RESEARCH (fit SUPPLEMENT TO THE WELDING JOURNAL, JUNE 1986 Sponsored by the American Welding Society and the Welding Research Council

Readers are advised that all papers published in the Welding Journal's Research Supplement undergo Peer Review before publication for: 1) originality of the contribution; 2) technical value to the welding community; 3) prior publication of the material being reviewed; 4) proper credit to others working in the same area; and 5) justification of the conclusions based on the results. The names of the more than 160 individuals serving on the AWS Peer Review Panel are published periodically. All are experts in their respective technical areas, and all are volunteers in the program.

The Role of Inclusions in Controlling HSLA Steel Weld Microstructures

Nucleation of is accelerated when there is a combination of large austenite grains and a high density of intragranular inclusions

BY S. LIU AND D. L. OLSON

ABSTRACT. The effects of inclusions on size of the weld metal inclusions lar ferrite was found to be the constituent acicular ferrite formation in niobium decreased with increasing concentration responsible for the high toughness (Refs. microalloyed steel submerged arc weld of oxygen. Inclusions were found located 1-6). Acicular ferrite is randomly oriented metal were investigated using experimen­ at the grain boundaries or within the short ferrite needles which are formed tal chemical reagent grade fluxes from austenite grains. Depending on the size intragranularly. The interlocking nature of the CaF2-CaO-Si02 system. Five different distribution, the partition of the inclusions acicular ferrite, together with its fine grain wires were used. Direct rela­ to the boundary and to the interior of the size, provides the maximum resistance to tionships between (1) weld metal oxygen grains may be different. An optimum crack propagation by cleavage. For this and inclusion content and (2) inclusion inclusion size distribution to obtain a large reason, it has become increasingly impor­ density and weld metal microstructure volume fraction of acicular ferrite is the tant to understand the factors which were established in this investigation, one that contains a low percentage of would maximize the volume fraction of explaining why it is possible to relate fine size particles, with the inclusion pop­ acicular ferrite in the weld metal. microstructure to weld metal oxygen ulation sizes resembling a normal, or In addition to the hardenability agents, content, although the inclusion density close to normal, distribution. The combi­ such as carbon, manganese, chromium, correlation would be fundamentally nation of large austenite grains and high etc., weld metal oxygen also affects the more correct. Finer particles and larger intragranular inclusion density is the key microstructures and the mechanical size variations were seen in the high to obtaining a refined microstructure. The properties of steel weld metal. Oxygen oxygen weld metal. The mean particle strain energy contribution to the total can exist in a metal in the form of solid free energy of austenite decomposition solution element, or in a combined form due to the inclusion/matrix differential as inclusions. Due to the low oxygen thermal contraction effect was found to Paper presented at the 66th Annual AWS solubility in iron, only very low free Meeting, held April 29 to May 3, 1985, in Las be insignificant when compared with the oxygen content, i.e., in the uncombined Vegas, Nev. austenite to ferrite transformation vol­ state, can be expected. A major amount ume free energy term. of the oxygen is present as oxide or S. LIU is with the Department of Industrial oxy-sulfide inclusions resulting from reac­ Engineering, Pennsylvania State University, tions of oxygen with iron alloying ele­ University Park, Pa. D. L. OLSON is with the Introduction Center for Welding Research, Department of ments such as silicon, manganese and Metallurgical Engineering, Colorado School of Factors affecting HSLA steel weld metal aluminum in the molten weld pool. In the Mines, Golden, Colo. toughness have been studied, and acicu­ submerged process, when

WELDING RESEARCH SUPPLEMENT 1139-s CC

UJ cr 1 <1- CE UJ / Sfff OL Wrr y UJ 1 ////, 5

r- 6 LOG TIME LOG TIME LOG TIME CC- TYPICAL SUBMERGED ARC WELD COOLING CURVE I -GRAIN BOUNDARY FERRITE 2-SIDEPLATE FERRITE 3-ACICULAR FERRITE 4- PEARLITE 5-BAINI TE 6-MARTENSITE

