WJ_2014_06_s205

JUNE 2014 / WELDING JOURNAL 205-s

Introduction Gas metal arc welding (GMAW) isone of the most widely used processesin the manufacturing industry. DuringGMAW, an electrode wire is fed to thecontact tip that is typically connectedto the positive terminal of the powersource. The arc melts the wire andmelted metal forms a droplet at the tipof the wire. After the droplet is

detached/transferred, a new dropletstarts to form and a new cycle starts.The type of metal transfer plays a crit-ical role in determining the quality andproductivity of GMAW. The American Welding Society clas-sifies metal transfer into three majortypes as follows: short circuiting, glob-ular, and spray transfer (Ref. 1). Theparameter that plays the most impor-tant role in determining the metal

transfer without short circuiting, i.e.,globular or spray transfer, is thecurrent. To detach a droplet withoutshort circuiting, the detaching force,primarily the electromagnetic forcethat is due to and increases with thewelding current and gravitational forceproportional to the droplet mass,must exceed the retaining force,primarily the surface tension at thesolid wire-liquid droplet interface thatis approximately fixed for the givenwire diameter and material. If the cur-rent is sufficient such that the detach-ing electromagnetic force is close to orexceeds the retaining surface tension,the droplet will be detached without aneed for additional gravitational force. Researchers proposed methods toachieve stable metal transfers (Refs.2–21) at currents lower than the tran-sition current (Ref. 1). At the Univer-sity of Kentucky, the laser-enhancedGMAW process was developed by pro-jecting a 1-kW laser stripe of 14 × 1mm onto the droplet (Refs. 22–25).The droplets were detached atcurrents lower than the transition cur-rent (Ref. 1) with the assistance of therecoil pressure from the laser.However, the length of the laser stripewas aligned with the wire and droplet.While this alignment provided an easyway to apply the auxiliary force fromthe laser on the droplet, only afraction of this 1-kW laser waseffectively utilized. The rest waswasted and adversely applied on the

Gas Metal Arc Welding Enhancedby Using a Pulsed Laser

Y. SHAO and Y. M. ZHANG ([email protected]) are with the Institute for Sustainable Manufacturing and Department of Electrical and Computer Engineering, University of Kentucky, Lexington, Ky.

ABSTRACT Laser-enhanced gas metal arc welding (GMAW) is a novel process where alaser is applied to provide an auxiliary force to help detach droplets at reducedcurrents. In previous studies, a continuous laser was applied. Since this auxiliaryforce is only needed each time the droplet needs to be detached and the detach-ment time is relatively short in a transfer cycle, the laser energy is greatly wasted.In addition, unnecessary application of the laser on the droplet producesadditional fumes. Hence, this study proposes to use a pulsed laser whose peak isapplied when the droplet is ready to detach. To implement this approach, thelaser aims at a predetermined desirable position, and the droplet’s position ismonitored in real time using a high-speed image processing system. When thedroplet moves to the position aimed by the laser, the closed-loop control systemcommands the laser to pulse to the peak. The current is also pulsed to the peak insynchronization with the laser to combine the increased electromagnetic forcewith the increased laser recoil force to detach the droplet. Experimental resultsverified the effectiveness of the proposed metal transfer control method thatmonitors the droplet in real time as well as applies synchronized laser andcurrent pulses when the droplet moves to the desired position aimed by the laser.

KEYWORDS •LaserEnhanced Gas Metal • Arc Welding (GMAW) • Auxiliary Force

• Droplet • RealTime Visual Feedback

Two series of experiments have been conducted with the system operatinga realtime visual feedback to apply a laser spot of 50W peak power

in synchronization with a current pulse

BY Y. SHAO AND Y. M. ZHANG

WELDING RESEARCH

Shao Supplement new_Layout 1 5/15/14 4:17 PM Page 205

WELDING JOURNAL / JUNE 2014, VOL. 93206-s

WELDING RESEARCH

workpiece. Furthermore, to ensure thelaser’s presence when needed, it wascontinuously applied. Its energy wasfurther wasted and adversely applied

onto the workpiece. More critically,because of the presence of the contin-uous force due to the laser, the detach-ment occurred anytime when the sur-face tension was balanced out by com-bined detaching forces due to thelaser, droplet mass (gravitationalforce), and electromagnetic force.Thus, the detachment time and masswere not controlled although smallerdroplets could be detached with rela-tively small currents. This study applies a pulsed laserspot onto the droplet at the right lo-cation/time where/when the detach-ing force from the laser is needed. As

a result, the laser-enhanced GMAWprocess becomes a pulsed process andchanges from an uncontrolled to acontrolled process. The droplet massand detaching time are controlled.The waste of the laser energy and theeffect of the unintentionalapplication of the laser on the work-piece are also eliminated. To realizethis goal, the metal transfer process ismonitored in real time. When thedroplet enters the location aimed bythe laser, a laser pulse is applied. Toreduce the need for laser energy to alevel achievable using the existinglaser spot, the current is pulsed

Fig. 1 — Block diagram of theproposed control algorithm.

Fig. 2 — Experimental system diagram.

