A Control System for Keyhole Plasma Arc Welding of Stainless Steel Plates with Medium Thickness

Wu 11-10 layout:Layout 1 10/11/10 8:45 AM Page 225

SUPPLEMENT TO THE WELDING JOURNAL, NOVEMBER 2010 Sponsored by the American Welding Society and the Welding Research Council A Control System for Keyhole Plasma Arc Welding of Stainless Steel Plates with Medium Thickness

The control system uses a specially designed current wavefrom with two slow dropping substages of variable slopes to adjust the keyhole size

BY C. S. WU, C. B. JIA, AND M. A. CHEN

way, while the establishment of the keyhole ABSTRACT ensures the desired complete penetration, Keyhole plasma arc welding (PAW) has the potential to achieve deep, narrow pen- the base current allows the melted metal to etration with low cost and high tolerance of joint preparation. However, the narrow solidify and the keyhole to close so that parameter window associated with the keyhole condition limits its wide application, es- melt-through is prevented. After a specified pecially for welding thicker plates. To reach its full potential, a control system for key- period of base current, the peak current is hole PAW is developed. The efflux plasma voltage signal is detected in real-time to applied again to reestablish the keyhole, be- characterize the keyhole size and dimension. The welding current waveform with two ginning a new pulse cycle. As a result, the slow dropping substages and variable slopes is designed to improve the controllabil- process is not maintained in the keyhole ity of keyhole dynamics via reducing both heat input and arc force when a keyhole mode as it is classically defined, but in a transforms from its opening status to the closure status. The keyhole PAW experi- mode of “one keyhole per pulse,” and the ments are conducted to examine the control effectiveness of the system. The devel- square waveform of the welding current is oped system operates in the mode of “one keyhole per pulse” reliably. Closed-loop usually used (Refs. 11, 12). However, this control tests of varied-thickness plates show that even if the plate thickness changes square waveform is associated with quick as much as 50% (from 8 to 4 mm), the developed system adjusts the peak current and step change of welding current from the its dropping slopes automatically to keep both weld width and penetration consistent. peak level to the base level, which causes the keyhole condition to vary in a non- Introduction open and stable to ensure weld quality. To smooth way, i.e., a rapid closure of the key- this end, the welding process parameters, hole occurs. Furthermore, such a current waveform is just suitable in welding of thin The keyhole mode of welding is the pri- especially the welding current, have to be sheets (Ref. 13). To avoid this problem, mary attribute of high-power-density weld- within a narrow range. If the welding cur- Zhang and Liu proposed another kind of

ing processes (plasma arc welding, laser rent is a little bit lower, the keyhole may be WELDING RESEARCH current waveform with two declining sub- beam welding, and electron beam welding) closed, but if the welding current is a little stages when the welding current changes that can penetrate thicker workpieces with bit higher, the keyhole expands too large from the peak level to the base level (Ref. a single pass and produce welds with a high and there is a tendency for the molten 14), but the first substage is with a variable aspect ratio. Compared to laser beam weld- metal to be blown away from the weld pool, slope while the second one is with a fixed ing and electron beam welding, keyhole causing melt-through or cutting instead of slope, and the welded workpiece thickness plasma arc welding (PAW) is more cost ef- joining. To solve this problem, Zhang et al. is still limited. Thus, it is necessary to in- fective and more tolerant of joint prepara- proposed a novel approach to operate key- vestigate further for improving the flexibil- tion (Refs. 1, 2). Thus, keyhole PAW has hole PAW through employing pulsed cur- ity of adjusting the keyhole condition for found wide application in industry (Refs. rent and to switch the current from the high thicker plates. 3–8). However, the keyhole dynamics, i.e., peak current to the low base current after Another critical issue for realizing au- the establishment, sustainment, and closure the keyhole is established (Ref. 11). In this tomatic control of keyhole PAW with a of the keyhole during the PAW process, is a wider parameter window is how to moni- critical issue in applying PAW (Ref. 9). In tor and sense the keyhole dynamics. Once keyhole PAW, the quality of the weld de- the feedback signals of the keyhole con- pends on the keyhole stability, which itself KEYWORDS dition are sensed, the welding process depends on a large number of factors, es- parameters can be adjusted synchronisti- pecially the physical characteristics of the Plasma Arc Welding cally to control the keyhole’s stability. material to be welded and the welding pa- Keyhole Various methods have been used to mon- rameters to be used (Ref. 10). In normal Control System itor the keyhole condition, such as efflux keyhole PAW, the keyhole is maintained Pulse Current Waveform plasma voltage, optical sensors, sound Efflux Plasma Voltage signal sensors, and plasma cloud charge Stainless Steel Plates C. S. WU ([email protected]), C. B. JIA, and M. sensors (Refs. 15–22). Each sensing Medium Thickness A. CHEN are with the Institute of Materials Join- method has some limitations. For exam- ing, Shandong University, Jinan, P.R. China. ple, optical sensors need a long time to

