observation and analysis of threedimensional weld pool ... · pdf filepool and verified...

Introduction The weld pool is where complexwelding phenomena originate, playing afundamental role in determining result-ant welds. It has been the subject ofmany recent studies (Refs. 1–7). Amongweld pool behaviors, oscillation has re-ceived attention from welding re-searchers around the world (Refs.8–28), since Kotecki et al. (Ref. 8) firstobserved the phenomena of weld pooloscillation in 1972, and has been stud-ied with regard to different interests.Particular interests included its correla-

tion with the weld penetration (Refs.18–23, 29–31) and its effect on thegrain refinement and defect inhibitionusing methods like magnetic force stir-ring and ultrasound vibration (Refs. 32,33). Kotecki et al. (Ref. 8) studied com-plete joint penetration (CJP) weld poolbehavior in the case of stationary gastungsten arc welding (GTAW) usinghigh-speed motion pictures. They firstbrought attention to the research com-munity regarding the oscillation phe-nomena of the weld pool. They founda relationship exists between the natu-

ral frequency of the oscillation and di-ameter of the weld pool. A theoreticalmodel for CJP was proposed based ona stretched membrane theory. A simi-lar relation was also derived by Zacksenhouse et al. (Ref. 34), who developed an analytical model for sta-tionary complete-joint-penetrationpool and verified their mode by experiments. Richardson et al. (Ref. 29) studiedthe oscillation frequency for a station-ary GTA weld pool by an arc voltagesignal and arc light intensity signals,and found that the natural oscillationfrequency is strongly dependent onthe pool geometry and correlates wellwith the inverse of the square root ofthe pool mass. They found a distinctboundary exists for pool oscillationfrequency in the intermediate statebetween the partial and CJP condi-tions (Ref. 30). They also later foundthat arc light is a better transducer todetect pool oscillation than is arc volt-age due to the volume effect of theplasma region (Ref. 30). In 1993, Yooet al. (Ref. 30) proposed three kinds ofoscillation modes for CJP welding,which were, respectively, symmetric,sloshing, and mixed modes. However,these modes were proposed based onsimulation without experimental verification. Xiao et al. (Refs. 18–20) studied weldpool oscillation under stationary andlow-speed welding conditions, and pro-posed two pool oscillation modes: onefor partial joint penetration and anoth-er for complete joint penetration understationary welding conditions. Further-more, in welding with low traveling

WELDING RESEARCH

MAY 2017 / WELDING JOURNAL 143-s

SUPPLEMENT TO THE WELDING JOURNAL, MAY 2017Sponsored by the American Welding Society and the Welding Research Council

Observation and Analysis of ThreeDimensionalWeld Pool Oscillation Dynamic Behaviors

A sensor was used to observe weld pool oscillation in three dimensions, and then the 3D data was used to analyze the oscillation behaviors

BY K. ZHANG, Y. M. ZHANG, J. S. CHEN, AND S. J. WU

ABSTRACT Pool oscillation frequency is thought to have a direct relationship with weld joint penetration, and appropriate oscillations help refine grains and reduce defect sensitivity. Thispaper uses a novel laser dotmatrix sensor to observe the pool oscillation three dimensionally (3D) and use the 3D data as an enhanced ability to analyze oscillation behaviors. To thisend, pool oscillations were excited using pulse current, oscillation images were captured at1000 frames per second by a highspeed camera, and the acquired highspeed images wereprocessed and observed. The processed/observed images provide experimental data aboutoscillation amplitude and mode to study oscillation behaviors. It was found that three oscillation modes associated with partial, complete, and critical penetration exist during the basetime period. From the observed 3D dynamic evolution process, the periodical contractionand expansion at its natural frequency as excited by the pulsing current were observed. Theamplitude of the oscillation was found to gradually decrease as the oscillation process proceeded and the liquid metal solidified during the base current period. Furthermore, the amplitude was found to increase as the peak current increased. The experiments underconstant current and highfrequency pulsing current provided supplemental data to understand the pool behaviors. Because of the enhanced ability of 3D observation of the oscillation, this study enhanced the understanding on the pool oscillation and experimentally verified certain theoretical derivations about oscillation that have not been verified previously.

KEYWORDS • Pulsed GTAW (GTAWP) • Oscillation Mode • Oscillation Behavior • Dynamic Evolution Process

Zhang Paper 2016110 May 2017_Layout 1 4/7/17 4:01 PM 43

speeds, there even exists a third oscilla-tion mode: asymmetrical. They furtherdeveloped and improved their corre-sponding theoretical model based onclassical hydrodynamics to the liquidmetal or a stretched membrane theoryon the basis of Kotecki’s research (Ref.8). They found the natural oscillationfrequency of a partial joint penetrationpool is considerably higher than thenatural oscillation frequency of a CJPpool and that an abrupt transition oc-curs between the partial and completejoint penetration oscillation modes. Inaddition, they found the oscillation fre-quency drops after the weld poolchanges from partial to complete jointpenetration because of the change inthe oscillation mode. This result is fun-damental by providing an effectivemethod to distinguish the partial andcomplete joint penetration. Ramos et al. (Ref. 31) studied theweld pool oscillation in pulsed gasmetal arc welding (GMAW-P) bymeans of Shadowgraphy image pro-cessing techniques, and obtained thefrequency spectra by fast Fouriertransform. They found the method ofShadowgraphy image processing formeasuring the weld pool oscillation ismore reliable than that from using thearc voltage signal. Furthermore, research has also re-vealed that, like arc vibration andweaving, ultrasonic vibration, andweld pool stirring by electromagneticforce, the pool oscillation duringpulsed arc welding can also promotegrain refinement of the weld and re-duce weld defects (Refs. 32, 33,35–38). Among the reasons support-ing this theory is that pulsed arc weld-ing can excite the pool oscillation by