Fig. 3 — Installation of the proposed system with d1 = 15 mm, d2 = 10 mm, and θ = 30deg. A — View 1; B — view 2

Table 1 — Major Parameters and Comparative Experiments

Experiment # Peak Current Peak Current Base Current Minimal Base Average Laser(A) Period (MS) (A) Current Period Current Power

(ms) (A) (W)

0 160 6 28 30 50 N/A1 150 6 28 30 48.3 N/A2 150 6 28 30 48.3 503 145 8 28 30 52.6 N/A4 145 8 28 30 52.6 50

A B

Shao Supplement new_Layout 1 5/15/14 4:17 PM 06


WELDING RESEARCH

simultaneously with the laser. Tomake sure the pulses of the laser andcurrent are correctly synchronized,monitoring of the metal transferprocess is done using a high-speedimaging system and real-time image-processing algorithm. The accuratemonitoring of the droplet using theimage method can eliminate theeffect of the variations in the wirefeed speed, melting speed, andcontact tube-to-work distance suchthat the synchronization can be accu-rately ensured.

Control Principle The principle is to monitor the po-sition of the droplet in real time andapply synchronized laser and currentpulses when it passes the laser aimingline such that the laser recoil force iscombined with the increased electro-magnetic force to detach the droplet. First, the droplet needs to be moni-tored in real time. To monitor the

welding process here, many methodshave been developed. Traditionalmethods for welding process monitor-ing include arc sensing (Ref. 26), radi-ography (Ref. 27), infrared sensors(Ref. 28), ultrasonic sensors (Refs. 29,30), and machine vision systems (Refs.31–38). They are also often fast andaccurate enough to monitor majorvariables in the welding process andhave been widely studied for monitor-ing the weld joint, weld profile, weldpenetration, and weld pool surface. A successful operation of pulsedlaser-enhanced GMAW relies on real-time accurate detection of the develop-ing droplet. Machine vision should bethe most direct method to detect thedroplet accurately. Researchers haveused high-speed machine visionsystems to observe the dropletthrough direct viewing and laser back-lighting for off-line analysis. At theUniversity of Kentucky WeldingResearch Lab, droplet trackingalgorithms using brightness-based

separation algorithms have been pro-posed (Refs. 39, 40). However, most ofthe works in this area have been basedon off-line analysis and feasibilitystudies. Real-timemonitoring/measurement of thedevelopment of droplets in GMAW re-mains a challenge because of the highspeed of the metal transfer process. Aswill be seen, the authors have made itpossible to real-time measure the de-veloping droplet and use the measureddroplet to effect the proposed controlthat determines when to apply thelaser and current pulses. Second, to accurately control themetal transfer such as the droplet isdetached, this work uses a constantcurrent (CC) power source thatoutputs the current per the controlcommand. Using machine vision-based monitoring, peak current andlaser pulses are applied when thedroplet crosses the desired location todetach the droplet. The droplet doesnot detach if it is not at the desired lo-

Fig. 4 — Droplet growth period during the base current time. A — Serial images in selected regions of interest; B — fitted edges; C —comparison of the fitted edges with the original images.

Table 2 — Parameters For Successful Detachment Experiments

Experiment # Peak Current Peak Current Base Current Minimal Base Average Laser Power(A) Period (ms) (A) Current Period (ms) Current (A) (W)

5 150 6 28 30 48.3 506 145 6 28 30 47.5 507 140 6 28 30 46.7 508 135 6 28 30 45.8 50

A B C



WELDING RESEARCH

cation aimed by the laser and will onlybe detached when it is at the desiredposition. The metal transfer process isaccurately controlled. Third, the proposed control systemutilizes a computer system to performboth image acquisition and controllerfunctions. The block diagram of thecontrol algorithm is given in Fig. 1.Specifically, an arc start program (ini-tialization) will first commence to es-tablish a stable arc between theelectrode and base metal before thecontrol algorithm starts to function.After this arc starting period, the con-trol system will take over and give outa voltage signal to the CC power sourceto apply the base current Ib for aperiod Tb0 during which it is impossi-ble for the droplet to reach the desired

position. Then the camera starts tocapture an image and store it in thepreallocated buffers through the framegrabber. The computer will process theacquired image using the imageprocessing algorithm to determine theposition of the droplet in real time. Ifthe droplet is determined to have notreached the desired position yet, a newimage will be acquired and processed.Otherwise, control commands will besent to the laser controller and powersource to apply the laser and currentpulses for a given peak period Tpbefore the laser and current are againreduced to their base levels to begin anew cycle. Fourth, in the control system, theframe grabber supports two kinds ofworking mechanisms — continuously

transmitting new frames from thebuffer (the normal mode) or workingin a wait-done mode. In the normalmode, the frame grabber continuouslyrefreshes the pointer to grab eachframe to the upper level of the controlsystem regardless of the actual timethe image-processing algorithm maytake to finish the processing task. Thedrawback is that if the speed of theimage processing is slower than that ofthe image acquisition/grabbing, theirdifferences would accumulate suchthat the control would inevitably fail.On the other hand, in the wait-donemode, if the image processingalgorithm takes too long to finish pro-cessing a frame, the frame grabber willwait until the processing is finishedbefore it neglects all other frames

Fig. 5 — Typical first pulsing cycle todetach a droplet without a laser using150 A of current for 6 ms. The intervalbetween consecutive images is 1 ms.Droplet size: 1.42 mm2 (image l).

Fig. 6 — A typical second pulsing cycleto detach a droplet without a laserusing 150 A of current for 6 ms. Dropletsize: 2.48 mm2 (image l).

