Weld penetration sensing in pulsed gas tungsten arc welding based on arc voltage

Weld penetration sensing in pulsed gas tungsten arc welding based on arc voltage

Journal of Materials Processing Technology 229 (2016) 520–527 Contents lists available at ScienceDirect Journal of Materials Processing Technology j...

2MB Sizes 2 Downloads 122 Views

Journal of Materials Processing Technology 229 (2016) 520–527

Contents lists available at ScienceDirect

Journal of Materials Processing Technology journal homepage: www.elsevier.com/locate/jmatprotec

Weld penetration sensing in pulsed gas tungsten arc welding based on arc voltage Zhang Shiqi, Hu Shengsun, Wang Zhijiang ∗ Tianjin Key Laboratory of Advanced Joining Technology, Tianjin University, Tianjin 300072, China

a r t i c l e

i n f o

Article history: Received 18 April 2015 Received in revised form 16 September 2015 Accepted 18 September 2015 Available online 25 September 2015 Keywords: Arc voltage Penetration sensing Medium-thickness steel plate Pulsed gas tungsten arc welding

a b s t r a c t An alternative to the conventional weld penetration sensing methods in pulsed gas tungsten arc welding is proposed for implementation at manufacturing sites. The fluctuation amplitude of arc voltage in the peak duration (Uk ) reflects the weld penetration status in a single-weld spot; a fully penetrated weld spot is obtained in the step-by-step welding process when the abrupt change of Uk (denoted as U’k ) is above the threshold (1.50 V herein). A full penetration control system that employs U’k as the feature signal for step-by-step pulsed gas tungsten arc welding can be easily installed at the manufacturing sites. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Medium-thickness steel plates, namely steel plates with thickness ranging between 4.5 mm and 25 mm, are usually used in structure construction, machinery manufacturing, container manufacturing, and so on. However, it is difficult to guarantee the welding quality for medium-thickness steel plates, especially the root pass. The penetration of the root pass, and therefore the final weld quality (e.g., mechanical properties and life time of the joint) (Fan et al., 2003), is affected by the good cooling condition of medium-thickness steel plates in the welding process and by the unfavorable welding conditions, such as misalignment of workpieces and uneven gaps between these plates. One of the effective methods frequently used for root pass welding is gas tungsten arc welding (GTAW). Because of the low heat input of GTAW, an automatic full penetration control at the manufacturing sites is important. The first step in the automatic weld penetration control is to find a feature signal that can be obtained and processed in real time to characterize the weld penetration status. Several techniques such as temperature measurement, imaging, acoustics, and X-ray have been proposed to determine this condition. Fan et al. (2003) studied the changes in temperature distribution surrounding the weld pool using an infrared sensor to detect different workpiece

∗ Corresponding author. E-mail address: [email protected] (Z. Wang). http://dx.doi.org/10.1016/j.jmatprotec.2015.09.034 0924-0136/© 2015 Elsevier B.V. All rights reserved.

inclination angles that have a significant influence on the weld penetration depth. Vasudevan et al. (2011) developed a computercontrolled GTAW machine that enables sensing of weld pool using an infrared camera mounted on a torch assembly. They revealed that an inverse linear relationship exists between the macroscopic temperature gradient, computed from the infrared thermal profile, and the measured penetration depth for the weld bead. Unfortunately, the results of infrared thermography are influenced by radiations from the arc and tungsten, the variation of the weld pool surface slope, and emissivity. Moreover, the infrared cameras are expensive and fragile under harsh conditions at the manufacturing sites. Chen et al. (2009) developed a welding robot system including two function modules (visual and data acquisition modules) and its corresponding software system. From the top view, they measured the weld penetration depth in the welding of aluminum alloy; however, the images were affected by strong arc light, and only the weld pool profile could be obtained. The depth of weld pool in two-dimensional images, which is considered as the most important information for weld penetration control, could not be extracted and applied. Saeed et al. (2004) proposed a technique for measuring the three-dimensional (3D) weld pool surface from specular reflection of laser beams, by using a simulated environment and a mathematical model for 3D surface measurement. This method could only be applied to the GTAW process in which the weld pool is relatively stable and its surface slopes are small. Furthermore, image processing, which utilizes concepts of optical flow, structured light, and feature point tracking, is complicated. Zhang et al. (2006) proposed a structured light method to visualize the