varying the energy input into the weldpool. This favors generating finergrains and more effective substruc-tures, thus better ensuring productionof quality welds. Along this direction, preliminaryinvestigations by Nakata et al. alsoshowed the effect of current pulsationon the weld solidification structure ofaluminum alloys (Ref. 32). They alsoexplored the effect of low-frequencyGMAW-P on grain refinement of theweld metal and improvement of solidi-fication crack susceptibility in alu-minum alloys (Ref. 33). They foundthat weld pool oscillation strongly af-fects grain refinement and, to someextent, also found that grain refine-ment has a beneficial effect on the so-lidification crack susceptibility of theweld metal. In previous studies, the pool oscilla-tion was studied based on analyzingthe frequency of the oscillation as aone-dimensional signal; for example,the weld penetration was experimen-tally correlated to the frequency fromsuch one-dimensional, signal-basedfrequency. While the one-dimension-based analysis on the weld pool hasthe advantage of simplicity, the weldpool is actually a distributive bodywhose oscillation is in all directions. Itis apparent that one-dimension meth-ods do not fully characterize the pooloscillation such that many phenomenaand characteristics are still not fullyrevealed. The pool oscillation shouldbe studied fully dimensionally, but themajor challenge lies in the lack of aneffective method to measure thethree-dimensional surface and thecomplexity of the analysis of thethree-dimensional oscillation of the

flexible distributive body. Along this direction, Shi et al. (Refs.21–23) made the first attempt to ana-lyze the pool oscillation based on athree-dimensional laser multiline sens-ing method to investigate weld pool os-cillation in pulsed GTAW. However, theyfocused on the correlation of the pooloscillation characteristic frequency withthe weld penetration. In this paper, anovel sensing system is used to observethe three-dimensional weld pool surfaceto analyze the oscillation modes three-dimensionally and their correlationwith the dynamic evolution process ofthe weld pool. In particular, in previous work(Refs. 39–43), a new laser dot-matrixsensing method was proposed. Thethree-dimensional surface of the weldpool can be monitored and measured.To this end, a 19 × 19 dot-matrix pat-tern was projected onto the weld poolsurface. Because of the specular natureof the weld pool surface, a laser dotmatrix reflected from the weld poolsurface can be intercepted/imagedby/on a diffusive imaging plane placeda distance from the center of the weldpool. The images are captured by ahigh-speed camera, and the three-dimensional shape of the weld poolsurface can be directly observed aswell as monitored and reconstructedfrom a two-dimensional to three-di-mensional freedom surface. Becausethe reflected position on the imagingplane differs with the curvature of aspecific position of the weld pool sur-face, variations of the reflected laserdot matrix can characterize the pooloscillation direction as well as the am-plitude of convexity and concavity ofthe weld pool surface. This paper focuses on the analysis ofthe dynamic behavior of the pool oscil-lation using the unique laser dot-matrixsensing method introduced previously.From the reflected laser dot matrix im-age captured by high-speed camera, theoscillation modes were clearly revealedand the dynamic real-time evolutionprocess for pool oscillation under differ-ent penetration states were explored ac-cording to the shape variation of surfaceoscillation of the weld pool under vari-ous welding conditions of pulse currentand penetration, such as complete, par-tial, and critical penetration, which isdefined as the states of penetration be-tween complete and partial penetration.

WELDING RESEARCH

WELDING JOURNAL / MAY 2017, VOL. 96144-s

Fig. 1 — Sensing system for pool oscillation experiment.

Zhang Paper 2016110 May 2017_Layout 1 4/7/17 4:01 PM Page 144

Furthermore, the variations of oscilla-tion amplitude related to the pulse cur-rent were analyzed to evaluate the pooloscillation process.

Methods and Principles

Sensing System

The proposed weld pool surfacesensing system is shown in Fig. 1 (Ref.43), where a 20-mW illumination laserwith an interbeam angle of 0.77 deg anda wavelength of 685 nm with variablefocus was used to generate a 19 × 19dot-matrix structured light pattern. Thelaser pattern was projected onto thearea under the electrode at 30 deg witha 50-mm distance from the electrodeand covering the entire possible weldpool surface. To intercept the reflectedlaser pattern, an imaging plane wasplaced at a known distance of approxi-mately 50 mm from the electrode,which could be as simple as a piece ofglass attached with a grid paper. A high-speed camera was used to record the re-flected images on the imaging plane. Tominimize the influence of the arc, thecamera was fitted with a band pass filterof 20-nm bandwidth centered at a 685-nm wavelength of the laser used. Asshown in Fig. 1, a universal coordinatesystem served to locate the positions forall the objects in the sensing system. Inthis system, the torch was on the z-axis,and the workpiece surface was on the x–y plane.

Characteristics for PoolOscillation

During the peak current period, thecenter of the weld pool surface was de-pressed by the arc jet pressure. Duringthe base current period, the balanceamong the arc jet pressure, surfacetension, and gravity, which were exert-ed on the weld pool, was broken afterthe arc plasma pressure was suddenlyreduced. As such, the surface tensionpulled the pool back toward a newequilibrium position, inducing an os-cillation to the pool. On the otherhand, the molten specular weld poolsurface could reflect, like a mirror,most of the incidental laser light. As aresult, the reflected laser dot matrixcould respond simultaneously to themotion of the weld pool surface to re-

flect its oscillation. When the laser dot matrix project-ed onto the weld pool surface, thelaser dots were reflected onto the im-aging plane. The reflected dots on theimaging plane had a different mappingrelationship corresponding to the dif-ferent shapes of the weld pool surface(Refs. 43, 44): convex, concave, andcombination — Fig. 2. Figure 2(1a–2a)are, respectively, the scheme of themapping relationship correspondingto Fig. 2(1b–4b), which are the reflect-ed laser dot matrix images of the dif-ferent weld pool surfaces. The figuresshow the adjacent laser dot distance atthe center of the weld pool changeswith the convex or concave state of theweld pool surface. According to themapping relationship between theprojected laser dots and reflected laserdots, the weld pool surface of Fig.2(1b) is moderately convex; the weldpool surface of Fig. 2(2b) is more con-vex because the distance of the reflect-ed laser dots in the center of the weldpool is larger than that of Fig. 2(1b);the weld pool surface of Fig. 2(3b) ismoderately concave because the dis-tance of the reflected laser dots in thecenter of the weld pool is close; and inFig. 2(4b), the distance of the reflectedlaser dots in the center of the weldpool is closer, showing that the weldpool surface is deeper than that of Fig.2(3b). If the weld pool surface weremuch deeper downward, the reflectedlaser dots on the imaging plane wouldprobably converge to a point or evencross up and down. Therefore, the am-plitude of the change in the pixel dis-tance between two adjacent laser dots

in the center of the weld pool can rep-resent the amplitude of the weld pooloscillation.