Fig. 7 — A typical pulsing cycle to detach a droplet with a laser using 150Acurrent for 6 ms. Droplet size: 1.42 mm2

(image j).

Table 1 — Major Parameters and Comparative Experiments

Experiment # Peak Current Peak Current Base Current Minimal Base Average Laser(A) Period (MS) (A) Current Period Current Power

(ms) (A) (W)

0 160 6 28 30 50 N/A1 150 6 28 30 48.3 N/A2 150 6 28 30 48.3 503 145 8 28 30 52.6 N/A4 145 8 28 30 52.6 50



WELDING RESEARCH

stored during the waiting time to givethe upper level to the newest frame ac-quired. While the latter may cause dis-continuities in processing results, thecontrol process itself would continueas the welding process runs.Fortunately, the authors havedeveloped an effective real-timeimage-processing algorithm (see thelater description on the image-processing algorithm) such that itsspeed is adequate for the normalmode. Hence, the normal mode is cho-sen over the wait-done mode althoughthe latter may be considered in casemore complex algorithms are needed. When the camera begins to captureimages, the captured data/images willbe written into the buffers. After anentire frame is written, the pointerwill jump to the current frame and theimage will be analyzed by the image-processing algorithm. If the droplethas not reached its desired positionaimed by the laser, the system will cap-ture and analyze an image again. Oth-erwise, the control system will sendsignals to apply the peak current andlaser pulses. Their combined detachingforce will detach the droplet. If for anyreason the droplet has passed the posi-tion aimed by the laser, or the image-processing algorithm fails to give a

valid result, a higher peak current canbe applied to forcefully detach thedroplet without the laser.

Experimental System andConditions Figure 2 is the diagram for theexperimental system established toimplement the pulsed laser-enhancedGMAW (pulsed LE-GMAW) andconduct experiments. The conventional GMAW systemused includes a CC power source,GMAW gun, wire feeder, and weldingpositioner and motion system. Steelpipes are fixed to a minipro weldingpositioner that rotates the pipe. A small spot laser with a lowerpower is used to replace the previouslyused laser stripe of much greaterpower to generate the auxiliarydetaching force to the droplet. It is afiber laser with maximum outputpower of 50 W and its power can beadjusted in real time by providing ananalog voltage signal. Its focused areais a spot whose diameter isapproximately 1 mm. Because thisfiber laser has a wavelength between960 and 980 nm in the infrared range,it is invisible and equipped with a visi-

ble guide laser. The high-speed machine vision sys-tem used to monitor the droplet inreal time consists of a micro lens thatallows aiming at a small object from along distance, a high-speed camera(camera 1), and a frame grabber. Thelong distance prevents the camera andlens from being damaged by possiblespatter from GMAW. The high-speedcamera is a Gazelle gzl-cl-22c5mequipped with a 200-mm f/4D microlens to capture the images reflectingthe droplet growth process from a rel-atively large distance. The gzl-cl-22c5m camera has a CMOS CMV2000sensor with a diagonal of 1.7 cm. Itsfastest shutter speed is 5.5 μm. For2048 × 1088 pixels resolution, thehighest frame rate is 280 frames persecond (fps). For this research wherethe frame rate needs to be relativelyhigh, the resolution is set to 200 × 800pixels to allow the camera to captureimages at 1200 fps. A frame grabbercompatible with this camera has beenselected. A second high-speed camera (cam-era 2) was also used to record imagesfor off-line analysis. This high-speedcamera is capable of recording themetal transfer at 33,000 fps. A band-pass filter centered at 810±2 nm with

Fig. 8 — A typical pulsing cycle to detach a droplet with a laser using 150 Aof current for 6 ms after a successful detachment. Droplet size: 1.23 mm2

(image l).

Fig. 9 — A typical first pulsing cycle todetach a droplet without a laser using145 A of current for 8 ms. Droplet size:1.43 mm2 (image l).

Fig. 10 — A typical second pulsingcycle to detach a droplet without alaser using 145 A of current for 8 ms.Droplet size: 2.29 mm2 (image k).



WELDING RESEARCH

full width at half maximum 10±2 nmwas used to observe the process andrecord the images. All imagespresented in this paper were recordedusing this high-speed camera at 3000fps with this filter. A data acquisition board, with two16-bit analog input channels with asample rate of 250 KS/s, two 16-bit

output channels with an update rate of833 KS/s, and eight digital input/out-put channels with a maximum clockrate of 1 MHz, is used to interfacewith the welding power source to con-trol its current waveform. The diameter of the steel wireused, i.e., ER70S-6, is 0.8 mm. Pureargon is used as the shielding gas and