WELDING JOURNAL 225-s Wu 11-10 layout:Layout 1 10/11/10 8:46 AM Page 226

Fig. 1 — The experimental setup. Fig. 2 — The controlled pulse current waveform and keyhole signal. WELDING RESEARCH

Fig. 3 — Schematic of the closed-loop control system. Fig. 4 — Relative definition of the pulse current waveform.

transfer and process video signals and Experimental System lish an electrical potential between the cost a lot, and sound signal sensors are workpiece and the measuring bar due to not accurate enough. The efflux plasma Figure 1 shows the developed control the phenomenon of plasma space charge charge sensor measures the electrical po- system for the keyhole PAW process. It (Refs. 11, 24). If the keyhole is not estab- tential resulting from the plasma efflux consists of the computer, PAW machine, lished, there will be no efflux plasma be- on the backside of the workpiece when data-acquisition unit, welding current sen- tween the workpiece and the measuring the keyhole is established (Refs. 20–22). sor, and efflux plasma voltage sensor. The bar, and thus, no electrical potential exists. Although such a sensing approach has computer is the central unit of the system. A simple sensor consisting of a resistor some disadvantages, i.e., the detection On one hand, it can adjust welding current and a capacitor is used to detect the elec- plate has to be mounted on the backside in real time and output any current wave- trical potential or the voltage between the of the workpiece, its cost is low, its struc- forms as user defined (Ref. 23). On the workpiece and the measuring bar. The ture is simple, and its reliability is high. other hand, it samples the transient signals larger the keyhole size, the higher the in- In this study, a control system for the of welding current and efflux plasma volt- tensity of the efflux plasma, thus, the keyhole PAW process is developed to age, and adjusts the output current wave- larger the efflux plasma voltage. There is a widen its parameter window. The efflux form parameters automatically to respond correlation of the keyhole diameter at the plasma voltage signal is used to character- to any changes in the keyhole signal. backside with the measured efflux plasma ize the keyhole size and weld dimension. As shown in Fig.1, the keyhole sensor is voltage signal (Ref. 25). The peak current and its duration, as well developed through measuring the efflux as its dropping slopes, are employed as the plasma voltage. The measuring bar is a Control Strategy controlling variables. The ideal one key- piece of mild steel sheet mounted under- hole per pulse condition is attempted for neath the workpiece. It is kept insulated Figure 2 shows the specially designed stainless steel plates with medium thick- electrically. The distance between the welding current waveform and keyhole ness. The flexibility of adjusting the key- workpiece and the measuring bar is fixed signal. As the welding current drops from hole condition is expected to improve at about 6 mm. If the keyhole is estab- the peak level to the base level, two sub- further. lished, the plasma jet must exit through stages of decreasing current with different the keyhole. The efflux plasma will estab- slopes of K1 and K2 are added. At instant