Experimental Procedureand Method

The experiment aimed to studypool oscillation behavior in differentweld penetrations for GTAW-P understationary conditions, which meantzero travel speed. Typical 304 stainlesssteel plates, 4.7 mm thick, were usedas the workpiece, which was placed ona thick copper plate with a backfill ofargon. Welding was carried out using a2% thoriated tungsten electrode witha diameter of 2.4 mm and a tip angleof 60 deg at direct current electrodenegative with 99.995% argon shield-ing gas at a flow rate of 15 L/min anda distance of 5 mm from the tungstenelectrode to the specimen. The welding parameter design isgiven in Table 1. The image-capturerate was set to 1000 frames per second(f/s). The base time was set to 20 msto observe the weld pool oscillationprocess at its natural frequency. Thebase time was set to 10 ms, mainly forinspecting the pool oscillation behav-ior of complete joint penetration,whereas the 3-ms base time served instudying the forced pool oscillation be-havior when the pulse frequency wasgreater than the natural frequency ofthe weld pool because of the veryshort base current time. Experiment 1was designed to judge the pool oscilla-tion under constant welding currentconditions. Experiments 2 and 5 were

WELDING RESEARCH


Fig. 2 — Characteristic principles for pool oscillation.

1A 2A 3A 4A

1B 2B 3B 4B


designed for exploring oscillation be-havior and partial penetration. Experi-ment 3 was designed for exploringforced oscillation. Experiments 4 and6 were designed for complete jointpenetration and oscillation behavior.Experiments 5 and 7 were designedfor critical penetration and oscillationbehavior. The other experiments weredesigned for analyzing pool oscillationbehavior or amplitude issues.

Experimental Results andDiscussion

Pool Oscillation Mode

The oscillating weld pool was im-aged/captured at 1000 f/s by the high-speed camera during welding. Both theconvexity and concavity of the weldpool surface under the pulse currentwere observed. The variation in the re-flected laser dot matrix presented the

dynamic variation in the weld poolsurface such that the pool oscillationbehavior could be clearly observed. Re-sults of the experiment showed threepool oscillation modes at differentdepths of penetration.

Symmetrical Oscillation under PartialPenetration

Figure 3(1b) and (2b) are the con-secutive reflected laser dot-matrix im-ages of the oscillating weld pool at par-tial penetration, such as those in Ex-periments 2 and 5 in Table 1. The pixeldistance of adjacent laser dots in theweld pool center in Fig. 3(1b) is rela-tively large, while in Fig. 3(2b) it ismuch smaller. This concurs with thecharacteristic principles discussed inthe section titled “characteristics forpool oscillation.” The center of theweld pool shrank toward the inside[Fig. 3(1b)] and, correspondingly, theshape of the weld pool oscillation sur-

face was concave, whereas the centerof the weld pool expanded toward theoutside [Fig. 3(2b)], correspondingly,and the shape of the weld pool oscilla-tion surface was convex. Figure 3(1b and 2b) shows that theweld pool oscillation in the partial pene-tration is symmetrical about the centerof the weld pool. As was presented inRefs. 18 and 20, the oscillation modelcan be described with the first harmonicmode of the Bessel function. The arc pressure is known to becaused by a magnetic pressure differen-tial along the length of the arc, whichaccelerates the arc plasma and entrainsthe gas toward the workpiece to form adynamic jet pressure (Refs. 8, 29). It in-creases with the square of the currentand decreases from electrode to work-piece as the arc radius increases. Thus,the arc jet pressure derived from thepeak current is much greater than thatderived from the base current. When the welding current is

WELDING RESEARCH


Fig. 3 — Symmetrical oscillation at the center of the pool underpartial penetration.

Fig. 4 — Symmetrical oscillation at the center of the pool undercomplete joint penetration.

1A 1A

1B 1B

2A 2A3A 3A

3B3B2B 2B

Table 1 — Welding Parameters

No Peak Current Base Current Peak Time Base Time Pulse Frequency Captured Rate Target of Study Ip/A Ib/A Tp/ms Tb/ms f/Hz f/s

1 60 60 — — — 1000 DC oscillation 2 60 20 20 20 25.0 1000 Partial / behavior 3 80 20 20 3 43.5 1000 Forced oscillation 4 80 20 20 5 40.0 1000 Complete / behavior 5 80 20 20 20 25.0 1000 Mode / behavior 6 100 20 20 10 33.33 1000 Complete / behavior 7 100 20 20 20 25.0 1000 Critical / behavior 8 120 20 20 5 40.0 1000 Oscillation behavior 9 140 20 20 5 40.0 1000 Oscillation behavior 10 160 20 20 5 40.0 1000 Oscillation behavior

* Partial – partial joint penetration; complete — complete joint penetration; critical – critical penetration; mode – partial and critical penetration; behavior – oscillation behavior.


switched to the base current from thepeak current, the arc jet pressure sud-denly releases from the top of the pooland induces pool oscillation at a natu-ral frequency. The surface tensionpulls the pool back toward its equilib-rium position. The pixel distance atthe central region of the weld pool en-larges [Fig. 3(2b)], and the weld poolcontinues to expand upward after itreaches its point of equilibrium, due tothe inertia. When the weld pool ex-pands upward to its highest position,the surface tension and gravity dragthe pool back toward its point of equi-librium, and the pixel distance at thecentral region of the weld pool shrinks— Fig. 3(1b). The weld pool continuesto depress downward after it reachesits point of equilibrium position, againdue to the inertia. More details on theevolution process can be seen in thesection titled pool oscillation underpartial penetration. Due to the support of the solidmetal in the bottom of the pool, theliquid metal around the pool edge waspushed upward when the liquid metalat the central pool was depresseddownward. The scheme of the pool os-cillation mode for partial penetrationis illustrated in Fig. 3(1a, 2a). The os-cillation mode was consistent with themode 1 proposed by Xiao et al. (Ref.18). Figure 3(3a and 3b) are the topand bottom for the weld associatedwith partial penetration.