Fig. 1—Schematic CCT diagrams showing the influence of oxygen on the weld metal phase transformations. A —High oxygen regime; B—Medium oxygen regime; C—Low oxygen regime given enough time, the inclusions will defined three regimes of transformation mation. grow in the melt and separate from the behavior with respect to the level of Ricks, Barritte and Howell (Ref. 16) bulk of the weld pool to the layer of oxygen. In the low weld metal oxygen studied the influence of second phase molten covering the weld puddle. content regime (<200 ppm), acicular fer­ particles on phase transformations in The separation of slags can take place as rite nucleation is restricted because of steels. They concluded that inert, non- a result of weld pool fluid flow or particle insufficient nucleants. As a result, Wid­ deformable inclusions are always less flotation (Refs. 7, 8). However, solidifica­ manstatten ferrite and bainite are the effective sites for nucleation than grain tion in submerged arc weld metal is predominant microstructures —Fig. 1A. boundaries. Nucleation on inclusions extremely fast and inclusions may Ferrite nucleation probably occurred on could only occur when site saturation of become trapped. The presence of these inclusions associated with the austenite the boundaries had taken place. second phase particles will affect both grain boundaries. In the intermediate Inclusion-assisted nucleation is a heter­ the mechanical properties and the phase weld metal oxygen content regime, intra­ ogeneous process in which interfacial transformation of the weld metal. granular acicular ferrite is the predomi­ energy at the nucleating interface is one In the early 1960's, investigators (Refs. nant microstructure. However, grain of the major controlling factors. Lattice 9-13) determined that mild steel weld boundary ferrite, ferrite sideplates and disregistry can be considered as one of metal contained nonmetallic inclusions of sometimes bainite are also present — Fig. the contributing factors. Disregistry can sizes ranging from a few tenths to several 1B. Nucleation probably occurred on be expressed by the Turnbull equation: microns. Most of the inclusions were inclusions lying at the prior austenite grain observed to be oxides and spherical in boundaries, resulting in grain boundary Aa0 8 = shape. An alignment of inclusions in rows ferrites. Acicular ferrite was suggested to (1) suggested that they were formed prior to have nucleated on intragranular inclu­ sions in a subsequent stage. In the high or during solidification (depending on the where Aa0 is the difference between the alloy composition). In the early 1970's, weld metal oxygen regime, inclusions lattice parameters of the substrate and Boniszewski (Ref. 14) did electron diffrac­ lying within the grain boundaries are the nucleated solid for a low index plane, considered to be ferrite nucleation sites. tion studies of carbon extraction replicas and a0 is the lattice parameter for the from mild steel weld metal and deter­ It was suggested that the grain boundary nucleated phase. Bramfitt (Ref. 17) modi­ mined that fine particles (0.005-0.060 ferrite nucleated and thickened in a direc­ fied Equation 1 and showed that the tion perpendicular to the grain boundary. micron) were cubic spinel MnO • Fe203 effectiveness of a compound as a nucle­ Widmanstatten sideplates, in the form of mixed with some Fe2Si04 cubic silicate. ating agent is related to the lattice dis­ parallel laths, grow across the grain However, it was not until the late registry between the nucleating agent boundary. Some coarse intragranular 1970's that the problem of weld metal and the nucleating phase. The greater the laths may also be observed — Fig. 1C. inclusions affecting ferrite nucleation was disregistry between the substrate and the considered in more detail, showing that Cochrane and Kirkwood (Ref. 3) also nucleating phase, the less effective the inclusion particles can affect phase trans­ suggested that oxygen in solution should compound is in promoting nucleation. formations by providing suitable nucle­ not be considered as the primary variant, Ito and Nakanishi (Ref. 18), Koukabi, et ation sites. Different mechanisms were and that the early nucleation of primary al. (Ref. 19), and Kanazawa, et al. (Ref. proposed. Abson, Dolby and Hart (Ref. 2) ferrite at the austenite grain boundaries 20), suggested that TiN was an effective and Abson (Ref. 15) showed that the prior to the growth of the sideplates is in nucleating agent for acicular ferrite during weld metal oxygen content affects main­ some way associated with the distribu­ the austenite decomposition. Heintze and ly the position of the nucleation curve for tion and nature of weld metal inclusions McPherson (Ref. 21) reported the same the acicular ferrite phase field. Schematic in the higher oxygen deposits. Like findings, that TiN acted as a weld metal CCT diagrams in Fig. 1 show the influence Abson, et al. (Refs. 2,15), they also raised grain refining agent and that the effec­ of weld metal oxygen on the austenite to the possibility of inclusion size distribution tiveness of titanium addition is enhanced ferrite phase transformation. The authors affecting the weld metal phase transfor­ in the presence of aluminum and boron.

