainless steel cladding or in a dissimilarmetal joint because its usual low ductil-ity tends to result in fracture during bendtesting of the weldment. In these cases,the Schaeffler diagram then provides atool for selecting filler metal to avoidmartensite in the weld metal (Ref. 6).

The Schaeffler diagram makes its pre-dictions in terms of “% ferrite,” but thereis no reproducible method of determin-ing % ferrite in weld metal. Round robintests within the Welding Research Coun-cil and within Commission II of the IIWshowed % ferrite values ranging from 0.6to 1.6 times the average value (Ref. 7)

(i.e., –40% to +60%) for a given sample,which is clearly unacceptable for speci-fications. On the other hand, in similarround robins, magnetic Ferrite Numbermeasurements showed interlaboratoryscatter of ±10% or less (about the aver-age value for a given sample), aftercalibration of instruments with primary(coating thickness) standards and of±14% or less after calibration of instru-ments with secondary (weld-metal-like)standards (Refs. 8, 9). So it is much moreattractive to use Ferrite Numbers, ratherthan “% ferrite,” in estimation and pre-diction of ferrite.

To expand the utility of the WRC-1992 diagram, it is of interest to deter-mine a martensite boundary for thatdiagram — the objective of the presentwork. The Schaeffler diagram shown inFig. 1 contains regions labeled “A + M”and “A + M + F,” in which martensite isexpected. The regions are boundedabove and below by a pair of diagonallines that proceed from upper left tolower right. It is the upper of these twolines that separates the “Austenite” re-gion from the “A + M” region along theleft-most part of the line and separates the“A + F” region from the “A + M + F”region along the right-most part of theline, which is the line of interest. Forcompositions above this line, no mar-tensite would be expected, while forcompositions below this line, somemartensite would be expected. It is thisline that is termed Schaeffler’s uppermartensite boundary.

Transposing Schaeffler’s UpperMartensite Boundary to theWRC-1992 Diagram

A logical way to begin to develop amartensite boundary for the WRC-1992diagram is to transpose the boundaryfrom the Schaeffler diagram. However,this is not a straightforward exercise,

D. J. KOTECKI is with The Lincoln ElectricCompany, Cleveland, Ohio.

A Martensite Boundaryon the WRC-1992 Diagram

BY D. J. KOTECKI

KEY WORDS

MartensiteStainless SteelWeld MetalWRC-1992 DiagramCladdingDissimilar Metal Joining

WELDING RESEARCH SUPPLEMENT | 181-s

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

because the Schaeffler chromium equiv-alent (Creq) is not the same as the WRC-1992 chromium equivalent, and neitheris the Schaeffler nickel equivalent (Nieq)the same as the WRC-1992 nickel equiv-alent. These equivalents, in which eachelement’s concentration is given inweight percent (a convention used every-where herein), are listed below:

Chromium EquivalentsSchaeffler:

% Cr + % Mo + 1.5x(% Si)+ 0.5x(% Nb)

WRC-1992:% Cr + % Mo + 0.7x(% Nb)

Nickel EquivalentsSchaeffler:

% Ni + 30x(% C) + 0.5x(%Mn)

WRC-1992:% Ni + 35x(% C) + 20x(% N)+ 0.25x(% Cu)

Because these equivalents are different,the method of transposing the martensiteboundary involves six steps, as follows:

Step 1. Fix the levels of the elements thatare different in the two sets of correspond-ing equivalent values. Since the transposedmartensite boundary was to be tested byusing single-pass submerged arc deposits,the elements to be fixed were chosen atlevels representative of the experimentaldata that followed. These levels are

C 0.10%Mn 1.00%N 0.02%Mo 0.00%Si 0.50%Nb 0.00%

Step 2. Arbitrarily choose two pointsalong Schaeffler’s upper martensiteboundary. These points are the pointwhere the martensite boundary intersectsthe Schaeffler 100% ferrite line and thepoint along the martensite boundarywhere the nickel equivalent is 15.00%.

Step 3. Use the Ferrite Predictor (Ref. 10)computer software to determine theSchaeffler Creq and Nieq correspondingto each of the two arbitrarily chosenpoints on Schaeffler’s upper martensiteboundary. Essentially by trial and errorwith the software, the coordinates ofthese two arbitrary points on the Schaef-fler diagram were found to be

Martensite boundary at 100% ferrite:Creq = 26.205Nieq = 4.536

Martensite boundary at 15.00% Nieq:Creq = 13.046

Nieq = 15.000

Step 4. Use these two pairs of Creq andNieq values, with the fixed values forother elements in step 2, to back calcu-late corresponding % Cr and % Ni levelsfrom the Schaeffler diagram. Calculationof the % Cr corresponding to each of thetwo chromium equivalents is straightfor-ward. Since Mo and Nb are presumed tobe zero, the only operation is to subtract1.5 times the chosen silicon content fromthe Schaeffler chromium equivalent fromstep 3 to find the % Cr

Fig. 1 — Schaeffler diagram.

Fig. 2 — WRC-1992 diagram, with four options for upper martensite boundaries transposed fromthe Schaeffler diagram.

Fig. 3 — Two longitudinal face bend test speci-mens. The specimen to the rear shows no cracks,while that to the front has numerous cracks.

182-s | MAY 1999

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

Martensite boundary at 100% ferrite:% Cr = Creq – 1.5x(% Si)

= 26.205 – 1.5x(0.5) = 25.455

Martensite boundary at 15.00% Nieq:% Cr = Creq – 1.5x(% Si)

= 13.046 – 1.5x(0.5) = 12.296

The % Ni calculations offer some op-tions, depending upon how nitrogen istreated. As originally published, theSchaeffler diagram did not consider ni-trogen. However, it has been realizedthat the Schaeffler nickel equivalent canbe adjusted for nitrogen.