Symmetrical Oscillation underComplete Joint Penetration

Images of the oscillating weld poolin Fig. 4(1b and 2b) show completejoint penetration occurring as it did inExperiments 4 and 6 (Table 1). Figure4(3a and 3b) are the top and bottom ofthe weld, respectively, correspondingto CJP oscillation. To a first approxi-mation, the oscillation under CJP canbe described in terms of a stretchedmembrane or classical hydrodynamicsto the liquid metal in the weld pool(Refs. 8, 18, 20). In Fig. 4(1b), a bright region appearsin the central region of the molten pool.The brightness is much higher therethan in other regions of the moltenpool. Under the CJP condition, the bot-tom metal of the workpiece has alsobeen melted, as shown in Fig. 4(3b).The bottom of the weld pool loses thesupport of the solid metal, and, instead,is maintained by the surface tension ofthe liquid metal from the bottom pool.Therefore, the liquid weld metal in theCJP pool has an extra degree of freedom(normal to the surface of the pool) (Ref.20), and the surface tension along thetop and bottom surface of the pool asthe main driving force has a significantinfluence on the oscillation behavior.When the arc jet pressure acts on thetop of the weld pool surface, the weldpool surface depresses to such a low po-

sition that the reflected laser dot-matrix shrinks into a large bright pointin the center of the weld pool. At thistime, the shape of the weld pool surfaceis concave (see the section titled charac-teristics for pool oscillation). Thescheme of the pool oscillation mode isshown in Fig. 4(1a). Figure 4(2b) shows a brighter circlein the edge of the pool. When the arcjet pressure on the top of the weldpool surface is suddenly released, thesurface tension on the weld pool topand bottom pulls the pool back towardits equilibrium position. The center ofthe weld pool first expands upward,and then the edge of the weld pool fol-lows with the motion of the weld poolcenter. However, because of the lack ofsupport of solid metal in the bottomof the pool, the weld pool surface ispushed down to a relatively low posi-tion by the high arc pressure that de-rived from the pulse current. There-fore, the edge of the weld pool doesnot easily drag back to the convexwhen the pool’s center reaches thehighest point, and most of the time,the reflected image of the weld poolsurface, like in Fig. 4(2b), is a brightcircle around the pool’s edge. The sec-tion titled Characteristics for Pool Os-cillation shows that the shape of theweld pool’s central region is convexand that of the weld pool edge is con-cave, as shown in Fig. 4(2b). Thescheme of the oscillation mode is illus-trated in Fig. 4(2a). It is apparent from the reflectedlaser dot characteristic that the pooloscillation under complete joint pene-tration is also radial symmetrical withrespect to the arc axis. However, boththe top and bottom of the weld poolare liquid film, and the pool oscillationmorphology is significantly different.The corresponding image of the oscil-lating weld pool surface [Fig. 4(1a)] isalso considerably different from theimage under partial penetration —Fig. 3(1a). Figure 4(1b and 2b) showsthat the pool oscillation amplitude ismuch greater than that under partialpenetration. The results coincide withmode 2 proposed by Xiao et al. (Refs.18–20), and the experimental phe-nomena also verify the hypothesisproposed by Richardson et al. (Ref.30), i.e., there exists a kind of symmet-rical oscillation mode under completejoint penetration as described earlier.

WELDING RESEARCH


Fig. 5 — Sloshing oscillation under critical penetration.

1A

1B

2A 3A

3B2B


Sloshing Oscillation under CriticalPenetration

Figure 5(1b and 2b) shows the pooloscillation under the critical penetra-tion, such as in Experiments 5 and 7,(Table 1). In Fig. 5(1b and 2b), the sym-bols and represent the weld pooloscillation valley and peak, respectively. The pixel distance minimum in Fig.5(1b) is located on the pool oscillationimage’s right side, whereas the pool os-cillation image’s minimum pixel dis-tance in Fig. 5(2b) is located on the leftside, neither at the center of the weldpool. Unlike the pool oscillation undercomplete or partial penetration, the ex-treme position (oscillation center here-after) of the oscillating pool under criti-cal penetration is not always at the weldpool center but instead shifts like awave of water in a pool. It is possiblethat the oscillation center would shift toanother position in the next moment.The oscillation behavior is somewhatlike a wave of water in a pool, so thepool oscillation mode is defined assloshing oscillation or swing oscillation.The scheme of the oscillation mode isgiven in Fig. 5(1a and 2a). Why did the oscillation center shift?When the weld pool grows, the bottommetal of the workpiece also graduallymelts into a liquid state. Because themicrostructure, crystal structure, anddefects/imperfections are not uniformand continuous, it is possible that theearliest spot that is melted is not exactlyat the weld pool center. The meltingprocess of the bottom pool is thus notperfectly uniform and symmetrical (instationary welding). Therefore, theasymmetrical position results in thatthe forces of the weld pool are off-bal-ance and not uniform, and the force of

the unbalance causesasymmetrical movement, like a sloshingwave of water, in the oscillating pool. In the situation of critical penetra-tion, only a very small part of the bot-tom metal melted. A large part of theweld pool was still backed by the solidmetal. Under the arc pressure, only asmall part of the liquid metal waspushed downward. With the increasedpool size, less liquid metal was backedby solid material and more liquid metalwas pushed downward. Only after thebottom size of the pool reached a cer-tain threshold, the oscillation becamethe mode of complete joint penetration. Our series of experimental resultshas also shown that this oscillation be-havior occurred under critical penetra-tion, i.e., after complete joint penetra-tion was achieved until the area of thebottom liquid surface reached approxi-mately 0.3 and 0.5 times that of the topsurface as suggested by literature (Ref.30). Thus, critical oscillation was not ata specific position as its oscillation cen-ter but in a transition stage. This is con-sistent with the results that were ob-tained in terms of oscillation frequencycharacteristics (Ref. 20). Figure 5(3aand 3b) is the top and bottom of thecorresponding weld with critical pene-tration in which the bottom pool sur-face area is less than 0.3 and 0.5 timesthat of the top pool surface.

Pool Oscillation Process forPulsed GTAW

To investigate the evolution processof pool oscillation during GTAW-P, aseries of consecutive reflected weldpool images based on laser dot-matrixwas captured at 1000 f/s. The peakand base times of the pulsing current

were both 20 ms. The other weldingparameters are given in Table 1. With-in a pulse period, 20 frames of imagesunder peak current and 20 frames ofimages under base current were cap-tured and processed. The featurepoints were extracted, the missingpoints were added by interpolation,and the point in the laser dot matrixin the image corresponding to the os-cillation center was replaced by thesymbol +, as shown in Fig. 6. As previously mentioned, the pixeldistance between two adjacent laserdots around the oscillation center of aweld pool is defined as the amplitude ofpool oscillation. When the currentswitches from the peak to base current,the arc plasma pressure suddenly releas-es. The surface tension pulls the poolback toward the equilibrium positionand the natural oscillation occurs due tothe surface tension and gravity. A differ-ent penetration state has a differentpool oscillation behavior that can be ob-served and analyzed from the consecu-tive oscillation behavior within a 40-mspulse period.