140-s | JUNE 1986 However, Mori, ef al. (Ref. 22), showed types of inclusions being associated with much higher than that of the oxide inclu­ that an increase in nitrogen in the weld particular microstructures may not be sions (Ref. 32). During cooling, the aus­ metal did not bring any microstructural generally valid. Instead, it suggests that tenite matrix is strained at the presence of change except for an increase in proeu­ the effectiveness of inclusions on phase inclusions and accelerates its transforma­ tectoid ferrite. Instead, they demon­ transformation behavior may depend on tion into ferrite. The local stresses around strated that titanium containing oxides tramp elements (either in the form of the inclusions can be so high that the can nucleate acicular ferrite within the surface coating or other ways) rather matrix adjacent to the particles becomes austenite grains. Using the model devel­ than on the macroanalysis of the inclu­ plastically deformed. The lattice distor­ oped by Bramfitt (Ref. 17), the disregis- sions. tion in'the vicinity of the dislocations can tries between several oxides and nitrides It also became quite clear that acicular assist nucleation in various ways, namely, (the nucleating agents) and ferrite (the ferrite formation is a complex phenome­ by reducing the total strain energy of the nucleated phase) were determined. non with multiple factors and mecha­ embryo and by providing growth Among the inclusions that were identified nisms operating simultaneously. Several enhancement of the embryo through the to be associated with high volume frac­ other mechanisms suggested are summa­ dislocation pipe diffusion mechanism. tions of acicular ferrite, i.e., fe^C, TiO, rized below. Harrison and Farrar (Ref. 29) Considering the nature and magnitude of Ti203, FeO • Ti02, a-AI203, titanium correlated weld metal oxygen content these stresses, Brooksbank and Andrews monoxide has the smallest disregistry with austenite grain size. A reduction in (Ref. 32) determined that such stress with ferrite, approximating three percent. oxide inclusion content (by laser remelt­ could also be responsible for crack initia­ They proposed then that titanium mon­ ing of the weld metal) developed large tion. oxide existed on the surface of oxide austenite grain size, suggesting that the The main objective of this investigation particles with high potency of nucleating inclusions may have significant effect on is to examine in detail the effects of acicular ferrite. However, no substantial the grain boundary pinning. They con­ inclusions on HSLA steel weld metal experimental evidence could be supplied cluded that the austenite to ferrite trans­ phase transformation and to determine with respect to the surface titanium mon­ formation temperature change was due the influence of the inclusion size distribu­ oxide layer. Opposite to titanium monox­ to the interaction of inclusions with aus­ tion on the austenite to ferrite transfor­ ide, aluminum oxide and zirconium diox­ tenite grain boundaries. Higher oxygen mation. ide have the largest disregistries with welds, i.e., higher inclusion content, tend nucleated ferrite, approximately 30 per­ to reduce the austenite grain size and, cent. therefore, favor high temperature trans­ Experimental Procedure formation. Nucleation of acicular ferrite is Along with the lattice disregistry mod­ The base metal chosen was a niobium aided by inclusions not associated with el, the inclusions type was also consid­ microalloyed steel. Five commercial elec­ the prior austenite grain boundaries. ered to affect the weld metal ferrite trode wires were used to make the Complete removal of the inclusions formation. Keville and Cochrane (Ref. 23) welds. The compositions of the base would lead to the development of bain­ determined that inclusions rich in alumi­ metal and the filler wires are given in ite. Ferrante and Farrar (Ref. 30) also num oxide are typically related to welds Table 1. The CaF -CaO-Si0 system observed similar effects. They concluded 2 2 which were made with basic fluxes, while was selected because of its low oxygen that grains of diameters larger than 45 aluminum oxide inclusions with MnS are activity and its ability to produce low microns tended to produce acicular fer­ associated with acidic fluxes. They sug­ oxygen welds. Based on the preliminary rite. Cochrane, Ward and Keville (Ref. gested that the behavior can be related flux behavior studies (Ref. 33) and weld 27), however, reported that for welds to weld pool deoxidation (Ref. 23). Using metal microstructural and toughness anal­ made on plates with a low aluminum particle scanning electron microscopy, yses results (Ref. 6) from bead-on-plate level in spite of smaller austenite grain Pargeter (Ref. 24) determined the chemi­ and double V-grooved welds, four fluxes boundary surface areas, the austenite to cal compositions of the inclusions and were selected to perform a complete ferrite transformation temperatures were observed the following relationships. series of welding experiments producing observed to be lower. From this, they Grain boundary ferrite and ferrite side­ different levels of weld metal oxygen concluded that the grain boundary pin­ plates are usually associated with inclu­ contents —Fig. 2. For investigating the ning is not the sole effect, and that the sions with manganese and silicon, with or inclusions behavior only one set of data inclusion types present in the low alumi­ without sulfur. Acicular ferrite appeared (welds made with Flux No. 13) will be num weld are most effective in promot­ to be associated with aluminum-bearing discussed. ing ferrite nucleation at prior austenite particles. grain boundaries. Single V-grooved welds were made on Devillers, ef al. (Ref. 25), found that !/2 in. (13 mm) thick 15 X 4.5 in. acicular ferrite nucleated, in particular, at Ferrante and Farrar (Ref. 30) reported (380 X 110 mm) strips. In order to avoid aluminum manganese silicate inclusions that the mean inclusion diameter contamination of the weld joints from mill which were not covered by sulfur coat­ increased as the weld metal oxygen con­ scales, rust and edge preparation, the ing and which contained some titanium. tent increased. Cochrane, Ward and surfaces of the plates were also Aluminum oxide particles were found, Keville (Ref. 27) also reported similar machined. The joint geometry is shown however, not to favor intragranular fer­ results. Ahlblom, et al. (Ref. 31), con­ in Fig. 3. In addition to the welds men­ rite nucleation. Ricks, ef al. (Ref. 26), cluded in their recent publication that tioned above, a bead-on-plate weld detected the presence of manganese high fractions of acicular ferrite were made with commercial consumables sulfide on the surface of some of the often seen associated with small (aver­ (containing titanium, boron and other inclusions. They also observed a decrease age) diameter inclusions. They also found microalloying elements) using a heat input in the inclusion titanium content in the that the presence of rather coarse inclu­ of 2.9 kj/mm (74 kj/in.) was also included sulfide coated particles. Cochrane, Ward sions was generally associated with in this study. The chemical composition and Keville (Ref. 27) and Sagesse, ef al. polygonal ferrite nucleation. of this second base metal and the titani­ (Ref. 28), reported that large proportions Another aspect that is often men­ um-boron electrode wire can also be of acicular ferrite were obtained in the tioned is that inclusions can accelerate found in Table 1. Together with other welds which contained stoichiometric nucleation through the difference in the specifications, the welding variables used alumina or alumina rich inclusions. These thermal expansion behavior of the inclu­ are listed in Table 2. observations, sometimes contradictory, sions and matrix. The thermal expansion Quantitative metallography work was seem to indicate that the idea of specific coefficient of the iron matrix (austenite) is done on all the weldments. Transverse

WELDING RESEARCH SUPPLEMENT 1141-s Table 1—Chemical Compositions of the Base Metals and the Welding Consumables Used(a)

Mn Cu Cr \i Mo Al Zr V Nb Base Metal 1 0.140 1.250 0.220 0.008 0.008 0.210 0.140 0.170 0.030 0.019 - 0.029

E70S3 0.124 1.152 0.560 0.021 — 0.070 - 0.010 E70SC 0.066 1.690 0.610 0.016 0.028 0.020 - - - 0.010 EH 14 0.113 2.004 0.012 0.004 0.078 0.180 EM 5K 0.036 1.169 0.446 0.008 0.016 0.020 0.100 0.083 0.065 TiBOR 22 0.058 1.480 0.035 0.008 0.006 0.770 0.117 0.100 0.330 0.028 0.031 70 Base Metal 2 0.093 1.580 0.200 0.012 0.008 0.120 0.120 0.120 0.210 0.056

Si02 AI2O3 MgO CaO MnO Ti02 CaF2 Na20 Fe203 C OP121 TT 10.7 17.3 31.0 6.6 1.1 0.86 24.1 0.78 1.9 0.35

^'Concentration expressed in weight percent, except boron-ppm. tb)Typical analysis.

low and consequently not used for fur­ ther analysis. Carbon extraction replicas were made on the welds in order to examine the inclusions and other second phases. A Phillips 400 TEM was used in the determi­ nation of the sizes, number densities, volume fractions, and the geometrical shapes of the inclusions. The phases asso­ ciated with the inclusions were also examined. For the size distribution deter­ mination and the calculation of volume fraction, number density, and the mean particle size of the inclusions, the Ashby and Ebeling equations (Ref. 34) were Fig. 2 - Partial ternary diagram for Ca?2-CaO-Si02 system sho wing the nominal compositions of the used. 28 fluxes melted. (0 — Fluxes used in welding experiments) Chemical analyses of the welds were performed with a Baird Atomic Spectro- sections were cut and their surfaces pre­ were observed and measured. Figure 4 vac Model 1000 emission spectrometer. pared for metallographic examination. shows this section view with the delin­ Weld metal oxygen, nitrogen and carbon The metallographic study also involved eated equiaxed grain structure. Thermal contents were determined using Leco prior austenite grain size determination. grooving techniques were also used to interstitial analyzers. Due to sectioning orientation effect, the duplicate the grain size measurements. measurement on a transverse section High magnification light micrographs of Results and Discussion may not correspond to the true austenite polished and unetched weld samples To study the relationship between grain size. Therefore, sections parallel to were obtained for the inclusion analyses. weld metal composition and microstruc­ the fusion line (perpendicular to the The values determined by light microsco­ ture, carbon equivalent type equations columnar grains) were prepared and py for the inclusions volume fraction, approximately equiaxed austenite grains areal density, and linear density were too were used. Several different forms of CE equations (Refs. 35-39) can be found in