226-s NOVEMBER 2010, VOL. 89 Wu 11-10 layout:Layout 1 10/11/10 8:46 AM Page 227

t1, the peak current IP is applied, and the system keeps detecting the signal from the efflux sensor. The signal of the efflux plasma voltage is around zero before the keyhole is established. At instant t2, the keyhole is established and the efflux plasma voltage exceeds a certain value. To avoid the keyhole size from expanding too much, the current starts to decrease with a dropping slope K1. Because of thermal in- ertia, the remaining keyhole volume expands slowly even though both heat input and arc force associated with the welding current start to drop during this stage. At instant t3, when the keyhole size reaches the preset value to meet the desired prac- tical requirements of weld quality, the current decreases at a steeper slope K2 ⎜ ⎜>⎜ ⎜ ( K2 K1 ), so that the keyhole stops expanding but starts to close. At instant t4, the keyhole is completely closed, and the efflux plasma voltage is zero. The current is switched to base level IB at instant t5. After IB is applied for a preselected period TB = t6 – t5, the current is switched to the peak level again to begin a new cycle. In this way, it can ensure keyhole establishment and complete penetration but avoid melt-through defects. To characterize the keyhole size and weld bead width at the backside of the workpiece, the average value (VEP) of efflux plasma voltage (VE) from t2 to t3 in each cycle is used as the feedback signal because it has been validated that the measured VEP is well correlated with the weld bead width at the backside of the workpiece (Ref. 25). Take the average value of efflux plasma voltage (VEP) as the Fig. 5 — The flowchart of the control process. controlled variable. Take the pulse current value (IP) and its two dropping slopes (K1 and K2) as the controlling variables. By ad- A justing IP, K1, and K2, the VEP is kept within the preselected range so that the weld penetration and weld bead width are WELDING RESEARCH controlled. Figure 3 shows the schematic diagram of the closed-loop control system. The PI (proportional and integral) controller is used to adjust IP, K1, and K2 ac- cording to the difference between- the B measured VEP and setpoint VEP. The ex- pression of discrete PI control algorithm may be written as

k ()= ()+ () uk KekP K1 ∑ ei − Fig. 6 — The weld appearance of 8-mm-thick workpiece. A — Top face; B — back face. i 0 (1)

where u(k) is the- output of PI controller, e(k) = VEP(k)–VEP is the error (input of PI controller), K and K are coefficients, P 1 I = I + 30% × (I –I ) = 0.3I +0.7I K = –(I –I )/T = –0.4(I –I )/T and k is the cycle number. K2 B P B P B 2 K1 K2 2 P B 2 (3) (5) As shown in Fig. 4, the base current IB is fixed, while IK1 and IK2 are correlated to Referring to Fig. 4, the dropping current where T1 and T2 are defined in Fig. 2. the peak level IP in the following way, slopes K1 and K2 may be written as Therefore, just through adjusting the peak level , the dropping current slopes K × 1 IK1 = IB + 70% (IP –IB) = 0.7IP +0.3IB K1 = –(IP –IK1)/T1 = –0.3(IP –IB)/T1 and K2 for two substages can be changed (2) (4) correspondingly, so that the waveform pa-

WELDING JOURNAL 227-s Wu 11-10 layout:Layout 1 10/11/10 8:47 AM Page 228

A B

Fig. 7 — A — The measured welding current; B — efflux plasma voltage.

A B WELDING RESEARCH

Fig. 8 — A — Part of measured welding current; B — efflux plasma voltage.

Fig. 9 — The measured VEP and Ip for 8-mm-thick plate. Fig. 10 — The test plate with varied thickness.

rameters can be set more easily but the thickness of the workpiece was 8 mm, and trolling parameter, the pulse current IP characteristics of the controlled pulse cur- its dimensions were 200 mm in length and was adjusted in-process, and other param- rent remain. Figure 5 is the control 80 mm in width. Bead-on-plate PAW was eters such as IK1, IK2, K1, and K2 were de- process flowchart of the system. carried out. The test conditions were as termined by Equations 2–5. The setpoint follows: the distance from the nozzle to for the desired value of the efflux plasma Control Experiment with 8-mm-Thick the workpiece surface was 6 mm; the voltage VEP was preset as 1000 mV. Workpiece shielding gas and the plasma gas were Figure 6 shows the weld appearance at argon with respective flow rates of 20 both top and back sides. The weld bead The developed control system was used L/min and 3 L/min; the welding speed was width at both sides looks consistent. Dur- to conduct closed-loop control experi- 120 mm/min. Other process parameters ing the welding process, the experimental ments. The experiments used stainless were set as follows: IB = 60 A, TB = 50 system captured the transient signals of steel (Type 304) for the workpiece. The ms, T1 = 50 ms, T2 = 50 ms. As the con- welding current and efflux plasma voltage,