Pool Oscillation under PartialPenetration

Again, 40 frames of consecutiveweld pool oscillation images were cap-tured and processed including 20frames for the peak and 20 frames forthe base times such as in Experiment5 (Table 1). Some typical oscillationimages taken after processing areshown in Fig. 6B. Figure 6 shows theoscillation dynamic evolution processis clearly observable. The reflectedlaser pattern is distorted because ofthe pool oscillation, and its variations

WELDING RESEARCH


Fig. 6 — Pool oscillation process under partial penetration (Ip = 80 A, Ib = 20A, f = 25 Hz, Tb = 20 ms): A — Amplitude variation of pool oscillation ; B —some typical pool oscillation images. (Welding base time from 3760 to3804 ms.)

A

B


respond to the regular and periodiccontraction and expansion during theentire period. As previously mentioned, the pixeldistance between two adjacent laserdots at the weld pool oscillation centeris defined as the amplitude of pool os-cillation. The amplitude in each imagewas calculated. The relationship be-tween the amplitude and welding timeis shown in Fig. 6A. The following phe-nomena can be discerned from Fig. 6A:During the base current time when thearc plasma pressure on the top of theweld pool surface is released, the weldpool oscillates at a natural frequency,and the amplitude gradually decreaseswith weld pool solidification becauseof the reduced heat input under thebase current. The variation in the am-plitude of the weld pool surface isfrom 19 to 36 pixels during the basetime period, while the variation in theamplitude of the weld pool surface isfrom 17 to 23 pixels during the peaktime period. Because of the support ofthe solid metal in the bottom (Fig. 3),it is difficult for the pool’s surface tobe pulled to a lower position only bythe surface tension than that by thepulse arc jet pressure; therefore, ingeneral, the lowest amplitude value isnot lower than the values during thepeak current period. During peak current time, the weldpool surface is also not completelystill, possibly because of the effect ofthe “natural” oscillation in the basecurrent period. The surface tension onthe weld pool surface prevents thepool oscillation from stopping imme-diately; instead oscillation continuesat a lower amplitude — Fig. 6A. In addition, the pool’s natural fre-quency can be easily calculated fromthe pool oscillation process shown inFig. 6B. The pool oscillation has 4.5 cy-

cles within 20 ms when the weldingtime is from 3783 to 3803 ms; thus,the oscillation frequency at the mo-ment is approximately 225 Hz. The os-cillation phenomena for the other ex-periments was similar to those in Ex-periment 5. Figure 7 shows the pool oscillationprocess when the welding time wasfrom 4608 to 4628 ms with the samewelding parameters as for the experi-ment shown in Fig. 6. Because the weld-ing time was longer than that of Fig. 6,the depth of penetration is correspond-ingly deeper, the oscillation process issharper, and the oscillation amplitudefrom 10 to 40 pixels is much greaterthan that of the oscillation of weldingtime from 3782 to 3802 ms. According to Fig. 7, it can be calcu-lated that the natural oscillation fre-quency is approximately 200 Hz,which is less than the 225 Hz of Fig. 6.The results are consistent with theother studies, indicating the naturalfrequency decreases with the increasein the geometry or mass of the weldpool (Refs. 8, 29). As can be seen in Figs. 6 and 7,compared with the arc voltage varia-

tion, arc light intensity (Ref. 30), orshadowgraphy image techniques(Ref. 31), with the help of a three-dimensional laser dot matrix sensingmethod and high-speed camera, thevariation of oscillation morphologycan be observed more easily, and thedynamic evolution process of weldpool surface oscillation can be clearlyobserved and analyzed.

Pool Oscillation under Complete JointPenetration

Figure 8 shows the variation of theoscillating weld pool surface and severalcorresponding pool oscillation imagesthat are typical for complete joint pene-tration, such as those in Experiment 6(Table 1), with the welding time t0 beingequal to 6080 ms. The reflected imagesfrom the pool oscillation surface weretaken at 1, 4, 7, and 10 ms after the arcplasma pressure was suddenly removed.The pool oscillation process was just asthe scheme of oscillation under the timeaxis of Fig. 8. As indicated in the section titledsymmetrical oscillation under completejoint penetration, the pool oscillation

WELDING RESEARCH


Fig. 7 — Pool oscillation process under greater partial penetration (Ip = 80 A, Ib = 20 A, f = 25 Hz, Tb = 20 ms): A — Amplitude variation ofpool oscillation; B — some typical pool oscillation images. (Welding base time from 4608 to 4628 ms.)

B

Fig. 8 — Variation of weld pool oscillation surface under complete joint penetration (Ip =100 A, Ib = 20 A, f = 33.3 Hz, Tb = 10 ms).

A


mode and phenomena of complete jointpenetration are similar to that of thepartial penetration. However, when the weld was at complete joint penetra-tion, the metal in the bottom of theweld pool was also molten such that thebottom of the weld pool lost the sup-port of the solid metal. Instead, it wasmaintained by the surface tension ofthe liquid metal in the bottom pool. Theweight of the weld pool itself had a sig-nificant effect on pool oscillation. Theamplitude of the pool oscillation wasmuch greater than that under partialpenetration, and the motion of weldpool oscillation was sharper — Fig. 8. Apparently, the pool surface waslower than the top of the workpiecedue to the effect of surface tension andthe lack of support from bottom solidmetal. When the welding currentswitched from peak to base currentsuch that the arc jet pressure on thetop of the pool was suddenly removed,the surface tensions of the top and bot-tom of the weld pool together pulledthe pool back toward its equilibriumposition and the oscillation occurred ata natural frequency. The oscillationprocess is as shown in Fig. 8. As can beseen, the weld pool gradually expandedup from the bottom to the top. Thepool morphology is as shown in thescheme of Fig. 8 in the correspondingimages t0 + 1 ms to t0 + 10 ms. Thenthe oscillation process repeated as de-scribed previously, but the amplitudeof the pool oscillation decreased with

the solidification of theliquid pool. During thenext 40-ms period, thepool oscillation processwas similar.