Option 1. If nitrogen is ignored, thecalculations of % Ni for the two arbitrarypoints are as follows:

Martensite boundary at 100% ferrite:% Ni = 4.536 – 30x(0.10% C)

– 0.5x(1.0% Mn) = 1.036

Martensite boundary at 15.00% Nieq:% Ni = 15.00 – 30x(0.10% C)

– 0.5x(1.0% Mn) = 11.50

Option 2. The Schaeffler diagram isbased entirely upon deposited weldmetal from covered electrodes. TheASME Code (Ref. 11) indicates a nitrogen

content of 0.06% is considered normal.Although not formalized, a commonpractice is to adjust the Schaeffler nickelequivalent of a weld that departs fromthis nitrogen value by the factor ∆Nieqgiven below:

Adjusted Nieq = Schaeffler Nieq + ∆Nieq

where ∆Nieq = 30x(% N – 0.06) andwhere the coefficient 30 is taken from theDeLong diagram (Ref. 4) coefficient fornitrogen. For low nitrogen weld metal,this factor can be negative, and ∆Nieq is–1.20 in the case of the nitrogen contentassumed in step 1 above. Applying thisapproach to find the nickel content at thetwo arbitrarily chosen points alongSchaeffler’s upper martensite boundaryinvolves simply subtracting the ∆Nieqfrom the nickel equivalent from option 1above, and produces the following:

Martensite boundary at 100% ferrite:% Ni = 1.036 – ∆Nieq

= 1.036 – (–1.20) = 2.236

Martensite boundary at 15.00% Nieq:% Ni = 11.50 – ∆Nieq

= 11.50 – (–1.20) = 12.70

Option 3. Espy (Ref. 12) proposed thatSchaeffler’s nickel equivalent be modi-fied by adding a different correction fac-tor to it, based upon his experimentaldata as follows:

Adjusted Nieq = Schaeffler Nieq + ∆Nieq

where ∆Nieq = 30x(% N – 0.045) =30x(0.02 – 0.45) = –0.75 in this case. Ap-plying this approach and subtracting thisnew ∆Nieq from the Nieq from option 1above produces the following:

Martensite boundary at 100% ferrite:% Ni = 1.036 – ∆Nieq

= 1.036 – (–0.75) = 1.786

Martensite boundary at 15.00% Nieq:% Ni = 11.50 – ∆Nieq

= 11.50 – (–0.75) = 12.25

Option 4. Schaeffler provided, ingraphical form, a nitrogen correction fac-tor that is dependent upon the chromiumcontent (Ref. 13). This can be reduced toan equation for the actual nitrogen con-tribution (∆Nieq) to Schaeffler’s nickelequivalent:

∆Nieq = 30x(Actual % N–[0.085x%Cr/18 – 0.031%])

At 25.455% Cr, Schaeffler’s correctionpredicts nitrogen of 0.089% and ∆Nieq =–2.07. At 12.296% Cr, Schaeffler’s cor-rection predicts nitrogen of 0.027% and∆Nieq = –0.21. Then new values for % Niare calculated as

Fig. 4 — WRC FN vs. measured FN (as-deposited) for crack-free bends.

Fig. 5 — Schaeffler diagram, with compositions having measured FN less than 1.0, or havingmeasured FN less than or equal to the WRC-1992 FN + 1, plotted as solid ellipses. No correctionfor nitrogen is used in plotting these data. The lower limit of these points is taken to be the “marten-site boundary based on FN measurements.”

WELDING RESEARCH SUPPLEMENT | 183-s

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

Martensite boundary at 100% ferrite:% Ni = 1.036 – ∆Nieq = 1.036 –

(–2.07) = 3.106Martensite boundary at 15.00% Nieq:

% Ni = 11.50 – ∆Nieq = 11.50 –(–0.21) = 11.71

Step 5. With the % Cr and % Ni valuescalculated in step 4, and the chosenvalues for other elements given in step 1,calculate the WRC-1992 chromiumequivalents and nickel equivalents forthe four options listed above, as follows:

Martensite boundary at 100% ferrite:WRC Creq = % Cr = 25.455

Option 1:WRC Nieq =

% Ni + 35x(% C) + 20x(% N)= 1.036 + 3.5 + 0.4 = 4.936

Option 2:WRC Nieq = 2.236 + 3.5 + 0.4 = 6.136

Option 3:WRC Nieq = 1.786 + 3.5 + 0.4 = 5.686

Option 4:WRC Nieq = 3.106 + 3.5 + 0.4 = 7.006

Martensite boundary at 15.00% Schaef-fler Nieq:

WRC Creq = % Cr = 12.296

Option 1:WRC Nieq = 11.5 + 3.5 + 0.4 = 15.4

Option 2:WRC Nieq = 12.7 + 3.5 + 0.4 = 16.6

Option 3:WRC Nieq = 12.25 + 3.5 + 0.4 = 16.15

Option 4:WRC Nieq = 11.71 + 3.5 + 0.4 = 15.61

Step 6. Plot the four possibilities for atransposed martensite boundary on theWRC-1992 diagram. The chromium

equivalents and nickel equivalents foreach option (step 5) provide two pointsalong a transposed martensite boundary.The two points on the WRC-1992 dia-gram were connected by a straight line,and this line was extrapolated to theupper edge of the diagram to form amartensite boundary.

Figure 2 shows the resulting marten-site boundaries transposed from theSchaeffler diagram to the WRC-1992 di-agram. The boundaries are fairly close to-gether, and three of the four are parallel.The boundary generated from option 4(with Schaeffler’s variable correction fornitrogen) is not parallel to the other threeboundaries.

Experimental Procedure

With four possibilities for a martensiteboundary on the WRC-1992 diagram, itis necessary to test the candidates with ex-perimental data. As this required testingof many compositions, a method was de-vised to obtain numerous deposit com-positions from a limited number of fillermetals. This consisted of single-passbead-on-plate welds using the submergedarc process. By varying wire feed speed(current) and, to a certain extent, voltage,it was possible to obtain different dilutionlevels on mild steel base metal (ASTMA36). In general, increasing the wire feedspeed (current) causes increasing dilu-tion, and this is a major effect. In general,increasing the voltage also increases dilu-tion, but the effect is not as strong as thatof wire feed speed. Changing from DCelectrode positive (DCEP) polarity to DCelectrode negative (DCEN) polarity alsogenerally reduces dilution.