228-s NOVEMBER 2010, VOL. 89 Wu 11-10 layout:Layout 1 10/11/10 8:47 AM Page 229

which are shown in Fig. 7. VE is the transient value of measured efflux plasma volt- A age, while VEP is the average value of VE in each cycle. To demonstrate the response of the pulse current with respect to the efflux plasma voltage more clearly, the signals during the period from 53 to 60.5 s in Fig. B 7 are magnified in the time scale. As shown in Fig. 8, one keyhole was established for every pulse. For each cycle, the Fig. 11 — The weld appearance of varied-thickness workpiece. A — Topside; B — backside. keyhole formed and closed as the peak current changed. Because of thermal iner- tia, there was a delay between the current change and the keyhole signal. But the mode of one keyhole per pulse was A achieved reliably. Figure 8A also illustrates the computer output (preset) and the measured current waveforms. It is validated that both are in good agreement, and the pulse current has a quick response to the efflux plasma voltage signal. As shown in Fig. 8, although the workpiece thickness was not varied but kept at 8 mm, there may have been other distur- bances during the welding process, so that the peak current IP of the five cycles were 213, 213, 218, 208, and 206 A, respectively. Due to the control action, the peak current, its duration and dropping slopes at two substages are not the same in each cycle, but are adjusted automatically based on the measured signal of efflux plasma B voltage. Figure 9 is the measured average values of the efflux plasma voltage and the peak current IP. It seems that the average value of the measured efflux plasma voltage in each cycle always fluctuates around the setpoint VEP within an acceptable limit. The pulse current makes a relevant adjustment based on the measured VEP. It can be seen that the pulse current IP varies with the signal of efflux plasma voltage in WELDING RESEARCH real-time so that the value of VEP in each cycle is within -a narrow range around the setpoint VEP. The response speed and ac- curacy are satisfactory. It shows marked control effectiveness in welding of a 8-mm Fig. 12 — Varied thickness test. A — The measured current; B — efflux plasma voltage. thick plate.

Control Test on Plate with Varied Thickness-

To test the control effectiveness of the plasma voltage VEP was preset as 1000 mV. with the change in plate thickness. developed system further, a test plate with Because the plate thickness was gradu- Figure 12 shows the measured tran- varied thickness was used, as shown in Fig. ally varied from 8 to 4 mm, the welding sient welding current and efflux plasma 10. Bead-on-plate welding was conducted process could not maintain consistency if voltage signals in the control test of a plate on such a workpiece from left (thickness 8 no appropriate control action was applied. with varied thickness. Figure 13 illustrates mm) to right (thickness 4 mm). The Using the developed system, the peak cur- the control action for the varied plate process parameters were set as follows: IB rent, along with -its duration and dropping thickness. During the welding process, as = 60 A, TB = 60 ms, T1 = 60 ms, and T2 = slopes, made a suitable adjustment based the plate thickness continuouly decreased, 60 ms. Other parameters were selected as on the measured efflux plasma voltage sig- the welding current waveform parameters aforementioned. The pulse current IP was nal as the thickness varies along the weld- made a suitable response, especially the adjusted in-process, and current waveform ing direction. Figure 11 shows the weld peak current was lowered, so that the key- parameters such as IK1, IK2, K1, and K2 appearance at both topside and backside. hole size was kept almost constant. The were calculated by Equations 2–5. The set- It appears the weld formation and pene- measured average value (VEP ) of efflux point for the desired value of the efflux tration were controlled very well, even plasma voltage fluctuates around the set-

WELDING JOURNAL 229-s Wu 11-10 layout:Layout 1 10/11/10 8:48 AM Page 230

Fig. 13 — The controlled variable VEP and controlling variable IP for varied Fig. 14 — Three segments of the weld on varied-thickness workpiece. thickness test.