the literature, but only the Pcm equation

Mn + Cr + Cu Si Pcm = C + 4- — (2) 20 30

V Mo Ni + — + + — + 5B 10 15 60 where the elements are represented in weight percent, was included in this dis­ cussion because it covers the major ele­ ments in the weld metal. Most of the CE type equations do not take into consider­ ation the weld cooling rate, prior austen­ ite grain size, inclusion effects, etc., but literature data (Refs. 27, 40) showed a clear trend that with increasing carbon equivalent, there is an initial increase in the acicular ferrite content, followed by progressively increasing amounts of bain­ Fig. 3 — Joint geometry of the single V-groove welds ite and martensite.

142-s | JUNE 1986 Table 2—Operating Variables Used to Make the Weld Specimens for this Inclusion IOO Investigation -

E70S3 E70SG EH 14 EM 5K TiBOR 22 • eo Base Metal 1 ? Flux 13 S313H SC13H EH13H EM13H Ti I3H • u tutr. n Welding Current: 310 A Travel Speed: 4.4 mm/s Of UJ Welding Voltage: 35 V Heat Input: 2.5 kj/mm (62 kj/in.] u. • S40 TiBOR 22 • • • Base Metal 2 TiOeH u OP121 TT o "20 Welding Current: 530 A Travel Speed: 5.2 mm/s Welding Voltage: 30 V Heat Input: 3.0 kj/mm (76 kj/in.] ' ' 0.I6 0 IB 0.20 0.22 0.24 Pcm (%)

Fig. 5-Effect ofPcm on the weld metal acicular ferrite content

Fig. 4 —Schematic diagram and light micrograph showing the sample preparation and the microstructural aspects of prior austenite grain size determination

The same tendency, however, was not welds SG13H, Ti13H and TiOeH, which observed in this work. Figure 5 is a plot showed similar Pcm and prior austenite showing the correlation between volume grain sizes, the weld metal microstruc­ fraction of acicular ferrite and Pcm. Within tures were very different, varying from the range of Pcm values considered, a 40 to above 80 percent of acicular ferrite. large scatter in acicular ferrite was This indicates that some other factors (or observed, from 40 to above 80 percent. combination of factors) must also con­ In other words, Pcm alone is insufficient to tribute to the final microstructure. characterize the weld metal microstruc­ The grain boundary ferrite transforma­ ture, even when the cooling rate is main­ tion can be described as a two step tained constant, as in the weldments process: under investigation. Two representative 1) Nucleation along the austenite grain weld metal micrographs of low and high boundaries. volume fraction of acicular ferrite are 2) Growth in a direction perpendicular shown in Fig. 6. to the interface, into the austenite Fig. 6—Light micrographs illustrating the low For a similar chemical composition and grain. and high volume fractions of acicular ferrite cooling rate, larger austenite grain size If nucleation is assumed to be easy and microstructures. A—S313H weld; 8— TiOeH steels usually show higher hardenability site saturation occurs readily, the prob­ weld due to a smaller grain surface area to lem in question resembles that of surface volume ratio (lower surface nucleation nucleation where the grain surface will be site availability). Consequently, less grain covered by a layer of transformed mate­ boundary ferrite can be expected in the rial. Further assumptions made are that final microstructure. Volume fractions of allotriomorph growth is diffusion con­ primary ferrite (here defined as the sum trolled and that austenite grains are cylin­ of grain boundary ferrite and Widman­ drical in shape. Using the Avrami type statten sideplates) were plotted in Fig. 7 equation approach, the following expres­ as a function of measured prior austenite sion was obtained: grain sizes. With increasing prior austen­ ite grain size, the amount of primary X = 1 - exp (^r^) (3) ferrite was observed to decrease. In the case of acicular ferrite, the opposite where X is the volume fraction of grain PRIOR AUSTENITE GRAIN SIZE trend appeared to be true, although with boundary ferrite transformed, t is the Fig. 7 — Effects of prior austenite grain size on a larger scatter. However, examining the cooling time from 800° to 500°C (1472° the weld metal primary ferrite content

WELDING RESEARCH SUPPLEMENT 1143-s E 0.5 - • - 0 i 0.2 um a. • - j? > 0.08 um • • _^— I- 0.4 - —" • < **• o -~r~" 0.3 |-- 0 = f ( C03 ) •