the flow rate is 14 L/min. (Usingmixed gas would add gas compositionas an additional parameter to compli-cate the study, but the methodpresented in this study should applyto argon-rich mixed gases.) The bead-on-plate experiments are conductedon 11.43-cm OD, 2.1-mm-thick pipes.The travel speed (converted from therotation speed) is 6 mm/s. The peakpower of the laser is set to itsmaximum output, 50 W. The contacttube-to-work distance (CTWD) isaround 1.5 cm, but it varies as thepipe surface is not perfectly symmet-rical about the rotation axle and thetorch position is fixed. The desiredwire extension at which the laserneeds to be applied is 1 cm. Figure 3A, B shows the importantparameters for the proposed laser-en-hanced GMAW system. Camera 1 isplaced along x-axle and sits 1.2 m awayfrom the torch. Camera 2 records theprocess in front of the whole systemwith the distance between camera 2and the torch to be approximately 1.5m. The view direction of camera 2 isparallel with the y-axle. To develop the image-processing al-gorithm, the pulsed laser-enhancedGMAW experiments have beenconducted using 150-A peak currentand 28-A base current to acquireimages. Images were grabbed usingcamera 1 that will be used in the real-time control system. The wavelengthof the laser is invisible for this camera. Figure 4A shows a series ofcaptured frames by camera 1 duringthe base current period. (In theproposed control system, the dropletis monitored in real time only duringthe base current period in order to de-termine when the peak current andlaser pulses should be applied.) Theseframes are captured from anuncontrolled GMAW process showinga droplet growth period during thebase current period. An algorithm hasbeen developed (Ref. 41) to extract thedroplet contours, and the extracteddroplet contours are given in Fig. 4B.As can be seen in Fig. 4C, the extractedcontours match well with theiroriginal images. The droplets can beextracted to provide feedback on thedroplet movement. Furthermore, theimage-processing algorithm has beenimproved to extract the droplet posi-tion in 1000 fps in order to providethe needed real-time feedback on the

Fig. 11 — A typical pulsing cycle to detach a droplet with a laser using 145 Aof current for 8 ms. Droplet size: 1.45mm2 (image j).

Fig. 13 — Typical two consecutive cycles in Experiment 6 with 145 A of peak current. A— Cycle one (1.36 mm2 in image h); B — cycle two (1.39 mm2 in image h).

Fig. 12 — A typical pulsing cycle to detach a droplet with a laser using 145 Aof current for 8 ms after a successfuldetachment. Droplet size: 1.43 mm2

(image j).

A B

A B



WELDING RESEARCH

position of the droplet in relation tothe prealigned laser (Ref. 41). The im-proved algorithm for the 1000 fps pro-cessing speed will be used in the con-trol system in this study.

Experiments

Comparative Experiments

Comparative experiments are con-ducted with/without laser usingexactly the same welding parametersin order to systematically verify the ef-fectiveness of the proposed modifica-tion to the laser-enhanced GMAWusing a pulsed laser spot to replace thecontinuous laser stripe. Table 1 liststhe major parameters for the designedcomparative experiments. Previous experiments have beendone by increasing the peak current in5-A increments. It was found that thepeak current must be at least 160 A orhigher in order to detach the dropletwithout a laser. Hence, 160 A was usedas the minimal peak current needed todetach the droplet for a 0.8-mm steelwire. For convenience, 160 A is used asthe reference peak current hereafter.Since the success in detachment alsodepends on the peak current duration,the peak current and its applicationtime will both be included in thediscussion and analysis in this paper.For convenience, this experiment islisted as Experiment 0 in Table 1. As can be seen, in all experiments,the minimal base current period dur-ing which the droplet is not monitoredis 30 ms although it can be easilychanged. After the minimal base cur-rent period, the droplet is monitoredin real time using the vision systemand can be controlled in exactly thesame ways except for the applicationof the laser. That is, after the base cur-rent is applied for the specified mini-mal base current time, i.e., 30 ms asshown in Table 1, the vision systemstarts to monitor the position of thedroplet. When the droplet reaches thedesired position, i.e., 10 mm below thecontact tip, as determined from the vi-sion system, the peak current specifiedin Table 1 is applied for the periodspecified in the table with or withoutthe application of the laser. During theapplication of the peak current (withor without laser) and the followingminimal base current application

period, no images are processed in realtime to detect the droplet position.

Experiment 1 — 150A Peak Currentfor 6 ms without a Laser

In Experiment 1, the peak currentis 150 A, 10 A lower than the referencepeak current, and its duration is 6 mswhile the base current and its minimalduration are 28 A and 30 ms,

respectively. The wire extension totrigger the peak current pulse, i.e., forthe droplet to reach the desired posi-tion, is 10 mm away from the contacttip. The corresponding average weld-ing current is 48.3 A if the dropletreaches the desired position fordetachment right after the minimalbase current period such that theactual base current period is the mini-mum as given in Table 1, i.e., 30 ms.



A B

A B



WELDING RESEARCH

Otherwise, it will be lower if an extraperiod of time takes place for thedroplet to reach the desired positionafter the minimal base current period.In this comparative experiment, alaser is not applied. Figure 5 shows a droplet being de-tached in Experiment 1. Frames athrough l are selected from therecorded images with 1 ms interval(also applies to the rest of the imageseries in the paper unless otherwisespecified) to show that the droplet isbeing pushed down by the metal wirestep by step. When the droplet reachesthe desired point at frame a, a peakcurrent of 150 A is applied. However,at this time, the size of the droplet isnot sufficient for its correspondinggravitational force to compensate forthe insufficiency of theelectromagnetic force in detaching thedroplet. As a result, after the peak cur-rent is applied for the specified dura-tion (from frames b to h), the dropletis still not detached. The one dropletper pulse mode needed to qualify for asuccessful metal transfer control is notachieved. Since the droplet volume requiresthree-dimensional measurements,which are not available, to calculate,the authors propose to use its two-dimensional area to measure its size.To calculate the area from the image,the droplet is approximated as anellipse such that the area can be calcu-lated from the height and width meas-ured from the image whosecoordinates have been calibrated toreal-world dimensions. Using thismethod, the droplet size in image l inFig. 5 is measured at 1.42 mm2. Figure 6 shows the next cycle thatfollows that in Fig. 5. After the previ-ous peak current period and another30-ms minimal base current period,the droplet has grown much larger. In