Because of the lack ofthe support of the bot-tom solid metal and theweight of the weld pool

itself, the weld pool shrank downwardto a lower position, and the reflectedlaser dot-matrix at the pool center grad-ually converged to a litter area as shownin the image of t0 + 1 ms in Fig. 8. Theimage of t0 + 4 ms in Fig. 8 shows theweld pool continues to shrink, and thebrightness is the largest in the entireweld pool area. When the pool surfaceexpanded upward, the distance betweenadjacent laser dots at the center of thepool gradually became larger. However,the shape of the weld pool edge mightstill be concave, and the reflected laserdot-matrix converged. As a result, thebrightness of the area is larger, lookinglike a bright ring, as shown in the imageof t0 + 7 ms — Fig. 8. The bright ring inthe image of t0 + 10 ms became bigger,and shows that the edge part of the poolexpanded upward. If the pool were tocontinue to expand upward, and theconcaveness of the pool edge graduallybecame flat, even convex, then thebright ring would gradually disappear. When the arc jet pressure, onceagain, was exerted on the top of theweld pool surface, the weld pool sur-face concaved downward to the verylow level because of the forced actionof the arc pressure, as shown in thelast scheme of Fig. 8.

Pool Oscillation under CriticalPenetration

As mentioned previously, when the

penetration is between the partial andcomplete, the pool oscillation dynamicprocess is in the critical penetrationstage as can be seen in Fig. 9, which in-cludes 20 frames from 20-ms base cur-rent time, such as those in Experiment5 (Table 1). The lowest positions of theoscillating pool surface in the follow-ing 1–20 images are labeled with the sign. The dynamic variation of pooloscillation by the shift of the sign inthe following 1–20 images is clearlyevident. The maximum amplitude andlowest oscillation position distributionare shown in Fig. 9. Figure 9 indicates the pool surfaceis not symmetrically and vertically os-cillated as is the case in partial pene-tration and complete joint penetra-tion, and the pool oscillation center isno longer fixed at the center of theweld pool but constantly moves likethe waves on a water surface. The poolbehavior looks like the swing or slosh-ing wave of water in a pool. Figure 10 shows the variations inmaximum and minimum amplitude forsloshing oscillation during a 20-ms basecurrent time. Again, the peak currenthas an amplitude of 80 A and a durationof 20 ms. The base current has an am-plitude of 20 A and duration of 20 ms.In Fig. 10, when the welding time reach-es 7084 ms, the critical penetrationprocess occurs, and the natural oscilla-tion process is forcibly stopped after thebase time of 20 ms, because the arc jetpressure, once again, is exerted on thetop of the weld pool surface, which isderived from the pulse current. Compared with partial penetration(Fig. 7) and complete joint penetration(Fig. 8), the oscillation amplitude ofcritical penetration (Fig. 9) is mini-mum, and the oscillation frequency isnot a fixed value but varies.

WELDING RESEARCH


Fig. 9 — Sloshing pool oscillation under critical penetration(Ip = 80 A, Ib = 20 A, f = 40 Hz, Tb = 20 ms).

Fig. 10 — Variation in maximum and minimum amplitude for sloshing oscillation during a 20ms base current time.


In the critical penetration, the oscil-lation center dynamically changes. Theweld pool surface morphology isabruptly different from those in partialand in complete joint penetration. Thepool oscillation mode is also essentiallydifferent from those in partial and incomplete joint penetration. It is appar-ent the morphology variation in theweld pool surface that is clearly observ-able from the reflected images is not ob-tainable from the one-dimensional arcvoltage or arc light signal that were usedin pool oscillation studies in literature.

Relationship betweenAmplitude and Pulse CurrentDifference

To observe the pool oscillation am-plitude in different peak currents, allother welding parameters were keptthe same: 20-A base current, 5-ms du-ration for the base current, and 20-msduration for the peak current. Differ-ent peak currents were applied in dif-ferent experiments (Table 1). To ensure the conditions were thesame, only after the weld pool hadgrown to the same size was the weld

pool oscillation observed. Figure 11Bshows typical pool oscillation imagesafter processing when the peak cur-rents were 60, 80, 100, 120, 140, and160 A, respectively. Figure 11A showsthe amplitude of pool oscillation at thedifferent peak currents. The arc pres-sure increases with the square of cur-rent and decreases from electrode toworkpiece as the arc radius increases.The pressure accelerates the arc plas-ma and entrained gas toward theworkpiece to form a dynamic jet pres-sure, and the jet acts on the pool sur-face to create pressure. As shown inFig. 11A, the amplitude of pool centeroscillation gradually increases from 30to approximately 100 pixels when thecorresponding welding peak currentgradually increases from 60 to 160 A. The conclusion can be drawn that,under the condition of the same basecurrent, the greater the peak currentthe greater the pool oscillation ampli-tude. The appropriate amplitude of pooloscillation helps to improve the weldquality, gain refinement, and defect in-hibition (Refs. 29, 30); however, im-proper amplitude might influence thestability of the welding process. Thus,the relationship between amplitude and

arc pressure needs to be understood.

Discussion

The authors have studied in detailthe oscillation modes and dynamic be-haviors. However, the studies weredone under pulsed current with mod-erate frequency. To gain a broader andmore complete view, the authorswished to briefly study the oscillationsunder extreme frequencies by using di-rect current (DC) as the extreme forlow frequency and pulsing currentwith high frequency.

Pool Dynamic Behavior under DirectCurrent GTAW

To study the behavior under ultra-low frequencies, (DC) was used. Asshown in Fig. 12, the amplitude of thepool center during DC-GTAW was alsomeasured based on the same methoddescribed previously. Figure 12Bshows typical images for DC-GTAW af-ter image processing. The symbol still represents the weld pool oscilla-tion center. Figure 12A shows ampli-tude variations from consecutive 1–40ms welding time from 1577 to 1617ms. Figure 12A shows the amplitudeof the pool oscillation center almostremains constant since the weldingcurrent stays nearly constant duringDC-GTAW, although some slightchanges for the amplitude might haveresulted from minor disturbances. Ac-cording to the sections titled Pool Os-cillation under Partial Penetration andPool Oscillation under Complete Jointpenetration, the pool behaviors duringDC-GTAW differ entirely from those inGTAW-P during the peak period.

Forced Pool Oscillation under HighFrequency Pulsing Current

As mentioned previously, when thearc jet pressure on the top of the weldpool is released, the pool oscillate is ex-cited at the natural frequency corre-sponding to the geometry of the weldpool. However, when the base time isshorter than the period of the pool os-cillation at the national frequency, theoscillation process would be forcibly ter-minated. Figure 7 shows that after thecurrent is switched at 4608 ms from 80-A peak current to 20-A base current,which lasts 20 ms, the natural oscilla-

WELDING RESEARCH


Fig. 11 — Behavior of weld pool oscillation under different peak currents (Ib = 20 A, f = 40Hz, Tb = 5 ms): A — Amplitude variation of pool oscillation in different peak currents; B —typical pool oscillation images for different peak currents.