•—ttmm • Fig. 8 — Correlation between experimentally 5 0.2 measured and calculated amounts of trans­ 0- formed grain boundary ferrite • < UJ 0.1 to 932°F), and d is the prior austenite grain diameter. As mentioned previously, HSi i i every effort has been made to minimize —i— • the sectioning orientation effect in the 250 300 350 400 austenite grain size determination such Z ( CO],cs] ) ( ppm ) that imprecision involved in the quantita­ Fig. 9—Variation of mean particle size of the inclusions with weld metal oxygen and sulfur tive metallography procedure will only content affect slightly the coefficient and the calculation of the equation. Experimentally measured volume frac­ important in weld metal phase transfor­ energy change of the inclusion formation, tions of grain boundary ferrite were plot­ mation. which is a function of the weld metal ted against the calculated values of X in As was reported in the introduction oxygen and sulfur content. Therefore, Fig. 8. Good correlation was obtained, (Refs. 27, 30, 31), the mean inclusion the higher the concentration of the inclu­ indicating that the grain boundary ferrite diameter was observed to increase with sion formers, the larger will be the num­ formation can be described by Equation an increasing weld metal oxygen. An ber of nuclei formed. Given the same 3. Nucleation originated on the austenite analysis of the transmission electron period of time of coalescence in the melt, grain boundaries, and the more the grain micrographs of the replicas in this work the weld metal with the larger number of boundary ferrite grows into the austenite showed that such observation could not nuclei will average smaller size particles. grain, the smaller will be the intragranular be verified. Higher weld metal oxygen The effects of number of nuclei in the region left for acicular ferrite. In the case samples indeed resulted in higher inclu­ melt on the rate of growth of oxide of high acicular ferrite content welds, this sion content than the lower oxygen inclusions were discussed thoroughly in also seems to suggest a much earlier welds. However, finer particles and larger the literature (Refs. 41-44) and seem to involvement of the intragranular inclu­ size variation were also seen in the higher agree with the present observation. sions in the weld metal phase transforma­ oxygen EM13H and S313H weld speci­ Inclusion volume fraction and number tion in blocking any further growth of the mens. This seems to agree with previous density were plotted as functions of weld allotriomorphs. observations reported by Dallam, et al. metal oxygen and sulfur content. Figure For a better understanding of the weid (Ref. 6). 10 shows the results of such correlation. metal phase transformation, one needs To verify the relationship between Inclusion volume fraction and number to analyze more closely all the factors mean particle diameter and the main density both increased with increasing that will affect the austenite to ferrite inclusion- elements in the weld weld metal oxygen and sulfur content. transformation. In addition to the weld metal (oxygen and sulfur), a linear regres­ The influence of inclusions on weld metal chemical hardenability, the cooling sion analysis was performed and the metal transformations can be explained in rate, and the prior austenite grain size, results are shown in Fig. 9. The straight two ways. First, they are concentrated in the presence of weld metal inclusions line passing through the full points the grain boundary region, pinning the and their effects on the austenite decom­ represents all the inclusions measured in austenite grain boundary and providing position should also be emphasized. this work, i.e., particles with diame­ nucleation sites. Too high inclusion densi­ Inclusions may affect both the thermody­ ters > 0.07 micron. It showed that the ty would lead to a more restricted aus­ namic and kinetic aspects of the weld mean particle size decreased with an tenite grain growth, resulting in a very metal transformations. First, the inclusions increasing concentration of oxygen and large surface area to volume ratio. This provide nucleation sites for the ferrite sulfur. This observation is different from increases the possibility of grain bound­ formation and alter the kinetic condition what was reported in the literature (Refs. ary ferrite occurrence. On the other of the process. Second, due to the differ­ 25, 28). Suspecting that the disagreement hand, a large austenite grain increases the ential thermal contraction between the might be due to the sampling and the hardenability of the weld metal by affect­ matrix and the inclusions, the austenite smallest size particles included in the ing the kinetic feasibility of grain bound­ matrix becomes strained during cooling, studies, inclusions smaller than 0.2 micron ary ferrite formation. To verify whether thus increasing the thermodynamic desire were excluded in a second calculation. inclusions were pinning the prior austen­ for the austenite to transform into fer­ This set of calculated values was plotted ite grain boundaries or not, the relation­ rite. in the same chart as full circles. This slope ship between the prior austenite grain Inclusion parameters such as mean par­ of the regression line is now positive, size and inclusion number density was ticle size and particle size mode (the indicating that particle size increased with represented graphically in Fig. 11. Very highest frequency class in the distribu­ increasing concentration of oxygen and good correlation was obtained, indicating tion), volume fraction, number density, sulfur. Nucleation rate is exponentially that weld metal oxygen and sulfur and size distribution are all fundamentally proportional to the inverse of the free expressed as number of inclusions did

144-slJUNE 1986 10.0 1 — 90 UJ • M • • (rt z 80 • cc

UJ 70 SE H 3 60 K O • • £ 50 60 8 0 10 0

INCLUSION NUMBER DENSITY IxlO'mm-') Fig. 11 — Effects of inclusion number density on the prior austenite grain size in the weld metal

250 300 350 400 450 00 I ( COD.CSD ) ( ppm) 80

10.0 60 • 40 • •

20 •

INCLUSION VOLUME FRACTION IXIO"1) Fig. 12—Variation of the percent of weld metal primary ferrite with the inclusion volume fraction

The inclusion size distribution effect can also be seen in the first group of weld metals. Both S313H and EM13H welds 250 300 350 400 showed very high occurrence frequency I ( COD , USD ) ( ppm ) of small size particles. Approximately 50 and 30 percent, respectively, of parti­ cles < 0.1 micron could be found. The Fig. 10-•Effects of weld metal oxygen and sulfur content on the inclusions volume fraction and number density mean particle sizes of the inclusions were also extremely fine, 0.14 and 0.16 micron, respectively. Those were also the welds that presented larger volumes of grain affect the prior austenite grain size. Sec­ the role of inclusions in weld metal phase boundary ferrite. This is reasonable ond, the inclusions that are within the transformation, the inclusion size distribu­ because smaller particles more effective­ austenite grains will provide nucleation tion and their effects have also to be ly pin the prior austenite grain bound­ sites for acicular ferrite formation, and an considered. The inclusion size distribu­ aries, resulting in smaller austenite grain adequate inclusion density is needed to tions of the different welds were deter­ sizes, as shown in Fig. 15. Consequently, obtain short, randomly oriented acicular mined. Basically, two types of distribu­ more grain boundary allotriomorph laths. A low inclusion density would lead tions were observed. They are the would be expected. This also shows that to a larger extent of ferrite lath growth, reverse J-shaped distribution and the pos­ the pinning of austenite grain boundary resulting in longer needles before they itively skewed, or right skewed, distribu­ by inclusions not only depends on the impinge on each other. Figure 12 illus­ tion. The first type was represented by volume fraction, but also varies with size trates the variation of the percentage of the welds S313H and EM13H-Fig. 13A. distribution. Zener's equation (Ref. 45) primary ferrite with the inclusion volume The second type included the SG13H, fraction. A general trend of primary fer­ EH13H, Ti13H and TiOeH welds-Fig. 2dcr rite increase was observed with the 13B. Among the welds from the second DL = Limiting Grain Size = (4) increasing inclusion density. group, inclusion populations in Ti13H and 3f Combining the results from Figs. 10- TiOeH approximated more closely the describes the effectiveness of inclusions 12, one can realize that the relationship normal distribution, showing larger parti­ pinning grain boundaries. Given the aus­ between weld metal oxygen and weld cle size modes (0.25 and 0.35 micron, tenite grain size and the inclusion volume metal microstructure is only an indirect respectively) and also larger mean parti­ fraction, f, the critical inclusion size, dcr one. It could be used to describe the cle sizes (0.21 and 0.24 micron, respec­ (above which inclusions will not restrain weld metal microstructure transition with tively). Different from some previous grain growth), can be calculated. The reasonable success solely because the observations (Ref. 27), i.e., in spite of the determined values of dcr can be found in weld metal inclusion population is directly larger inclusions sizes, these weld speci­ Table 3. To relate this to the measured proportional to the weld metal oxygen mens showed the finest microstructures, size distributions, the critical inclusion content, since it is the main inclusion predominantly acicular ferrite. Figure 14 sizes were used to determine the fraction former. shows the effect of inclusion particle size of inclusions associated with the grain on acicular ferrite content. For a more complete understanding of boundaries (F). Multiplying the total num-