addition, it has reached/exceeded thedesired position for detachment as canbe seen by comparing frames athrough c (they should still be in the30-ms minimal base current period) inFig. 6 with frame a in Fig. 5. Unfortu-nately, despite the much increaseddroplet size, the peak current pulsestill failed to detach the droplet duringthe peak current period (frames dthrough i). By the end of the peak cur-rent period (frame i), the droplet grewmuch further. The gravitational forcebecame the dominant force exceedingthe retaining force provided by thesurface tension at the interfacebetween the solid wire and moltendroplet. As a result, the droplet wasdetached from the wire in thebeginning of the following basecurrent period (frame l). Repeated experiments have beenconducted with exactly the same nom-inal experimental parameters. Therecorded videos were analyzedcarefully. It is found that the multiplepulses one drop phenomenonobserved in Figs. 5 and 6 occurscontinuously.

Experiment 2 — 150A Peak Currentfor 6 ms with a Laser

In Experiment 2, all the nominalparameters remain the same as thosein Experiment 1, but the 50-W laser isapplied to the droplet when it reachesthe desired position. Again, this posi-tion for applying the laser and currentpulses to detach the droplet is still 10mm away from the contact tip. Although a laser has been appliedto the droplet, neither camera is ableto capture the laser because the 960-nm wavelength of the laser is too highfor the sensors in the high-speed cam-eras used. Hence, the aiming positionand power of the laser have been pre-

calibrated and pretested. Figure 7 shows the metal transferphenomenon in Experiment 2. At thebeginning of the cycle, the size of thedroplet is at an average level. Whenthe peak current is applied, the laser issimultaneously applied to thepredefined aiming point. This time, al-though the size of the droplet beforethe pulse is applied is similar as that inframe a in Fig. 5 (compare frame b inFig. 7 with frame a in Fig. 5), thedroplet is successfully detached, as canbe seen in frame j in Fig. 7. Figure 8 shows the metal transferin the next cycle that follows. Becausethe previous droplet has beendetached, after the minimal base cur-rent period, the size of this droplet(1.23 mm2 in frame l) is similar to thatin Fig. 7 (1.42 mm2 in image j) andconsiderably smaller than the size ofthe droplet in Fig. 6 (2.48 mm2 inimage l). This time, when the peak cur-rent and laser pulses are applied, thedroplet is again detached as can beseen in Fig. 8 (frames k and l). Repeated experiments have beenconducted with exactly the samenominal experimental parameters.The recorded videos were analyzedcarefully. It is found that the desiredone drop per pulse in Figs. 7 and 8continuously repeats. The applicationof the laser made the difference tochange the metal transfer from theundesirable multiple pulses one dropof large droplets to the desired onedrop per pulse with small droplets.

Experiment 3 — 145A Peak Currentfor 8 ms without a Laser

In Experiment 3, the peak currentis reduced to 145 A, but the peak cur-rent period is increased to 8 ms. Thebase current and minimal base currentperiod are still 28 A and 30 ms

Table 2 — Parameters For Successful Detachment Experiments

Experiment # Peak Current Peak Current Base Current Minimal Base Average Laser Power(A) Period (ms) (A) Current Period (ms) Current (A) (W)

5 150 6 28 30 48.3 506 145 6 28 30 47.5 507 140 6 28 30 46.7 508 135 6 28 30 45.8 50


without changes. The minimal averagecurrent is increased to 52.6 A due tothe increase in the peak currentperiod. Although the average weldingcurrent is slightly increased, the elec-tromagnetic force is lower in compari-son to Experiments 1 and 2. Figure 9 shows a typical metaltransfer cycle in Experiment 3 whenthe peak current has been reduced to145 A. When the droplet reaches thedesired detaching point, a peakcurrent is applied without the laser.The 145 A of peak current is lowerthan the reference peak current for thewire used in this study. It can beobserved that although the peak cur-rent time has been extended 2 ms, thesurface tension still retains the dropletfirmly. The necking phenomenon thatis typically observed before thedetachment starts is not observed.This indicates that when the peak cur-rent is lower than the reference peakcurrent, simply extending the peakcurrent time by a small additional pe-riod may still not detach the dropletbecause the combined detaching forcemay still not be sufficient to overcomethe surface tension. Hence, to detachthe droplet, increasing the peakcurrent period beyond 6 ms may notbe effective and necessary. Figure 10 shows the metal transfercycle following the one in Fig. 9 whereno droplet is detached. When the peakcurrent is applied, the droplet hasgrown significantly greater than it wasbefore the peak current was applied inFig. 9. The gravitational force hasincreased such that the combined de-taching force exceeds the retainingforce. As can be seen, the droplet is de-tached during this second peak period(frames j–l). Hence, the undesirablemultiple pulses one-drop phenomenonis observed again. Repeatedexperiments under the same nominalparameters confirm this finding.