Two types of fluxes were used as an-other means of varying deposit composi-tion with a given wire — highly basicunalloyed flux and acid flux containing

considerable free metallic chromium. Ahighly basic unalloyed flux produces alower ratio of chromium to nickel, with aparticular electrode at a given dilutionlevel, than does a high-chromium flux.Most of the welds with basic flux weremade with 880 flux, but another basicflux, 882, was used for a few DCENwelds because this flux offers betterwelding characteristics with DCEN. Theacid high-chromium flux, A-100, weldsacceptably with both DCEP and DCEN.

The wires used were all 3⁄32 in.(2.4 mm) in diameter. Two were standardcommercially available solid wires —AWS A5.9 Classes ER308L and ER309L.Nine were laboratory-made, metal-coredwires designed to vary the ratio ofchromium to nickel well outside of thenormal range available in ER308L orER309L. Calculated compositions of themetal-cored wires and measured compo-sitions of the solid wires are given inTable 1. Also shown is the compositionof two six-layer weld deposits using Wire119 — one with the basic flux and onewith the acid high-chromium flux. Itshould be noted that the chromium con-tent of the six-layer deposit with the acidhigh-chromium flux and Wire 119 isabout 8% higher than the chromium con-tent of the corresponding deposit withthe basic flux. Table 1 includes a repre-sentative composition of the ASTM A36steel base metal used in this program.Not all of the base metal pieces camefrom the same heat, so there may be veryminor differences in composition forsome of the test pieces. Since the welddeposit was chemically analyzed, thisshould not cloud the results.

Evidence of martensite in the depositswas sought in three ways. Magnetic mea-surements of the deposits were com-pared with predictions of the Schaefflerand WRC-1992 diagrams. Since both

Table1 — Filler Wire Compositions and Ferrite and Representative Base Metal Composition

Measured WRC SchaefflerC Mn P S Si Cr Ni Mo Cu Nb N FN FN % Ferrite

Calculated Wire 119 0.07 1.54 0.01 0.01 0.61 19.21 14.84 0.00 0.00 0.00 0.00 N/A 0.0 0.0Wire 119, Six Layers with 880 Flux 0.06 1.16 0.02 0.01 0.73 19.88 13.40 0.00 0.00 0.00 0.02 0.1 0.0 0.0Wire 119, 6 Layers with A-100 Flux 0.07 1.71 0.03 0.01 0.74 27.83 12.29 0.03 0.00 0.01 0.04 27.1 31.8 19.1Measured ER309L 0.02 1.78 0.02 0.01 0.46 24.20 13.95 0.16 0.07 0.02 0.06 N/A 12.2 9.4Calculated Wire 120 0.07 1.54 0.01 0.01 0.61 19.21 15.64 0.00 0.00 0.00 0.00 N/A 0.0 0.0Calculated Wire 121 0.07 1.24 0.01 0.01 0.45 14.57 10.92 0.00 0.00 0.00 0.00 N/A 0.0 0.0Measured ER308L 0.02 1.91 0.02 0.01 0.51 19.54 9.97 0.05 0.02 N/D 0.04 N/A 9.2 8.4Calculated Wire 65N1693 0.14 2.15 0.03 0.01 0.38 20.12 19.77 0.30 0.20 0.00 0.03 N/A 0.0 0.0Calculated Wire 65N1694 0.14 2.12 0.02 0.01 0.38 19.69 26.31 0.30 0.20 0.00 0.03 N/A 0.0 0.0Calculated Wire 65N1695 0.14 2.12 0.03 0.01 0.38 20.12 24.82 0.30 0.20 0.00 0.03 N/A 0.0 0.0Calculated Wire 65N1716 0.13 1.92 0.02 0.01 0.34 17.61 17.86 0.28 0.20 0.00 0.03 N/A 0.0 0.0Calculated Wire 65N1717 0.14 1.87 0.02 0.01 0.34 17.61 27.78 0.28 0.20 0.00 0.03 N/A 0.0 0.0Calculated Wire 65N1743 0.04 2.08 0.03 0.01 0.42 27.05 6.66 0.28 0.20 0.00 0.03 N/A >100 80.6ASTM A36 Base Metal 0.16 0.51 0.01 0.02 0.03 0.08 0.06 0.02 0.06 0.00 N/D N/A N/A N/A

N/A = Not Applicable N/D = Not Determined

184-s | MAY 1999

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

martensite and ferrite are ferro-magnetic,it is not possible to be absolutely certainwhich phase (or which proportions ofeach phase) is responsible for a givenmagnetic response. Nevertheless, mag-netic measurements turn out to be veryuseful. Second, the longitudinal bend testof each weld deposit was examined forcracking. Cracks occur when the weldhas insufficient ductility to pass the bendtest, and this is considered to be due tomartensite. Third, selected deposits were

examined after bending for metallo-graphic evidence of martensite by etchedappearance.

The test coupon for martensite deter-mination was an ASME-type longitudinalface bend test. This is normally a bendtest in which the bend specimen thick-ness is twice the bending radius (a 2Tbend test) that requires 20% minimumelongation to pass. Martensite makes theweld brittle, resulting in specimen cracksduring bending, so this is a go/no-go test

that is very quickly executed. Since theA36 base metal was 1⁄2-in. (12.7-mm)thick, the bend specimen was bentaround a 1-in. (25.4-mm) radius bar. Theweld bead reinforcement extends abovethe surface of the base metal by perhaps3⁄32 in. (2.4 mm) to 1⁄8 in. (3.2 mm), so thatthe elongation requirement for the weldmetal is actually somewhat more than20%, if the weld is to pass the bend test.Thus, the results should be somewhatconservative. Figure 3 shows two suchbend test specimens, one passing the testand one failing the test (as evidenced bytransverse cracks after bending). Beforebending, the top surface of each weld de-posit was lightly ground on a belt sander,with the grinding marks parallel to thelength of the specimen, to smooth thesurface for ferrite measurement and toprevent surface ripples from influencingthe bend test. Cracks found after bendinginvariably extended completely acrossthe welds, so it could not be determinedwith certainty the point at which crack-ing initiated.