A

B


tion frequency of the weld pool is 200Hz. The oscillation period is 5 ms.Clearly, if the base time is less than 5ms, such as in Experiment 3 (Table 1),the oscillation process will be terminat-ed before a complete natural oscillationcycle is finished. Natural oscillation willnot occur. The pool oscillation frequen-cy is expected to be equal to that of thepulse frequency of GTAW-P. In this case,the oscillation will be much strongersuch that the images become blurrier. To analyze such oscillation, the im-ages have been enhanced to increase thecontrast as in Fig. 13 where the basecurrent is 3 ms where the fourth imagewas acquired 3 ms after the peak cur-rent had been applied. Despite the rela-tively blurrier images, it is still clear thatthe reflection in the fourth image showsa concave pool surface (as indicated bythe bright spot in the center) ratherthan a convex pool surface as it shouldbe if the natural oscillation continues.

Conclusion

Pool oscillation behavior has beencharacterized by and analyzed fromthe reflection laser dot-matrix patternfrom the oscillating weld pool surface.From reflection laser dot-matrix pattern-based analyses, the followingconclusions can be drawn for the pooloscillation in GTAW-P: 1) Pool oscillation dynamic behav-iors can be clearly observed and easilymeasured according to the reflectedimages derived from the laser dot-matrix sensing method. 2) The dynamic evolution processof the oscillating weld pool and thevariation in the weld pool surface mor-phology can be clearly presented usingthe innovative three-dimensional weldpool surface sensing method. 3) There exists three oscillationmodes: symmetrical oscillation forpartial penetration, sloshing oscilla-tion for critical penetration, and sym-metrical oscillation for complete jointpenetration, which differs significant-ly from that of partial penetration. 4) The amplitude of the pool oscil-lation gradually increases as the peakcurrent increases. This is because thearc plasma pressure on top of the poolsurface increases as the peak currentincreases. 5) The natural frequency oscillationcan be excited after the current is re-

duced to the base level but its continu-ation to finish a complete cycle re-quires the base current period be suffi-ciently long in comparison with theperiod of natural frequency. A relative-ly short base current period will resultin a forced oscillation at the frequencyof the pulsed current. 6) Oscillations with smaller ampli-tude still exist in the weld pool duringthe peak current period. However,during welding with a constant cur-rent, the oscillation in the weld pool isinsignificant.

The financial support from the Chi-na Scholarship Council is greatly ap-preciated. This work is also partiallysupported by the National ScienceFoundation under grant NSF 1208420and the University of Kentucky Insti-tute for Sustainable Manufacturing.

1. Zhang, Z. Z., and Wu, C. S. 2015. Ef-fect of fluid flow in the weld pool on thenumerical simulation accuracy of the ther-mal field in hybrid welding. Journal of Man-ufacturing Processes 20: 215–223. 2. Wu, C. S., Wang, L., Ren, W. J., andZhang, X. Y. 2014. Plasma arc welding:Process, sensing, control and modeling.Journal of Manufacturing Processes 16:74–85. 3. Liu, Z. M., Wu, C. S., Liu, Y. K., andLuo, Z. 2015. Keyhole behaviors influenceweld defects in plasma arc welding process.Welding Journal 94(9): 281-s to 290-s. 4. Chen, S. B., and Lv, N. 2014. Re-search evolution on intelligentized tech-nologies for arc welding process. Journal ofManufacturing Processes 16: 109–122. 5. Bahrami, A., Aidun, D. K., and Valen-tine, D. T. 2014. Interaction of gravityforces in spot GTA weld pool. Welding Jour-nal 93(4): 139-s to 144-s. 6. Bahrami, A., and Aidun, D. K. 2014.Modeling of carbon steel-duplex stainlesssteel GTA weld pool. Welding Journal 93(7):

WELDING RESEARCH


Fig. 12 — Pool dynamic behavior under DCGTAW (welding current I = 60 A): A — Amplitudevariation during DCGTAW; B — some typical images for DCGTAW.

Fig. 13 — Forced pool oscillation under highfrequency pulse current (Ip = 80 A, Ib = 20 A, f= 43.5 Hz, Tb = 3 ms).

A

A B C D

B

References

Acknowledgments


262-s to 270-s. 7. Ahsan, M. R. U., Kim, Y. R., Ashiri,R., Cho, Y. J., Jeong, C., and Park, Y. D.2016. Cold metal transfer (CMT) GMAWof zinc-coated steel. Welding Journal 95(4):120-s to 132-s. 8. Kotecki, D., Cheever, D., and How-den, D. 1972. Mechanism of ripple forma-tion during weld solidification. WeldingJournal 51(8): 368. 9. Yudodibroto, B., Hermans, M., Hira-ta, Y., and den Ouden, G. 2013. Influenceof filler wire addition on weld pool oscilla-tion during gas tungsten arc welding. Sci-ence and Technology of Welding & Joining. 10. Tewari, S. 1999. Effects of oscilla-tion on impact property of weldments. ISIJInternational 39: 809–812. 11. Methong, T., and Poopat, B. 2013.The effect of ultrasonic vibration on prop-erties of weld metal. Key Engineering Mate-rials, Trans Tech Publications, pp. 177–181. 12. Kageler, C., and Schmidt, M. 2010.Frequency-based analysis of weld pool dy-namics and keyhole oscillations at laserbeam welding of galvanized steel sheets.Physics Procedia 5: 447–453. 13. Jose, M., Kumar, S.S., and Sharma,A. 2016. Vibration assisted weldingprocesses and their influence on quality ofwelds. Science and Technology of Welding andJoining 21: 243–258. 14. Huang, C., and Kou, S. 2003. Liqua-tion cracking in partial-penetration alu-minum welds: Effect of penetration oscilla-tion and backfilling. Welding Journal 82(7):184-s to 194-s. 15. Hermans, M., and den Ouden, G.1999. Process behavior and stability inshort circuit gas metal arc welding. WeldingJournal 78(4): 137-s to 141-s. 16. Dowden, J., and Kapadia, P. 1999.Oscillations of a weld pool formed by melt-ing through a thin workpiece. Lasers in En-gineering (UK) 8: 311–318. 17. Cho, J., Farson, D. F., Hollis, K. J.,and Milewski, J. O. 2015. Numerical analy-sis of weld pool oscillation in laser welding.Journal of Mechanical Science and Technology29: 1715–1722. 18. Xiao, Y., and den Ouden, G. 1990. Astudy of GTA weld pool oscillation. WeldingJournal 69(8): 289-s to293-s. 19. Xiao, Y., and den Ouden, G. 1992.Direct observation of GRA weld pool oscil-lation. International Trends in Welding Sci-ence and Technology, pp. 423–426. 20. Xiao, Y., and den Ouden, G. 1993.Weld pool oscillation during GTA weldingof mild steel. Welding Journal 72(8): 428-sto 434-s.