Experiment 4 — 145A Peak Currentfor 8 ms with a Laser

In Experiment 4, the nominal valuesfor all the welding parameters remainthe same as those in Experiment 3except for the application of the laseronto the droplet. As can be seen in Fig.11, when the peak current and laserpulses are applied, the droplet issuccessfully detached during the peakcurrent period. The droplet is

transferred with a smaller sizecompared to the size of the droplet thatwas transferred in Experiment 3. Figure 12 shows the metal transfercycle that follows the one in Fig. 11where a droplet has been successfullydetached. It can be observed thatwhen the peak current is applied withthe laser, the droplet is detached in themiddle of the peak current period. Acareful observation shows that thedroplet is greater than that in Fig. 11when the pulses are applied. An obser-vation of Fig. 11 shows that thedroplet detached there has actuallyformed approximately in Fig. 11h.Hence, the metal melted later duringthe peak current period accumulatedto the initial droplet in Fig. 12. The in-creased initial mass of the droplet re-duces the time to detach the droplet.From this point of view, 8 ms may beslightly greater than needed to detachthe droplet with the laser while the ef-fect of the laser is also verified. In thefollowing peak current reductionexperiments, the peak current periodwill be set at 6 ms, as can be seen inTable 2.

Peak Current Reduction Experiments A series of Experiments, #5–8, havebeen designed/performed using exactlythe same welding parameters except forthe peak current. They begin with Ex-periment 5 (the same as #2) using 150-A peak current, which has beenconducted earlier (Figs. 7 and 8) andproven to be successful for metal trans-fer control. When a peak current is con-firmed to be able to achieve the desiredone drop per pulse, the amperage forthe peak current is reduced by 5 A tosee if the peak current may be furtherreduced while still being able to detacheach droplet. Table 2 shows the param-eters for the successful detachment ex-periments. The parameters in each suc-cessful detachment experiment inTable 2 have been used to conductrepeated experiments in order to acceptthem as a success. The lowest peak current in Table 2is 135 A. Experiments have been con-ducted using 130-A peak current, butthe desired one drop per pulse was notalways achieved. Hence, 135 A is con-sidered as the lowest peak current forsuccessful detachment.

The experiment illustrated in Figs.5 and 6 has shown that 150 A is notsufficient to detach each droplet in 6ms without a laser. In the previous sec-tion, it was introduced that 160 A isthe lowest peak current possible forsuccessful detachment without a laser. Figures 13–15 give typical two con-secutive cycles for successfulExperiments 6–8. (The results fromExperiment 5 can be seen in Figs. 7and 8 and are not repeated here.) Eachof these figures has A and B with thecycle in B to follow the one in A. Ascan be seen, the droplet is detached ineach cycle during the peak current pe-riod by the synchronized peak currentand laser pulses. The size of thedroplet when it reaches the detachingposition may vary. However, thesuccess in the detachment is notaffected. The detachment is ensuredby the combined detaching forces fromthe electromagnetic force (due to thecurrent) and the recoil force (due tothe laser). The needed gravitationalforce, if any, can be provided by themass from the droplet formed duringthe present base and peak period with-out dependence on the accumulationfrom previous cycles.

Analysis and Discussion From the systematic and repeatedcomparative Experiments, #1–4,with/without a laser using exactly thesame welding parameters, one can eas-ily conclude that while the desired onedrop per pulse mode is achieved witheach laser pulse, it is never producedunder the same welding parameterswithout a laser. The success of thereal-time vision system in detectingthe droplet position and effectivenessof a pulsed laser spot as a replacementfor continuous laser stripe are system-atically verified. The proposed methodto modify the laser-enhanced GMAWto eliminate the waste of the laser en-ergy, reduce the effect of the uninten-tional laser application on theworkpiece, and achieve accuratecontrols on the droplet mass anddetaching time is verified. The result-ant laser-enhanced GMAW process be-comes controllable and morerepeatable. As can be seen from Experiments5–8, pulsed laser-enhanced GMAWsuccessfully reduces the minimal peak


WELDING RESEARCH


current for the desired one drop perpulse mode from 160 to 135 A. Thisreduction is achieved using a 50-Wpeak power laser spot. This reductionis similar to that achieved in the previ-ous effort by Huang (Refs. 22–25) forthe first laser-enhanced GMAW withthe same wire (material and diameter).The laser used in his effort was a con-tinuous 1-kW laser, but the effectivelaser power was approximately 850 Wand the length of the laser stripe was14 mm. For the droplet detached, thelength of the droplet along which thelaser stripe is intercepted may be con-sidered approximately 1.0 mm. The ef-fective power projected on the dropletis approximately 60 W, similar to thatused in this research. The presentstudy verifies that the pulsed laserwith 6-ms pulse duration has approxi-mately the same effect on the detach-ment enhancement. Continuous appli-cation of the laser is not necessary.Furthermore, the continuous applica-tion produces unintentional adverseeffects including overheating thedroplet, generating extra fume, andwasting energy. Previous efforts by Huang (Refs.22–25) have shown that the laser en-hances the detachment by its recoilforce when the laser vaporizes part ofthe droplet metal. The arc also vapor-izes the droplet metal due to itsextremely high temperature at theanode. This vaporization increaseswith the anode power, which increaseslinearly with the current. However, therecoil force due to arc-caused evapora-tion is against the detachment. Onemay expect that the increase in theevaporation due to the laserapplication should be smaller than thereduction in the arc caused by evapo-ration due to the reduction in the cur-rent. As can be seen, a 50-W laserreduces the peak current by 25 A. Forthe anode of a steel wire, its voltage is12 V (Ref. 1). The reduction in theanode power is 300 W. The net vapors,thus, should be reduced. The recoil force of the laserincreases more than linearly with thelaser power intensity intercepted bythe droplet (Refs. 22–25). Hence,increasing the peak laser power is ex-pected to reduce the peak currentneeded to detach the droplet. With amuch increased peak laser power, it isexpected that the droplet may bedetached at any (reasonable) size at

any (reasonable) arc variables (weldingcurrent). From this point of view, thisresearch for pulsed laser-enhancedGMAW using a pulsed laser spotestablished the foundation to lead tothe ideal laser-enhanced GMAW.