A second means of detecting the pres-ence of martensite in some samples wasto use a ferrite measuring instrument (Fis-cher Feritscope Model MP-3), calibratedaccording to AWS A4.2, before and afterbending. Martensite is ferromagnetic,just as is ferrite. If a magnetic response (a“Ferrite Number”) was found in a com-position above and to the left of the 0 FNline on the WRC-1992 diagram, it couldbe concluded to be due to martensite,not ferrite. Furthermore, weld depositsthat exhibited a magnetic response ap-preciably higher than the FN predictedby the WRC-1992 diagram could be in-ferred to contain martensite as well as fer-rite. Also, some weld deposits showed anincrease in magnetic response as a resultof bending, indicating that some marten-site was forming during bending. Themagnetic response of only a few weldswas examined after bending, but almostall of the welds were measured for “Fer-rite Number” before bending. Note that“Ferrite Number” or “FN” is used in quo-tation marks here because the responsecould be due to ferrite and/or martensite.

After the bend test of each sample wascompleted, the sample was reverse-bentflat again. Then chips were milled fromthe weld metal in the vicinity of theformer apex of the bend for analysis ofcarbon, sulfur and nitrogen by fusionmethods (given in ASTM E1019). Afterthe chips were removed, the remainderof the weld surface in the vicinity of theapex of the bend was finish-ground to adistance of about 1⁄32 in. (0.8 mm) abovethe original base metal surface, and opti-cal emission spectrophotometry (OES, as

Fig. 6 — WRC-1992 diagram, with compositions having measured FN less than 1.0, or havingmeasured FN less than or equal to the WRC-1992 FN + 1, plotted as solid ellipses. The lower leftlimit of these points is taken to be the “martensite boundary based on FN measurements.”

Fig. 7 — Schaeffler diagram, with compositions that passed the 2T bend test indicated as solid el-lipses and compositions that failed the 2T bend test indicated as open rectangles.

WELD

ING

RESEA

RC

H SU

PPLEMEN

T | 185-s

RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT

Table 2 — Experimental Weld Deposit Compositions, Bend Test Results and Welding Conditions

186-s| M

AY 1999

RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT

Table 2 — Experimental Weld Deposit Compositions, Bend Test Results and Welding Conditions (Continued)

described in ASTM E1086) was used toanalyze for Mn, P, Si, Cr, Ni, Mo, Nband Cu. Many of the chip sampleswere also analyzed for Ni and Cr bywet laboratory methods (given inASTM E353) to check on the accuracyof the OES results. In general, the twomethods of Cr and Ni analysis agreedwell. When both methods were used,the wet analysis result is reportedherein. With the chemical analysisdata, the chromium equivalent andnickel equivalent of each depositcould be plotted on the WRC-1992 di-agram, with a different symbol forwelds that passed the bend test vs.welds that cracked during bending.

After OES chemical analysis, se-lected OES samples were cross sec-tioned in the vicinity of the apex of thebend and were examined metallo-graphically for evidence of martensite.A variety of etches were tried to revealmartensite. Vilella’s etch, commonlyused for revealing martensite in 12%Cr stainless steels, was only somewhatsuccessful, as many of the welds wererather highly alloyed and would notetch. The etch that was found to bemost successful in revealing marten-site in all welds so examined was di-luted Kane’s etch. Kane’s etch consistsof a solution of 6 g CuCl2, 60 mL ofHCl and 6 mL of distilled water. Thiswas diluted with an equal volume ofwater to make it less aggressive. Thesample was immersed for 5 to 10 s.The etch darkens martensite, outlinesferrite and leaves austenite un-touched.

Experimental Results

In all, more than 100 deposits wereprepared and tested. For each deposit,the deposit composition, calculatedchromium and nickel equivalents, cal-culated and measured “ferrite” con-tent, bend test result and weldingconditions to produce the deposit areall listed in Table 2. Welds in Table 2that have the same prefix were madewith the same wire (see Table 1), withcomposition changes from one depositto the next due to varying dilution and,sometimes, to varying flux.

Magnetic Measurements

The magnetic measurements of ferritein Table 2 provide a means of assessingwhether martensite is present. First, itshould be recognized that welds contain-ing neither ferrite nor martensite can havea trace magnetic response. This can beseen especially well in the data for sam-

ples 65N1694-1 through -5. These weldsare high enough in nickel (15 to morethan 20%) and chromium (13 to 15%) tobe fully austenitic; yet they produce mag-netic response equivalent to 0.3–0.9 FN.It is a common experience with fullyaustenitic weld metals, such as Type 310(nominally 25% Cr, 20% Ni), to also findtrace magnetic response, so this did notcome as a surprise. This leads to the con-clusion that any weld deposit that mea-sured less than 1 “FN” in the as-deposited

condition must be at the upper edge of, orabove, the range of compositions in eitherthe Schaeffler or WRC-1992 diagram,where martensite is to be found. There-fore, a line along the lower left edge of thearea (where these compositions of lessthan 1 measured “FN”) should define partof the martensite boundary, in the area ofthe diagram where ferrite is not consid-ered to exist.

The remainder of the martensiteboundary in either diagram will pass

WELDING RESEARCH SUPPLEMENT | 187-s

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

Fig. 8 — WRC-1992 diagram, with compositions that passed the 2T bend test indicated as solidellipses and compositions that cracked indicated as open rectangles. The two heavy diagonal linesbound the region of compositions of mixed bending behavior — some bent and some broken.

Fig. 9 — WRC-1992 diagram showing locations of compositions examined for microstructure.