S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527

weld pool in GTAW using the specular property of the weld pool and eliminating the effect of the strong arc; however, the reconstruction of the weld pool was complicated (Song and Zhang, 2007); moreover, the method was applied in gas metal arc welding (Ma and Zhang, 2009). Mnich et al. (2004) developed a method for sensing 3D weld pool information using binocular vision; however, in this method, the arc hinders obtaining information about the geometry underneath the torch, the synchronization of two cameras is rigorous because of the dynamics of the weld pool surface, and the reconstruction algorithm is difficult and complex because of the lack of features on the weld pool surface. As can be seen, all the 3D vision methods of weld pool are usually constricted by complex computation, which makes it difficult to use them in real time at manufacturing sites. Tao et al. (2014) investigated acoustic emissions to detect welding porosity and incomplete penetration. They distinguished two types of welding defects using the amplitude and centroid frequency of acoustic emissions. The acoustic emission technique used at manufacturing sites requires noisereduction methods and an efficient and fast data processing system when welding defects are encountered. Guu and Rokhlin (1989) developed a computerized radiographic weld penetration control system, in which weld pool depression is used as feedback. The Xray systems are expensive and well-designed radiation protection systems should be installed at manufacturing sites for the sake of safety of the workers. The applicability of most of the above-mentioned sensing techniques is limited under practical conditions. Therefore, to characterize the weld penetration status, electric signals (namely welding current and arc voltage), which are easily obtained and processed at the manufacturing sites, are proposed. A full penetration weld-sensing concept for medium-thickness steel plates is presented in Section 2, in which the feature signal, which is discussed in Section 3, plays the key role. A series of experiments were designed and conducted, as described in Section 4, to prove that the fluctuation of the arc voltage can reflect the weld penetration status. The experimental results are processed and discussed in detail in Section 5.

2. Full penetration welding conception GTAW has a low heat input and thereby has a low efficiency. To obtain a larger penetration, pulse current is used, namely pulsed GTAW (GTAW-P), but a full penetration weld cannot be guaranteed in a GTAW-P process. As a result, a “step-by-step welding” method is proposed, shown in Fig. 1, in which the welding torch remains stagnant at a spot until the arc heats the weld pool to achieve full penetration and then travels the “step distance” to heat up the next weld spot. In this way, a full penetration pass can be obtained as long as the step distance is reasonable and penetration sensing works well. It is apparent that the key to realize this concept is to obtain a feature signal for full penetrations at each weld spot.

3. Proposed feature signal To mimic the behavior of a skilled welder in full penetration welding, the weld pool surface should be captured by visual sensors, which function as the eyes of the welder, and be processed by an algorithm, which works as the brain of the welder, in an automatic welding system. However, it is difficult to obtain a clear image of the weld pool surface because of the strong arc and specular reflection, as well as the feature signal from a dynamic weld pool surface. Therefore, a systematic analysis of the GTAW-P process is required for understanding different physical phenomena


Fig. 1. Step-by-step welding.