21. Shi, Y., Zhang, G., Li, C., Gu, Y., andFan, D. 2015. Weld pool oscillation fre-quency in pulsed gas tungsten arc weldingwith varying weld penetration. IEEE Inter-national Conference on Automation Scienceand Engineering (CASE). IEEE, pp.401–406. 22. Shi, Y., Zhang, G., Ma, X., Gu, Y.,Huang, J., and Fan, D. 2015. Laser-vision-based measurement and analysis of weldpool oscillation frequency in GTAW-P.Welding Journal 94(5): 176-s to 187-s. 23. Yu, S., Chunkai, L., Leiming, D.,YuFen, G., and Ming, Z. 2016. Frequencycharacteristics of weld pool oscillation inpulsed gas tungsten arc welding. Journal ofManufacturing Processes 24: 145–151. 24. Tarn, A., and Hardt, D. 1989. Weldpool impedance for pool geometry meas-urement: Stationary and non-stationarypools. Journal of Dynamic Systems, Measure-ment, and Control 101: 545–553. 25. Kovacevic, R., Zhang, Y., and Li, L.1996. Monitoring of weld joint penetrationsbased on weld pool geometrical appearance.Welding Journal 75(10): 317-s to 329-s. 26. Joshi, Y., Dutta, P., Schupp, P., andEspinosa, D. 1997. Nonaxisymmetric con-vection in stationary gas tungsten arc weldpools. Journal of Heat Transfer 119:164–172. 27. Andersen, K., Cook, G.E., Barnett,R. J., and Strauss, A. M. 1997. Synchro-nous weld pool oscillation for monitoringand control. IEEE Transactions on IndustryApplications 33: 464–471. 28. Aendenroomer, A., and den Ouden,G. 1998. Weld pool oscillation as a tool forpenetration sensing during pulsed GTAwelding. Welding Journal 77(5): 181-s to187-s. 29. Renwick, R., and Richardson, R.1983. Experimental investigation of GTAweld pool oscillations. Welding Journal62(3): 29-s to 35-s. 30. Yoo, C. D., and Richardson, R. W.1993. An experimental study on sensitvityand signal characteristics of welds pool os-cillation. Transactions of the Japan WeldingSociety 24: 54–62. 31. Ramos, E. G., Carvalho, G. C. D.,and Alfaro, S. C. A. 2013. Analysis of weldpool oscillation in P-GMAW by means ofshadowgraphy image processing. Soldagem& Inspeçao 18: 39–49. 32. Matsuda, F., Ushio, M., Nakata, K.,and Miyanaga, Y. 1978. Effects of currentpulsation on weld solidification structureof aluminum alloys. Transactions of JWRI 7:287–289. 33. Yamamoto, H., Harada, S., Ueyama,

T., Ogawa, S., Matsuda, F., and Nakata, K.1993. Beneficial effects of low‐frequencypulsed MIG welding on grain refinement ofweld metal and improvement of solidifica-tion crack susceptibility of aluminium al-loys: Study of low‐frequency pulsed MIGwelding. Welding International 7: 456–461. 34. Zacksenhouse, M., and Hardt, D.1983. Weld pool impedance identificationfor size measurement and control. Journalof Dynamic Systems, Measurement, and Con-trol 105: 179–184. 35. Pearce, B., and Kerr, H. 1981. Grainrefinement in magnetically stirred GTAwelds of aluminum alloys. MetallurgicalTransactions B 12: 479–486. 36. Brown, D., Crossley, F., Rudy, J., andSchwartzbart, H. 1962. The effect of elec-tromagnetic stirring and mechanical vibra-tion on arc welds. Welding Journal 41(6):241-s to 250-s. 37. Mukae, S., Katoh, M., Nishio, K.,and Tsutsui, T. Effect of electromagneticstirring on weld solidification structure of5052 alloy. Kei Kinzoku Yosetsu (J. LightMet. Weld. Constr.) 22: 299–309. 38. Vives, C. 1989. Electromagnetic re-fining of aluminum alloys by the CREMprocess: Part I. Working principle and met-allurgical results. Metallurgical TransactionsB 20: 623–629. 39. Zhang, W., Liu, Y., Wang, X., andZhang, Y. 2012. Characterization of threedimensional weld pool surface in GTAW.Welding Journal 91(7): 195-s to 203-s. 40. Zhang, W. 2014. Machine-human co-operative control of welding process. PhDdissertation. University of Kentucky, Dept.of Electrical Engineering, Lexington, Ky. 41. Zhang, W., Wang, X., and Zhang, Y.2013. Analytical real-time measurement ofa three-dimensional weld pool surface.Measurement Science and Technology 24:115011. 42. Zhang, Y. M., Beardsley, H. E., andKovacevic, R. 1994. Real-time image pro-cessing for 3D measurement of weld poolsurface. The 1994 International MechanicalEngineering Congress and Exposition, pp.255–262. 43. Song, H., and Zhang, Y. 2007. Animage processing scheme for measurementof specular weld pool surface. 2007 2nd

IEEE Conference on Industrial Electronics andApplications, IEEE, pp. 5–10. 44. Song, H., and Zhang, Y. M. 2008.Measurement and analysis of three-dimen-sional specular gas tungsten arc weld poolsurface. Welding Journal 87(4): 85-s to 95-s.

WELDING RESEARCH


K. E. ZHANG ([email protected]), SHAOJIE WU, YUMING ZHANG ([email protected]), and JINSONG CHEN are with the Institute for Sustainable Manufacturing, University of Kentucky, Lexington, Ky. K. E. ZHANG is also with the Welding and Laser Processing Instite, Shanghai Jiaotong University, Shanghai, China.


observation and analysis of three­dimensional weld pool ... · pdf filepool and verified...

Documents

observation and analysis of threedimensional weld pool ... · pdf filepool and verified...