A B WELDING RESEARCH

Fig. 15 — Segment 1 for time interval 30–35.5 s. A — The measured welding current; B — efflux plasma voltage.

point VEP within the allowable range. 153, 154, 145, 144, and 143 A, respectively. 3) The experimental results showed To observe the response of welding cur- Therefore, the experimental results show that the sampled efflux plasma voltage sig- rent to the efflux plasma voltage more that even if the plate thickness changes as nal responded to the variation in the con- clearly, three time intervals of the tran- much as 50% (from 8 to 4 mm), the devel- trolled pulse current waveform very well. sient signals in Fig. 12 are magnified in oped system is capable of adjusting the The developed system operates in the time-scale, i.e., the signals within three peak current and its dropping slopes auto- mode of “one keyhole per pulse” reliably time intervals- 30~35.5 s, 38.5~44 s, and matically so that the keyhole size stays al- for welding of thick materials. 68~73 s. These three time intervals corre- most constant. Therefore, both weld width 4) The varied thickness plates were spond to the segments on the workpiece, and penetration are consistent, and con- used to conduct the control experiments. as shown in Fig. 14. The magnified results trol action responds reliably and smoothly. The test results showed that even when the for three segments are shown in Figs. plate thickness changed as much as 50% 15–17, respectively. For segment 1, the Conclusions (from 8 to 4 mm), the developed system plate thickness was not varied, but kept at was capable of adjusting the peak current 8 mm. The peak current was constant (220 1) A control system for keyhole plasma and its dropping slopes automatically so A), and the measured efflux plasma volt- arc welding was developed for application that the keyhole size keeps almost con- age agrees with the setpoint. Segment 2 lo- on thin sheet to thick plate. The efflux stant, and both weld width and penetra- cated at the middle part of the workpiece, plasma voltage was measured in real-time tion are consistent. and the thickness changed continuously. to characterize the keyhole size and weld Thus, the peak current was gradually de- dimension. The peak current and its dura- Acknowledgment creased from 220 to 200 A to keep the key- tion, as well as its dropping slopes, are em- hole size nearly constant. The six pulses in ployed as the controlling variables. The authors are grateful for the finan- Fig. 16A are with values of peak current 2) The specially designed pulse current cial support for this research from the Na- 220, 220, 215, 210, 205, and 200 A, respec- waveform, including two slow dropping tional Natural Science Foundation of China tively. For segment 3, the workpiece thick- substages with both variable slopes, is used (Key Program Grant No. 50936003). ness became thinner. Thus, the peak to transform the welding current from its current was lowered further. It changed peak level to base level. The keyhole con- References gradually from 153 to 143 A to adapt to dition can smoothly transform from the the thinner thickness. The five pulses in opening status to the closure status under 1. Lancaster, J. F. 1984. The Physics of Weld- Fig. 17A are with values of peak current the action of such current waveform. ing. pp. 268–280, Oxford, Perganon Press.

230-s NOVEMBER 2010, VOL. 89 Wu 11-10 layout:Layout 1 10/11/10 8:49 AM Page 231

A B

Fig. 16 — Segment 2 for time interval 38.5–44 s. A — The measured welding current; B — efflux plasma voltage.

A B

Fig. 17 — Segment 3 for time interval 68–73 s. A — The measured welding current; B — efflux plasma voltage.