188-s | MAY 1999

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

through composition ranges where fer-rite is present. In this region of eitherdiagram, it can be considered that, if thepredicted FN from the WRC-1992 dia-gram agrees well with the measured FN,or if the predicted FN is appreciablyhigher than the measured FN, then themeasured magnetic response is probablydue only to ferrite, not to martensite. Thiswould allow the magnetic measurementdata to extend the martensite boundaryfrom the fully austenitic compositionrange into the composition range wheresome ferrite is also present. First, how-ever, it is appropriate to examine howwell the WRC-1992 diagram predictsmeasured ferrite for the experimentalweld deposits. This examination is doneonly for weld deposits that passed thebend test, so that little or no martensitecould be expected to be present to influ-ence the results. The WRC-1992 FerriteNumbers (calculated with the Ferrite Pre-dictor software) are plotted against themeasured “Ferrite Numbers,” for suc-cessful bends only, in Fig. 4. There aretwo types of data points plotted: Compo-sitions within the region of isoferritenumber lines in the WRC-1992 diagramare plotted as solid elliptical symbols;and compositions outside the region ofiso-ferrite number lines, whose FN thesoftware estimates by extrapolation, areplotted as open triangles. The perfect cor-relation line is also included. There isreasonable agreement with the perfectcorrelation line, with a few outliers.

Deposit compositions whose mea-sured FN is less than 1.0 and composi-tions whose measured FN is less than orequal to the WRC-1992 FN + 1 can then

be concluded to be essentially free ofmartensite. All of the compositions meet-ing one or the other of these two criteriaare plotted in Fig. 5. Note that, in calcu-lating the Schaeffler Nieq, nitrogen wasignored, as in option 1, when transpos-ing the Schaeffler martensite boundary tothe WRC-1992 diagram. If any other op-tion were used, the Schaeffler Nieq wouldbe reduced somewhat, moving all of thedeposit composition points downward.Figure 5 also includes a line that consti-tutes the lower left boundary of all ofthese compositions plotted in the dia-gram. This line is labeled the “martensiteboundary based on FN measurements”in Fig. 5. It does not agree very well withthe line labeled “upper martensiteboundary according to Schaeffler.” Itsslope is not the same and it is about 1Nieq below near the upper left end of theline and about 4 Nieq below near thelower right end. It should be noted thatany of the other three options for dealingwith nitrogen in the Schaeffler Nieqwould produce even poorer agreementbetween the “martensite boundary basedon FN measurements” and the “uppermartensite boundary according to Scha-effler.”

These same compositions can beplotted on the WRC-1992 diagram usingthe WRC Creq and Nieq. This is done inFig. 6, and, again, the “martensiteboundary based on FN measurements”is drawn as a line representing the lowerleft limit of compositions meeting one orthe other of the two criteria for conclud-ing that the composition is essentiallyfree of martensite. Figure 6 indicates thatthis “martensite boundary based on FN

measurements” does not agree well withany of the four options for transposingSchaeffler’s upper martensite boundary.At the upper left end, it is close to options1 and 4, but at the lower right end, it isnearly 4 Nieq, or more, below any of thefour options.

2T Bend Tests

The second method of examiningmartensite formation on the Schaefflerand WRC-1992 diagrams is to plot allcompositions that failed the 2T bend testwith a different symbol than is used toplot all compositions that passed the 2Tbend test. This is done on the Schaefflerdiagram in Fig. 7, again using no correc-tion for nitrogen. Two heavy lines aredrawn on the Schaeffler diagram. Theupper of these two lines is the boundaryabove and to the right of which all com-positions tested passed the 2T bend test.The lower of these two lines is the bound-ary below and to the left of which allcompositions tested failed the 2T bendtest. Between these two heavy lines inFig. 7, the 2T bend test produced mixedresults — some compositions passed andsome failed. It is noteworthy the “marten-site boundary based on FN measure-ments” lies between the two bend testlines. Agreement with Schaeffler’s uppermartensite boundary is poor.

Similarly, all compositions that passedthe 2T bend test are plotted on theWRC-1992 diagram (Fig. 8), with a dif-ferent symbol than that used to plot allcompositions tested that failed the 2Tbend test. Again, a heavy line is drawn inthe figure to indicate the lower left limitabove which all compositions passed.

Fig. 10 — Microstructure of weld 65N1716-4R after bending. DilutedKane’s Etch, X200. Magnetic response equal to that of 0 FN before bend-ing and 4.5 FN after bending indicates a small amount of martensite for-mation during bending. The dark etching regions contain somemartensite. No ferrite is present.

Fig. 11 — Microstructure of weld 65N1716-2R after bending. DilutedKane’s Etch, X200. Magnetic response equal to that of 0.05 FN beforebending and 37.1 FN after bending indicates that considerable marten-site formed during bending. The dark etching regions contain somemartensite. No ferrite is visible.

And a second heavy line is drawn to in-dicate the upper right limit below whichall compositions tested failed the 2Tbend test. Between these two lines is anarrow band of compositions, some ofwhich passed the 2T bend test and someof which failed. The width of this band ofcompositions having uncertain bendingbehavior is about 1.4 Creq in the hori-zontal direction, or about 1.4 Nieq in thevertical direction. Note also that the“martensite boundary based on FNmeasurements” lies between these twolines, just as it did on the Schaeffler dia-gram. Agreement with any of the trans-posed martensite boundaries from theSchaeffler diagram is poor.