WELDING RESEARCH SUPPLEMENT 1145-s 005 0 IO 0 15 0.20 0.25 0.30 0.33 PARTICLE SIZE MODE (aim)

Fig. 14 - Effects of inclusion particle size mode on the percent of acicular ferrite in the weld metal

MEAN PARTICLE SIZE Fig. 15- Variation of prior austenite grain size with the mean particle size of the inclusions

grow. In the case of EH13H weld, the number of intragranular inclusions seems to be low, with a calculated mean free spacing between inclusions of approxi­ mately five microns, allowing the ferrite 02 04 06 0 8 10 12 14 16 2 0 laths to grow. However, the formation of PARTICLE DIAMETER ( pm ) bainite is also due to the higher chemical Fig. 13 — Simple and cumulative size distribution of the inclusions extracted from EM 13H and TH3H hardenability found in the weldment. In weld metal the other direction are the specimens Ti13H and TiOeH, which showed slightly higher intragranular inclusion density than the EH13H weld. The calculated mean ber density (Nv) by the fraction (F), the within the grains. The larger number of inclusion density associated with the prior intragranular inclusions in S313H and free spacing between laths is only two austenite grain boundaries (%) is deter­ EM13H weld metals did not guarantee a microns. Because of that, more ferrite mined. Prior austenite grain size is plotted high acicular ferrite content. This is laths may nucleate and grow to impinge­ as a function of the inclusions found in because of the smaller prior austenite ment, resulting in a much finer micro- the boundaries in Fig. 16. The excellent grain size (held by the extremely large structure with higher acicular ferrite con­ correlation indicates that the inclusion number of small inclusions associated tent. The inclusion size distribution effect size distribution effect determines the with the grain boundaries), providing can also be seen in Fig. 17, where acicular number of inclusions in the boundaries, conditions for the grain boundary allotrio­ ferrite content is plotted as a function of indirectly controlling the austenite grain morphs to nucleate and grow. Once the intragranular inclusion density. The maxi­ mum acicular ferrite content corresponds size. The difference between Nv and nb is grain boundary ferrite is formed, it the intragranular inclusion density, n\, i.e., decreases the physical space in the grain to the welds with the optimum inclusion the potential sites for ferrite nucleation interior for acicular ferrite to form and size distributions.

Table 3—Summary of the Results of Calculation of the Critical Inclusion Sizes (dcr) and the Inclusion Densities Associated with Austenite Grain Boundaries and Interior of Grains

Tgs dcr Nv nb Nv - nb (Mm) f (Mm) (108 mm"3) F (108 mm"3) (108 mm"3)

S313H 55 0.033 0.27 7.60 0.82 6.23 1.37 SG13H 85 0.0040 0.51 2.01 0.85 1.71 0.30 EH13H 85 0.033 0.42 2.99 0.88 2.63 0.36 EM13H 55 0.0051 0.42 6.40 0.87 5.57 0.83 2 0 3 0 5 0 6 0 Ti13H 80 0.0033 0.40 3.27 0.86 2.81 0.46 TiOeH 88 0.0025 0.33 1.54 0.72 1.11 0.43 Fig. 76 — Effects of inclusions associated with the grain boundaries on the prior austenite grain size