2. Fortain, J. M. 2008. Plasma welding evo- 1401–1411. 19. Dong, C. L., Zhu, Y. F., Zhang, H., Shao, lution and challenges. IIW Doc.XII-1948-08, 12. Lu, W., Zhang, Y. M., and Lin, W. Y. Y. C., and Wu, L. 2001. Study on front side arc 2004. Nonlinear interval model control of quasi- light sensing in keyhole plasma arc welding. pp. 1–11. WELDING RESEARCH 3. Martikainen, J. K., and Moisio, T. J. I. keyhole arc welding process. Automatica 40: Chinese Journal of Mechanical Engineering 1993. Investigation of the effect of welding pa- 805–813. 37(3): 30–33. rameters on weld quality of plasma arc keyhole 13. Dong, C. L., Wu, L., and Shao, Y. C. 20. Steffens, H. D., and Kayser, H. 1972. Au- welding of structural steels. Welding Journal 2000. The history and state-of-the-art of key- tomatic control for plasma arc welding with 72(7): 329-s to 340-s. hole mode plasma arc welding. China Mechan- constant keyhole diameter. Welding Journal 4. Vilkas, E. P. 1991. Plasma arc welding of ical Engineering 11(5): 577–581. 51(06): 408–418. exhaust pipe system components. Welding Jour- 14. Zhang, Y. M., and Liu, Y. C. 2007. Con- 21. Zhang, S. B., and Zhang, Y. M. 2001. Ef- nal 70(4): 49–52. trol of dynamic keyhole welding process. Auto- flux plasma charge-based sensing and control of 5. Irving, B. 1997. Why aren’t airplanes matica 43: 876–884. joint penetration during keyhole plasma arc welded? Welding Journal 76(1): 34–41. 15. Zhang, Y. M., Zhang, S. B., and Liu, Y. welding. Welding Journal 80(7): 157-s to 162-s. 6. Irving, B. 1992. Plasma arc welding takes C. 2001. A plasma cloud charge sensor for pulse 22. Dong, C. L., Wu, L., Zhu, Y. F., Zhang, on the advanced solid rocket motor. Welding keyhole process control. Measurement Science H., and Shao, Y. C. 1995. Detection of efflux Journal 71(12): 49–52. and Technology 12(8): 1365–1370. plasma voltage in keyhole plasma arc welding. 7. Nunes, A. C. 1984. Variable polarity 16. Wang, Y. W., Chen, Q., Sun, Z. G., Sun, Aeronautical Manufacturing Technology (5): plasma arc welding on the space shuttle external J. W., and Wang, H. Y. 2001. Sound sensing of 60–62. tank. Welding Journal 63(9): 27–35. the keyhole behaviors in plasma arc welding. 23. Jia, C. B., Wu, C. S., and Zhang, Y. M. 8. Keanini, R. G., and Rubinsky, B. 1990. Chinese Journal of Mechanical Engineering 2007. A computer-based control system for key- Plasma arc welding under normal and zero 37(1): 53–56. hole plasma arc welding. China Welding 16(1): gravity. Welding Journal 69(6): 41–50. 17. Chen, Q., Sun, Z. G., Sun, J. W., and 7–10. 9. Zhang, Y. M., and Zhang, S. B. 1999. Ob- Wang, Y. W. 2004. Closed-loop control of weld- 24. Li, L., Brookfield, D. J., and Steen, W. servation of the keyhole during plasma arc ing penetration in keyhole plasma arc welding. M. 1996. Plasma charge sensor for in-process welding. Welding Journal 75(2): 53-s to 59-s. Transactions of Nonferrous Metals Society of non-contact monitoring of laser welding 10. Metcalfe, J. C., and Quigley, M. B. C. China 14(1): 116–120. process. Measurement Science and Technology 7: 1975. Keyhole stability in plasma arc welding. 18. Zheng, B., Wang, H. J., Wang, Q. I., and 615–626. Welding Journal 54(11): 401-s to 404-s. Kovacevic, R. 2000. Control of weld penetra- 25. Jia, C. B., Wu, C. S., and Zhang, Y. M. 11. Zhang, Y. M., and Liu, Y. C. 2003. Mod- tion in VPPAW of aluminum alloys using the 2009. Sensing controlled pulse key holing condi- eling and control of quasi-keyhole arc welding front weld pool image signal. Welding Journal tion in plasma arc welding. Transactions of Non- process. Control Engineering Practice 11: 79(12): 363-s to 371-s. ferrous Metals Society of China 19(2): 341–346.

WELDING JOURNAL 231-s