Microstructures

The third way to observe martensite isto examine selected weld metal mi-crostructures. Not enough metallo-graphic examination was done to definea martensite boundary based solely onthis evidence, but the results serve to cor-roborate the magnetic and bending find-ings. Five compositions are chosen forthis examination. The first is a composi-tion (65N1716-4R) with calculated FN =0, which lies in the range where all com-positions passed the 2T bend test and areabove the “martensite boundary based onFN measurements.” The second is a com-position (65N1716-2R) with calculatedFN = 0, which lies in the range of com-positions where some passed (includingthis one) and some failed the 2T bend test.The third is a composition (120-11) withcalculated FN = 0, which lies in the rangewhere all compositions failed the 2T bend

test. These first three compositions pro-vide a transition across the martensiteboundary region at nearly constant Creq,without likelihood of finding any ferrite.The fourth composition (308-8), like thesecond, lies in the range of compositionsin which some (including this one) passedthe 2T bend test and some failed, but ex-trapolation of the iso-ferrite lines in theWRC-1992 diagram results in the expec-tation that this composition would con-tain some ferrite (4.6 FN, Table 2). Thefifth composition (65N1743-11) also liesin the range of compositions in whichsome passed the bend test and some (in-cluding this one) failed, and extrapolationof the iso-ferrite lines results in the ex-pectation that this composition wouldcontain considerable ferrite (39.7 FN).Figure 9 isolates these five compositionson the WRC-1992 diagram.

Figure 10 shows the microstructure ofWeld 65N1716-4R (WRC-1992 Creq =12.00, Nieq = 15.84), which passed thebend test. It can be seen in Fig. 9 that thiscombination of Creq and Nieq lies virtu-ally on the transposed Schaeffler marten-site lines using option 1 or 4, but belowthe transposed martensite lines usingoption 2 or 3. Before bending, this weldshowed no magnetic response (0.0 FN),indicating that it was free of martensite atthat time. But after bending (and straight-ening), its magnetic response was equalto 4.5 FN, which indicates that a smallamount of martensite formed duringbending. The photomicrograph of Fig. 10shows small patches of dark-etchingmartensite (which include some remnantaustenite) in an austenite matrix. Themartensitic areas seem to be concen-

trated in dendrite cores, which would beexpected to be leaner in alloying ele-ments than the interdendritic spaces, andtherefore could be expected to transformto martensite first. This observation of asmall amount of martensite after bend-ing, taken together with the magneticresponse in this sample equivalent to4.5 FN after bending, indicates clearlythat the magnetic method of detectingmartensite is quite sensitive.

Figure 11 shows the microstructure ofWeld 65N1716-2R (WRC-1992 Creq =12.07, Nieq = 14.41), which also passedthe bend test. With lower Nieq than theweld of Fig. 10, but virtually the sameCreq, this composition lies below all ofthe transposed Schaeffler martensiteboundaries in Fig. 9. Its composition lieswithin the band of compositions inwhich some passed the bend test andsome failed the bend test. This depositpassed the bend test. As deposited, itsmagnetic response was equal to that of0.05 FN, indicating that there was virtu-ally no martensite present. But afterbending, the magnetic response wasequal to that of 37.1 FN, so a great dealof martensite formed during bending.Figure 11 shows the extensive marten-site-containing regions, concentrated indendrite cores.

Figure 12 shows the microstructure ofWeld 120-11 (WRC-1992 Creq = 11.85,Nieq = 12.44), which cracked in the bendtest. This weld has virtually the same Creqas that of the two previously examinedwelds, but its Nieq is lower. It lies in theregion in Fig. 9 in which all of the sam-ples cracked during bending. Figure 12shows that this deposit is very heavily

WELDING RESEARCH SUPPLEMENT | 189-s

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

Fig. 12 — Microstructure of weld 120-11 after bending. Diluted Kane’sEtch, X500. Magnetic response was not determined before bending thesample, nor before discarding it. The dark etching regions contain somemartensite. The light etching regions are austenite. No ferrite is visible.

Fig. 13 — Microstructure of weld 308-8 after bending. Diluted Kane’sEtch, X1000. Magnetic response equal to that of 9.3 FN before bending,and 86.1 FN after bending, indicates that considerable martensite formedduring bending. Ferrite appears as thin elongated islands. Martensiteetches more darkly within the austenite matrix.

martensitic. Unfortunately, this samplewas discarded without evaluating mag-netic response. Taken together, Figs.10–12 provide a spectrum of microstruc-tures at about 12 Creq with decreasingNieq making a transition from no marten-site as deposited to very high martensite,all with no ferrite in the microstructure.

Weld deposits containing ferrite,along with austenite and martensite afterbending, were also produced, and it is in-teresting to examine a few of these. Fig-ure 13 shows the microstructure of Weld308-8 (WRC-1992 Creq = 16.60, Nieq =9.06). It lies in the band of compositionsin Fig. 9 in which some deposits bent andsome cracked during bending. As de-posited, its magnetic response was equalto that of 9.3 FN, which is slightly greaterthan the 4.6 FN the Ferrite Predictor soft-ware estimates by extrapolating the iso-ferrite lines. After bending, however, itsmagnetic response was equal to that of86.1 FN, indicating considerable trans-formation of austenite to martensite dur-ing bending. Nevertheless, it passed the2T bend test. The microstructure showsthin elongated islands of ferrite and ex-tensive martensite within the austenitematrix.

Figure 14 shows the microstructure ofWeld 65N1743-11 (WRC-1992 Creq =19.43, Nieq = 7.02). It also lies in theband of compositions in Fig. 9 in whichsome deposits bent and some cracked.As deposited, its magnetic response wasequal to that of 50.3 FN, which is a littlegreater than the 39.7 FN the Ferrite Pre-dictor software estimates by extrapola-

tion of the iso-fer-rite lines. It shouldbe noted that thehigher FN iso-fer-rite lines appear toconverge when ex-trapolated towardtheir lower left, soit is not possible tobe sure that all ofthe magnetic re-sponse beforebending was due toferrite — there mayhave been somemartensite presentas well. After bend-ing, the magneticresponse rose toequal that of 75.7FN, indicating thatsome transforma-tion of austenite tomartensite oc-curred duringbending. This de-posit cracked dur-

ing bending. Figure 14 showsmicrohardness impressions in the ferriteand in the martensite. The martensitereading was 430 Vickers, while that ofthe ferrite was 320 Vickers. No austeniteis clearly discernible in Fig. 14, but it isconsidered that the dark etching areasactually consist of a very fine mixtureof martensite and austenite — otherwise,the magnetic response would corre-spond to well over 100 FN.