146-S | JUNE 1986 IUU \ X ' < -;'•

g^ -:-••, ( ^ ,— i £ 80 — t

Ul t- m r i SE so

UJ <• u. """"i A;y • -K < i • 5 40 - • • . V _J • ./•»' V 3 ', 1 O

WELDING RESEARCH SUPPLEMENT 1147-s ferrite transformation. 8. Lau, T.W., and HoSang, A. 1984. Inclu­ HSLA steel weld metals, lournal of Materials 2. Grain boundary allotriomorph sion study, research report. Welding Institute Science 17:732. * growth is diffusion controlled and can be of Canada, RC200/2/84. 27. Cochrane, R.C, Ward, J.L., and Keville, described by an Avrami type equation as 9. Irvine, K.J., and Pickering, F.B. 1960. Rela­ B.R. 1983. The influence of deoxidation and/ tionship between microstructure and mechan­ or desulphurisation practice on the weld metal a function of prior austenite grain size ical properties of mild steej weld deposits. toughness of high dilution welds. Proc. Intl. and cooling rate. British Welding lournal 7:353. Conf. on The Effects of Residual, Impurity and 3. The inclusion size distribution deter­ 10. Wheatley, I.M., and Baker, R.G. 1962. Micro-alloying Elements on and mines whether the inclusions will be asso­ Mechanical properties of a mild steel weld Weld Properties, London, England, The Weld­ ciated with grain boundaries or located metal deposited by the metal-arc process. ing Institute, Paper 16. intragranularly, controlling the final weld British Welding lournal 9:378. 28. Saggesse, M.E., Bhatti, A.R., Hawkins, metal microstructure. 11. Gloor, K., Christensen, N., Maehle, G., D.N., and Whiteman, J.A. 1983. Factors influ­ 4. Weld metals with high acicular fer­ and Simonsen, T. 1963. Non-metallic inclusions encing inclusions chemistry and microstructure in weld metal. IIW DOC ll-A-106-63. in submerged arc welds.. Proc. Intl. Conf. on rite content are found to be associated 12. Tichelaar, G.W. 1964. Brittle fracture of The Effects of Residual, Impurity and Micro- with coarse austenite grains and with a mild steel weld metal at low temperatures. alloying Elements on Weldability and Weld large number of inclusions of diame­ Proc. Welding Symposium, Rotterdam, p. 31. Properties, London, England, The Welding ters > 0.2 micron. 13. Katoh, K. 1965. Investigation of non- Institute, Paper 15. 5. High oxygen content weld metals metallic inclusions in mild steel weld metals. 29. Harrison, P.L., and Farrar, R.A. 1981. with higher grain boundary ferrite con­ IIW DOC ll-A-158-65. Influence of oxygen-rich inclusions on the tent are found to be associated with fine 14. Boniszewski, T. 1972. Fine oxide parti­ austenite to ferrite phase transformation in austenite grains and with a large number cles in mild steel CO2 weld metal. Welding high-strength low-alloy (HSLA) steel weld met­ of smaller size inclusions of diame­ lournal 51(1):19-s to 22-s. als, lournal of Materials Science 16:2218. ters < 0.1 micron. 15. Abson, D.|. 1978. The Role of Inclusions 30. Ferrante, M., and Farrar, R.A. 1982. The in Controlling Weld Metal Microstructures in role of oxygen rich inclusions in determining 6. As weld metal oxygen content C-Mn Steels. The Welding Institute, Cam­ the microstructure of weld metal deposits. increases, the mean particle size of the bridge, England lournal of Materials Science 17:3293. inclusions decreases because of a higher 16. Ricks, R.A., Barritte, G.S., and Howell, 31. Ahlblom, B., Bergstrom, H., Hannerz, occurrence frequency of very fine parti­ PR. 1981. The influence of second phase N.E., and Werlefors, I. 1983. Influence of cles of diameters < 0.1 micron. particles on diffusional phase transformations welding parameters on nitrogen content and in steels. Proc. Intl. Conf. on Solid State Phase microstructure of submerged arc weld metal. Transformations, Pittsburgh, Pa., AIME, p. Proc. Intl. Conf. on The Effects of Residual, A ckno wledgments 463. Impurity and Micro-alloying Elements on Weld­ The authors acknowledge the research 17. Bramfitt, Bl. 1970. The effect of car­ ability and Weld Properties, London, England, The Welding Institute, Paper 38. and fellowship support of the United bide and nitride additions on the heteroge­ neous nucleation behavior of liquid iron. Met- 32. Brooksbank, D., and Andrews, K.W. States Army Research Office, the Consel- allurgical Transactions 1A (1):1987. 1972. Stress fields around inclusions and their ho Nacional de Pesquisa e Desenvolvi- 18. Ito, Y„ and Nakanishi, M. 1976. Study relation to mechanical properties, lournal of mento do Brasil, and the material and on Charpy impact properties of weld metals the Iron and Steel Institute 210:246. equipment support of Lukens Steel Com­ with submerged arc welding. The Sumitomo 33. Liu, S., Dallam, C.B., and Olson, D.L. pany and Hobart Brothers Company. Search 15:42. 1982. Performance of the CaF2-CaO-Si02 sys­ 19. Koukabi, A., North, T.H., and Bell, H.B. tem as a submerged arc welding flux for a 1978. Flux formation, sulphur, oxygen and niobium based HSLA steel. Proc. of ASM Intl. References rare-earth additions in submerged arc welding. Conf. on Welding Technology for Energy 1. Glover, A.G., McGrath, J.T., Tinkler, M.J., Proc. Intl. Conf. on Trends in Steels and Con­ Applications, Gatlinburg, Tenn., p. 445. and Weatherly, G.C. 1977. The influence of sumables for Welding, London, England, The 34. Ashby, M.F., and Ebeling, R. 1966. On cooling rate and composition on weld metal Welding Institute, p. 281. the deformation of the number, size, spacing, microstructure in C-Mn HSLA steel. Welding 20. Kanazawa, S., Nakashima, A., Okamo­ and volume fraction of spherical second phase /ournal 56(9):267-s to 273-s. to, K., and Kanaya, K. 1976. Improvement of particles from extraction replicas. Trans. AIME. 2. Abson, D.J., Doiby, RE., and Hart, P.H.M. weld fusion zone toughness by fine TiN. 236:1396. 1978. The role of non-metallic inclusions in Transactions ISII 16:486. 35. Granjon, H. 1967. Note on the carbon ferrite nucleation in carbon steel weld metals. 21. Heintze, G.W., and /McPherson, R. equivalent. IIW DOC IX-555-67. Proc. Intl. Conf. on Trends in Steels and Con­ 1983. Grain refinement of steels by titanium 36. Ito, Y„ and Bessyo, K. 1968. Weldability sumables for Welding, London, England, The innoculation during submerged arc welding. formula of high strength steel related to heat Welding Institute, p. 75. Australian Welding lournal 28:37. affected zone cracking. IIW DOC IX-576-68. 3. Cochrane, RC, and Kirkwood, P.R. 22. Mori, N., Homma, H., Okita, S„ and 37. Duren, CF. 1977. Significance of the 1978. The effect of oxygen on weld metal Asano, K. 1980. The behavior of B and N in implant test for assessment of the field weld­ microstructure. Proc. Intl. Conf. on Trends in notch toughness improvement of Ti-B bearing ability of large diameter pipes. Mannessmann Steels and Consumables for Welding, London, weld metals. IIW DOC IX-1158-80. Research Institute Report, Duisberg, PRG. England, The Welding Institute, p. 103. 23. Keville, B.R., and Cochrane, R.C 1982. 38. Tanaka, )., and Kitada, T. 1976. Implant 4. North, T.H., Bell, H.B., Koukabi, A., and Factors controlling the microstructure and test for studying cold cracking. IIW DOC Craig, I. 1979. Notch toughness of low oxygen toughness of submerged arc weldments. Proc. IX-959-76. content submerged arc deposits. Welding 30th Annual Convention on Welding Technol­ 39. Sekiguchi, H. 1976. Fundamental journal 58(12):343-s to 354-s. ogy 82, Hobart, The Australian Welding Insti­ research on the welding heat affected zone of 5. Dolby, RE. 1976. Factors Controlling tute, p. 263. steels. Nikkan Kogyo Shimbum, Tokyo, Weld Toughness — The Present Position. Part 24. Pargeter, R.|. 1981. Investigation of Sub­ Japan. II- Weld Metals. The Welding Institute, Cam­ merged Arc Weld Metal Inclusions. The Weld­ 40. Graville, B.A. 1980. Factors affecting the bridge, England. ing Institute, Cambridge, England. toughness of submerged arc weld metal - part 6. Dallam, C.B., Liu, S„ and Olson, D.L. 25. Devillers, L, Kaplan, D., Marandet, B., IL Technology Focus — Welding Institute of 1985. Flux composition dependence of micro- Ribes, A., and Riboud, P.V. 1983. The effect of Canada 2:1. structure and toughness of submerged arc low level concentrations of some elements on 41. Kiessling, R. 1978. Non-Metallic Inclu­ HSLA weldments. Welding lournal 64(5): 140-s the toughness of submerged arc welded C-Mn sions in Steel Parts l-IV, London, England, The to 151-s. welds. Proc. Intl. Conf. on The Effects of Metal Society. 7. Grong, O., Siewert, T.A., Martins, G.P., Residual, Impurity and Micro-alloying Elements 42. Twidwell, L.G. Physical Chemistry of and Olson, D.L. 1985. A model for the on Weldability and Weld Properties, London, Iron and Steelmaking. Butte, Mont., College of sequence of reactions occurring during silicon- England, The Welding Institute, p. 1. Mineral Science and Technology. manganese deoxidation of mild and low alloy 26. Ricks. R.A., Howell, P.R., and Barritte, 43. U.S. Steel Corporation. 1971. The Mak­ steel weld metals. Submitted for publication. G.S. 1982. The nature of acicular ferrite in ing, Shaping, and Treating of Steels. 9th Edi-