Conclusions and FutureWork With the developed system thatuses a real-time visual feedback toapply a laser spot of 50-W peak powerin synchronization with a currentpulse, two series of systematically de-signed experiments have beenconducted repeatedly. Experimentalresults and analysis suggest thefollowing: • The laser does not need to be con-tinuously applied onto the droplet inorder to enhance the metal transfercontrol. • For the experimental conditionsin this study, 6 ms appears to be an ad-equate duration for the laser pulse. • The laser does not need to be ap-plied onto the position other than thedroplet to enhance the metal transfercontrol. The laser stripe should be re-placed with a laser spot as long as thelaser spot can be projected to thedroplet when the droplet needs to bedetached. • The minimal peak current neededto achieve the desired one drop perpulse metal transfer mode can bereduced from 160 to 135 A, i.e., 16%reduction, for the 0.8-mm-diametersteel wire of and the 50-W peak powerlaser used. The reduction for the peak currentfrom 160 to 135 A corresponds to thelaser power, i.e., 50-W, used in thisstudy. Based on the positive definiteverification for the effectiveness ofpulsed laser and real-time image pro-cessing-based control, future workmay use a higher peak power laser tostudy how the peak current reduces asthe peak laser power increases.

This work is funded by the NationalScience Foundation under grantCMMI-0825956 titled “Control ofMetal Transfer at Given Arc Variables.”The assistance and help from

colleagues at the Welding ResearchLaboratory at the University ofKentucky, including Dr. ZhenzhouWang, WeiJie Zhang, and Jun Xiao,are greatly appreciated.

1. O’Brien, R. L. 1991. Welding Handbook.Vol. 2: Welding Processes. 8th edition. Miami,Fla.: American Welding Society. 2. Pan, J. 2005. Arc Welding Control.Cambridge, United Kingdom: WoodheadPublishing Limited. 3. Zhang, Y. M., Liguo, E., and Kovacevic, R.1998. Active metal transfer control by monitor-ing excited droplet oscillation. Welding Journal77(9): 388-s to 395-s. 4. Zhang, Y. M., and Li, P. J. 2001. Modifiedactive control of metal transfer and pulsedGMAW of titanium. Welding Journal 80(2): 54-sto 61-s. 5. Li, K. H., and Zhang, Y. M. 2007. Metaltransfer in double-electrode gas metal arc weld-ing. Journal of Manufacturing Science andEngineering-Transactions of the ASME 129(6):991–999. 6. Thomsen, J. S. 2006. Control of pulsed gasmetal arc welding. International Journal of Model-ling, Identification, and Control 1(2): 115–125. 7. Parks, J. M., and Stava, E. K. 1991. Appa-ratus and method of short circuiting arc weld-ing. U.S. Patent 5,003,154. 8. Stava, E. K. 1993. A new, low-spatter arcwelding machine. Welding Journal 72(1): 25–29. 9. Parks, J. M., and Stava, E. K. 1991. Appa-ratus and method of short circuiting arc weld-ing. U.S. Patent 5,003,154. 10. Stava, E. K. 1992. System and method ofshort circuiting arc welding. U.S. Patent5,148,001. 11. Stava, E. K. 2006. Short circuit arcwelder and method of controlling same. U.S.Patent 7,109,439. 12. Zhang, H. T., and Feng, J. C. 2009. Thearc characteristics and metal transfer behaviorof cold metal transfer and its use in joining alu-minum to zinc-coated steel. Materials Scienceand Engineering A 499: 111–113. 13. Pickin, C. G., and Young, K. 2006. Evalu-ation of cold metal transfer (CMT) process forwelding aluminum alloy. Science and Technologyof Welding and Joining 11(5): 583–585. 14. Praveen, P., Yarlagadda, P. K. D. V., andKangb, M. J. 2005. Advancements in pulse gasmetal arc welding. Journal of Materials ProcessingTechnology 164–165(May): 1113–1119. 15. Ono, M., Shinbo, Y., Yoshitake, andOhmura, A. M. 2002. Development of laser-archybrid-welding. NKK Techn. 86: 8–12. 16. Kim, Y. S., and Eagar, T. W. 1993. Analy-sis of metal transfer in gas metal arc-welding.Welding Journal 72(6): 269-s to 278-s. 17. Kim, Y. S. 1989. Metal transfer in gasmetal arc welding. PhD dissertation, Massachu-setts Institute of Technology. 18. Kim, Y. S., and Eagar, T. W. 1993. Metaltransfer in pulsed current gas metal arc-welding.Welding Journal 72(7): 279s to 287s. 19. DeRuntz, B. D. 2003. Assessing the bene-fits of surface tension transfer welding to indus-