Discussion of Results

It has been noted that the measure-ments of magnetic response lead to a“martensite boundary based on magneticmeasurements” that does not agree verywell with the upper martensite boundaryindicated in Schaeffler’s diagram — Fig.5. This comparison is based upon theassumption that Schaeffler’s uppermartensite boundary is specific to theas-deposited weld metal (i.e., beforebending). It is well known that plasticdeformation will induce martensite for-mation in compositions that are fullyaustenitic in the annealed condition,such as 304L stainless steel. Hull (Ref. 14)observed deformation-induced marten-site in a number of chill cast stainlesssteels while trying to determine marten-site-start temperatures. Deformation wasshown to raise the martensite-start tem-perature. So, as a corollary at ambienttemperatures, on the WRC-1992 dia-gram there would have to be a marten-site boundary before bending and asecond martensite boundary after bend-

ing, at richer compositions than the firstmartensite boundary (i.e., above and tothe right of the first martensite boundary).The present work concentrated on mag-netic measurements before bending,with only a few measurements afterbending, so it is not possible to proposeherein a martensite boundary afterbending.

One possibility for explaining the dif-ference between Schaeffler’s uppermartensite boundary and the “martensiteboundary based on magnetic mea-surements” developed herein is thatSchaeffler examined samples that hadundergone a small amount of plasticdeformation, causing the martensiteboundary to be shifted upward as com-pared to that for deformation-free weldmetal. A second possibility is that therewas a bias in Schaeffler’s chemical anal-yses and/or in the present chemical anal-yses. The present analyses were checkedby comparing classical wet results withspectrographic results, and NIST stan-dards were used for calibration of allmethods used herein. A third possibilityis a misinterpretation of when martensitefirst appears. It is not possible to deter-mine the reason for the difference and itis not proposed to modify the Schaefflerdiagram.

It is proposed to modify the WRC-1992 diagram to include a martensiteboundary. Figure 6 offers a “martensiteboundary based on magnetic measure-ments” that is appreciably different fromany of the four options considered fortransposing Schaeffler’s upper martensiteboundary to the WRC-1992 diagram.The upper left portion of that boundary israther clear since there is no ferrite to pro-duce possible confounding of the inter-pretation of magnetic responses. Thatportion of the boundary is based uponmeasured FN less than 1.0, but the lowerright portion of that boundary is less cer-tain. It is based upon the measured FNbeing less than or equal to the WRC-1992 calculated FN + 1.0. The problemhere is that calculated FN and measuredFN cannot be expected to agree exactly.Further, that portion is entirely outside ofthe region containing iso-ferrite lines inthe diagram, so it is necessary to use ex-trapolated iso-ferrite lines for calculationof the WRC-1992 FN, as was done withthe Ferrite Predictor software. These ex-trapolated iso-ferrite lines must be con-sidered as somewhat doubtful becausethey are beyond the limits of the databaseused in developing the earlier WRC-1988 diagram (Ref. 15), which is the di-rect parent of the WRC-1992 diagram.

There is corroborating evidence forthe “martensite boundary based on mag-

190-s | MAY 1999

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

Fig. 14 — Microstructure of weld 65N43-11 after bending. Diluted Kane’sEtch, X500. Magnetic response equal to that of 50.3 FN before bendingand 75.7 FN after bending indicates that some martensite formed duringbending. Ferrite appears light, while martensite in austenite appears dark.Vickers hardness impressions measured 430 in the martensite-containingregions, 320 in the ferrite.

netic measurements” in the 2T bend testresults. Figure 8 shows that this boundarylies in the region in which some bendtests passed and some failed. It is notquite parallel to the two boundaries ofthat region (the line above and to theright of which all bends passed, and theline below and to the left of which allbends failed), but it is nearly parallel. Sothe bend test results reflect the magneticresults very well.

It is evident from examination of Fig.8 that none of the four options for anupper martensite boundary, transposedfrom the Schaeffler diagram to the WRC-1992 diagram, agrees well with theboundary between experimental compo-sitions that all pass the longitudinal facebend test, and compositions that may ormay not pass this test. The transposedboundary from option 1 (which ignoresany correction for nitrogen) comes clos-est to this boundary, but it is overly con-servative. That is, if it were used as thebasis for accept/reject decisions (reject-ing deposit compositions below and tothe left of this line), the result would berejection of a range of compositions thatwill not contain martensite and that willpass a bend test.

The metallurgist wants to know whichcompositions will contain martensite.The engineer wants to know whether abend test can be passed because the abil-ity to pass a bend test is often required inwelding procedure qualifications. Bothconcerns are addressed by the narrowband of compositions included betweentwo lines in Fig. 8 — the line above andto the right of which all compositionspassed the 2T bend test, and the linebelow and to the left of which all com-positions failed this bend test. The“martensite boundary based on FN mea-surements” lies within this region. It mustbe appreciated that there is a degree ofuncertainty in any determination of thissort. The degree of uncertainty in deter-mining a martensite boundary in theWRC-1992 diagram is probably compa-rable to the width of the region betweenthese two lines. So this region, which in-cludes a degree of uncertainty, is pro-posed as the upper martensite boundaryin the WRC-1992 diagram. For clarity,the WRC-1992 diagram is redrawn inFig. 15 with only the proposed marten-site boundary added. The region to theleft of the upper portion of this boundaryis labeled “A + M” because both austen-ite and martensite would be expected insuch compositions. Lower and to the leftof this boundary, the region is labeled “A+ M + F” because deposits could be ex-pected to contain austenite, martensiteand ferrite, based upon extrapolation

(not shown) of the iso-ferrite lines.It is noteworthy that the magnetic de-

tection limit for martensite and the bendtest result boundary coincide ratherclosely. This seems to say that the pres-ence of martensite before bending isdetrimental to bend test performance. Butformation of martensite during bending isnot detrimental. If the weld deposit is free,or virtually free, of martensite beforebending, it will pass the bend test. Thatshould not be surprising — 304 stainlesssteel often forms some martensite duringbending, but bends successfully. If, on theother hand, the weld deposit contains ap-preciable martensite before bending, itwill probably fail the bend test. Note,however, that all of the data herein is forweld deposits of 0.05% C or higher. Thiswas a deliberate choice at the beginningof the program to ensure that martensiteformed would be relatively brittle. It iswell known that very low carbon marten-site (0.03% C or less) can pass the bendtest (e.g., Type 410NiMo weld metal). Forvery low carbon martensite, the predic-tion of martensite according to the exper-imental results herein should be correct,but the prediction of bend test failuresmay not be correct.