148-s|JUNE 1986 tion, Pittsburgh, Pa., p. 281. Dislocations, pp. 30-58, New York, N.Y., acicular ferrite nucleation in submerged arc 44. Turkdogan, E.T. 1972. Deoxidation of McGraw Hill. HSLA welds. Journal of Materials Science Let­ steel, lournal of The Iron and Steel Institute 47. Barritte, G.S., Ricks, R.A., and Howell, ters 2:123. 210:21. P.R. 1981. Application of STEM/EDS to study 49. Indacochea, I.E., and Olson, D.L. 1983. 45. Smith, CS. 1948. Grains, phases and of microstructural development in HSLA steel Relationship of weld metal microstructure and interfaces: an interpretation of microstruc­ weld metals. Proc. on Quantitative Microanal­ penetration to weld metal oxygen content. tures. Trans. AIME 175:15. ysis with High Spatial Resolution, London, lournal of Materials for Energy Systems 46. Hirth, J.P., and Lothe,). 1982. Theory of England, The Metals Society, p. 122. 5:139. 48. Kayali, E.S., Corbett, J.M., and Kerr, H.W. 1983. Observations on inclusions and

WRC Bulletin 309 November 1985

Development of a Production Test Procedure for Gaskets By A. Bazergui, L. Marchand and H. D. Raut

This report presents detailed results of tests carried out on five styles of gaskets with properties covering the broad range of gaskets commercially available. Three types of tests have been performed: mechanical (stress-deflection and creep); leakage tests on a hydraulic test rig; and leakage tests on a bolted-up rig. The report also presents selected Milestone Gasket Test Program results which pertain to four other styles of gaskets. The publication of this report was sponsored by the Task Group on Gasket Testing of the Subcommittee on Bolted Flanged Connections of the Welding Research Council. The price of WRC Bulletin 309 is $14.00 per copy, plus $5.00 for postage and handling. Orders should be sent with payment to the Welding Research Council, Ste. 1301, 345 E. 47th St., New York, NY 10017.

WRC Bulletin 310 December 1985

Damage Studies in Pressure Vessel Components By F. A. Leckie

The theory of damage mechanics is applied to pressure vessel components operating at high temperatures. It is suggested that some existing design procedures can be readily adapted to deal with high temperature problems. The publication of this report was sponsored by the Subcommittee on Elevated Temperature Design of the Pressure Vessel Research Committee of the Welding Research Council. The price of WRC Bulletin 310 is $14.00 per copy, plus $5.00 for postage and handling. Orders should be sent with payment to the Welding Research Council, Ste. 1301, 345 E. 47th St., New York, NY 10017.

WRC Bulletin 311 January 1986

Assessment of the Significance of Weld Discontinuities: Effects of Microstructure and Discontinuities upon Fracture Morphology By C. D. Lundin and C. R. Patriarca

The purpose of this report was to systematically investigate the metallurgical influence of weld metal microstructure, hydrogen presence and loading rate on fracture morphology in the presence of different types of discontinuities. In order to assess the metallurgical significance of weld discontinuities on fracture characteristics, mechanical and nondestructive testing, metallographic and fractographic examination, and hydrogen determinations were used to evaluate E7018, E9018 and El 1018 weld metals. Publication of this report was sponsored by the Subcommittee on Significance of Weld Discontinuities of the Pressure Vessel Research Committee of the Welding Research Council. The price of WRC Bulletin 311 is $14.00 per copy, plus $5.00 for postage and handling. Orders should be sent with payment to the Welding Research Council, Ste. 1301, 345 E. 47th St., New York, NY 10017.

WELDING RESEARCH SUPPLEMENT 1149-s