WELDING RESEARCH

Acknowledgments

References


try. Journal of Industrial Technology 19(4): 2–8. 20. Wu, Y., and Kovacevic, R. 2002.Mechanically assisted droplet transfer processin gas metal arc welding. Proc. Instn. Mech.Engrs. Part B: Journal of Engineering Manufac-ture 216: 555–564. 21. Yang, S. Y. 1998. Projected droplet trans-fer control with additional mechanical forces(AMF) in MIG/MAG welding process. PhD dis-sertation, Harbin Institute of Technology. 22. Huang, Y., and Zhang, Y3. M. 2010.Laser-enhanced GMAW. Welding Journal 89(9):181-s to 188-s. 23. Huang, Y., and Zhang, Y. M. 2011. Laser-enhanced metal transfer: Part I: System and ob-servations. Welding Journal 90(10): 183-s to190-s. 24. Huang, Y., and Zhang, Y. M. 2011. Laser-enhanced metal transfer: Part II: Analysis andinfluence factors. Welding Journal 90(11): 205-sto 210-s. 25. Huang, Y., Shao, Y., and Zhang, Y. M.2012. Nonlinear modeling of dynamic metaltransfer in laser enhanced GMAW. Welding Jour-nal 91(5): 140-s to 148-s. 26. Kim, J., Na, W., and S. J. 1991. A studyon an arc sensor for gas metal arc welding ofhorizontal fillets. Welding Journal 70(8): 216-sto 221-s. 27. Rokhlin, S. I., and Guu, A. C. 1990. Com-puterized radiographic sensing and control of an

arc welding process. Welding Journal 69: 83-s to97-s. 28. Nagarajan, S., Banerjee, P., Chen, W., andChin, B. A. 1992. Control of the welding processusing infrared sensors. IEEE Transactions on Ro-botics and Automation 8(1): 86–93. 29. Hardt, D. E., and Katz, J. M. 1984. Ultra-sonic measurement of weld penetration. WeldingJournal 63(9): 273-s to 281-s. 30. Siores, E. 1988. Development of a real-time ultrasonic sensing system for automatedand robotic welding. Brunel University. PhD dis-sertation. 31. Richardson, R. W. 1984. Coaxial arc weldpool viewing for process monitoring andcontrol. Welding Journal 63(3): 43–50. 32. Pietrzak, K. A., and Packer, M. 1994. Vi-sion-based weld pool width control. Journal ofEngineering for Industry – Transaction of theASME 116:1: 86–92. 33. Kovacevic, R., Zhang, Y. M. et al. 1995.Sensing and control of weld pool geometry forautomated GTA welding. Journal of Engineeringfor Industry – Transaction of the ASME 117(2):210–222. 34. Kovacevic, R., and Zhang, Y. M. 1996.Sensing free surface of arc weld pool using spec-ular reflection principle and analysis.Proceedings of the Institution of Mechanical Engi-neers, Part B, Journal of Engineering Manufactur-ing 210(6): 553–564.

35. Steele, J. P., Minch, H. C. et al. 2005. De-velopment of closed-loop control of roboticwelding processes. Industrial Robot: An Interna-tional Journal 32(4): 350–355. 36. Mnich, C. 2004. Development of a syn-chronized, high-speed, stereovision system forin situ weld pool measurement. MS thesis, Engi-neering Division, Colorado School of Mines. 37. Wu, C. S., Gao, J. Q. et al. 2003. Vision-based measurement of weld pool geometry inconstant-current gas tungsten arc welding. Proc.Instn Mech. Engrs Part B: J. Engineering Manufac-ture 217: 879–882. 38. Ma, X. J., and Zhang, Y. M. 2009. Reflec-tion of illumination laser from gas metal arcweld pool surface. Meas. Sci. Technol. 20(11):115105. 39. Wang, Z. Z., and Zhang, Y. M. 2009.Brightness based selection and edge detectionbased enhancement separation algorithm forlow resolution metal transfer images. IEEETransactions on Automation Science and Engineer-ing 6(1): 181–187. 40. Shao, Y., and Zhang, Y. M. 2011. Moni-toring of liquid droplets in laser-enhancedGMAW. International Journal of Advanced Manu-facturing Technology 57: 203–214. 41. Shao, Y. 2012. Vision based real-timemonitoring and control of metal transfer inlaser enhanced gas metal arc welding. PhD dis-sertation, University of Kentucky.


WELDING RESEARCH

Peer review of research papers is now managedthrough an online system using Editorial Managersoftware. Papers can be submitted into the systemdirectly from the Welding Journal page on theAWS website (www.aws.org) by clicking on “sub-mit papers.” You can also access the new site di-rectly at www.editorialmanager.com/wj/. Followthe instructions to register or log in. This onlinesystem streamlines the review process, and makesit easier to submit papers and track their progress.

By publishing in the Welding Journal, more than69,000 members will receive the results of yourresearch.

Additionally, your full paper is posted on theAmerican Welding Society Web site for FREE ac-cess around the globe. There are no page charges,and articles are published in full color. By far, themost people, at the least cost, will recognize yourresearch when you publish in the world-respectedWelding Journal.

Authors: Submit Research Papers Online


WJ_2014_06_s205

Documents

Transcript of WJ_2014_06_s205