Future Work

Since the WRC-1992 diagram doesnot include manganese in the nickelequivalent, as the Schaeffler and DeLong

diagrams did, Fig. 15 has to be inter-preted as specific to the manganese levelexamined herein — about 1% Mn.Szumachowski, et al. (Ref. 16), estab-lished that there is no effect of man-ganese on ferrite vs. austenite formationat high temperatures in Cr-Ni stainlesssteel weld metals. However, Self, et al.(Ref. 17), examined the effect of Mn onaustenite stability as regards transforma-tion to martensite at low temperatures,and found the effect of Mn to vary withchromium content. In particular, theyfound Mn to be a more powerful austen-ite stabilizer at low chromium contentsthan at high chromium contents. This isbeing taken into account in further workto extend the examination of martensiteappearance to very high Mn weld metals,such as might be obtained from type 307or 209 welding filler metals.

Conclusions

From results of bend tests of more than100 compositions, a relatively thinboundary can be drawn on the WRC-1992 diagram shown in Fig. 15, aboveand to the right of which no martensitewould be expected in the as-depositedweld metal and the weld metal would beexpected to pass a 2T bend test. Belowand to the left of this boundary, marten-site can be expected in the weld metal,and it can be expected to fail a 2T bendtest. Within the boundary region, results

WELDING RESEARCH SUPPLEMENT | 191-s

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

Fig. 15 — WRC-1992 diagram with the martensite boundary for as-deposited weld metal. Com-positions below and to the left of the shaded boundary are expected to contain some martensiteand to fail an ASME 2T bend test. Compositions above and to the right of the boundary are ex-pected to be free of martensite and to pass an ASME 2T bend test. The behavior of compositionswithin the boundary is considered to be unpredictable.

192-s | MAY 1999

RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

/RE

SE

AR

CH

/DE

VE

LO

PM

EN

T/R

ES

EA

RC

H/D

EV

EL

OP

ME

NT

are expected to be unpredictable. Thisboundary is appreciably different fromany of four possibilities for transposingthe upper martensite boundary from theSchaeffler diagram. This boundary is spe-cific to compositions containing approx-imately 1% Mn.

Acknowledgments

The author is indebted to PeerReviewers who, on behalf of the AWSTechnical Papers Committee, reviewedthe original draft of this paper and madea number of insightful comments andsuggestions. Their efforts contributedmightily to improving the quality of thisfinal version of the paper. The author isalso grateful to The Lincoln Electric Co.for providing the opportunity to pursuethis investigation.

References

1. Kotecki, D. J., and Siewert, T. A. 1992.WRC-1992 constitution diagram for stainlesssteel weld metals: a modification of the WRC-1988 diagram. Welding Journal 71(5): 171-s to178-s.

2. Lefebvre, J. 1993. Guidance on specifi-cations of ferrite in stainless steel weld metal.Welding in the World 31(6): 390–407.

3. Kotecki, D. J. 1998. Standards and in-dustrial methods for ferrite measurement.Welding Journal 77(5): 49–52.

4. DeLong, W. T. 1974. Ferrite inaustenitic stainless steel weld metal. WeldingJournal 53(7): 273-s to 286-s.

5. Schaeffler, A. L. 1949. Constitution di-agram for stainless steel weld metal. MetalProgress 56(11): 680–680B.

6. Davis, J. R. 1993. Hardfacing, weldcladding, and dissimilar metal joining. MetalsHandbook, vol. 6, pp. 817–818.

7. ANSI/AWS A4.2-97, Standard Proce-dures for Calibrating Magnetic Instrumentsto Measure the Delta Ferrite Content ofAustenitic and Duplex Ferritic-AusteniticStainless Steel Weld Metal. American WeldingSociety, Miami, Fla.

8. Kotecki, D. J. 1997. Ferrite determina-tion in stainless steel welds — advances since1974. Welding Journal 76(1): 24-s to 37-s.

9. Kotecki, D. J. 1997. FN measurementround robin using shop and field instrumentsafter calibration by secondary standards —summary report. IIW Document II-1318-97.International Institute of Welding, Paris NordII, France.

10. Ferrite Predictor computer software,©1992, American Welding Institute,Knoxville, Tenn., now distributed by The Lin-

coln Electric Co. Cleveland, Ohio.11. ASME Boiler and Pressure Vessel

Code. 1995. Section III, Division I, FigureNB-2433.1-1. The American Society ofMechanical Engineers, New York, N.Y.

12. Espy, R. H. 1982. Weldability of nitro-gen-strengthened stainless steels. WeldingJournal 61(5): 149-s to 156-s.

13. Schaeffler, A. L. 1991. Privatecommunication.

14. Hull, F. C. 1973. Delta ferrite andmartensite formation in stainless steels. Weld-ing Journal 52(5): 193-s to 203-s.

15. Siewert, T. A., McCowan, C. N., andOlson, D. L. 1988. Ferrite Number predictionto 100 FN in stainless steel weld metal. Weld-ing Journal 67(12): 289-s to 298-s.

16. Szumachowski, E. R., and Kotecki,D. J. 1984. Effect of manganese on stainlesssteel weld metal ferrite. Welding Journal 63(5):156-s to 161-s.

17. Self, J. A., Matlock, D. K., and Olson,D. L. 1984. An evaluation of austenitic Fe-Mn-Ni weld metal for dissimilar metal welding.Welding Journal 63(9): 282-s to 288-s.