(Zhao et al., 2004), including the weld pool surface behaviors that reflect the weld penetration status. During welding, the shape and size of the weld pool change with the variation of arc force, heat input, and boundary constraints around the weld pool, among others. A fully penetrated weld pool shows a clear difference in the bottom boundary constraint compared to a partially penetrated weld pool, as can be seen in Fig. 2. Under the condition of full penetration, the weld pool will fluctuate intensely in the absence of the solid boundary constraint at the bottom (in liquid boundary constraint). From the point of view of weld pool growth in single-weld-spot welding, the fluctuation of weld pool will show an abrupt change at the same time when the solid boundary constraint at the bottom recedes, which also indicates full penetration of the weld spot. The fluctuation of the weld pool can be expressed indirectly by the change of the arc voltage in the GTAW-P process (as the tungsten will not be consumed in the GTAW, it can be considered as a constant constraint for arc, the fluctuation of the arc is only affected by the constraint at the side of weld pool); therefore, the abrupt change in the fluctuation amplitude of arc voltage in the peak duration (U’k ) is proposed to be the feature signal to identify the time when full penetration is achieved in each weld spot. If U’k acts as an effective indicator of the penetration status, the step-by-step GTAW-P system shown in Fig. 3 will function efficiently. 4. Experiment design A series of experiments were designed and conducted in order to prove the concept presented in Section 3. The practicability of the concept and stability of the feature signal were also verified. A fixed single-weld spot, which is representative of the weld spots in the pass, is welded by the step-by-step GTAW-P method. Although there are some differences between two weld spots in the step-by-step welding because of heat accumulation, weld penetration sensing is not affected, as discussed in Section 3. The boundary constraint at the bottom and the arc force affect the fluctuation amplitude of the weld pool. In order to study the effect of only the boundary constraint at the bottom, the peak current (Ip ) is set at a constant value (400 A) for eliminating the effect of different arc forces. To obtain the desired information in the peak duration, the base current and its duration are kept constant;


S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527

Fig. 2. Diagram for GTAW-P processes.

Fig. 3. The step-by-step GTAW-P system. Table 1 Main experimental parameter settings. Experiment No.

tp /ms



1 2 3 4 5 6 7

950 450 283 200 150 117 93

95 90 85 80 75 70 65

1 2 3 4 5 6 7

therefore, the base current amperage (Ib ) and base current duration (tb ) are set as 240 A and 50 ms in the study. The peak current duration (tp ) is only altered, as shown in Table 1. Moreover, the duty cycle (D) and the frequency of welding current (f) are changed with the peak current duration. Other parameters are set as follows: steel Q345-B (GB/T 15912008) with dimensions of 300 mm × 150 mm × 20 mm is used as the base material; the workpieces are machined to obtain

a groove of 30◦ and a root face of 2 mm for butt welding; the distance from the electrode to the workpiece is 5 mm; the shielding gas is argon (99.99%) and its flow rate is set to 15 L/min. The experiments mentioned above were conducted using the setup shown in Fig. 4. The power supply is set at a constant current (CC) mode. A video camera captures images from the backside of the weld at 30 frames per second into a SD card. Simultaneously, the arc voltage and actual welding current are measured by the Hall sensors, sampled by the DAQ card, and stored in the computer.

5. Results and discussion 5.1. Experimental results and data processing The electrical signals and the image data for different peak current durations were collected from the seven experiments.

S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527


Fig. 4. Experimental system.

17 16 15


14 13 12 11 10 9 8 7







Time/s Fig. 5. Original voltage signals.

Figs. 5 and 6 illustrate the original electrical signals collected from 1 s to 2 s in Experiment No. 4. The fluctuation amplitude of arc voltage in the peak duration (Uk ) can be calculated from Eq. (1), Uk = Uk−max − Uk−min


where, Uk−max and Uk−min represent the maximum and the minimum of the arc voltage (in Volts) in the peak duration k. The variation of Uk with time (or the number of pulse cycles) is illustrated in Fig. 7a–g according to the experiment number, and the corresponding images of the backside weld at the time of the abrupt change in Uk (at the end point (tep )) are also displayed in Fig. 7a–g. Fig. 7 shows an abrupt change of Uk from Experiment No. 1 to No. 7 and the corresponding images show full penetration. The experimental results agree with the proposed feature signal. Uk

fluctuates before and after the points of the sudden variation of Uk , but these fluctuations are not drastic. The variations of Uk before the abrupt change are less because of the boundary restraint at the bottom, while the fluctuations of Uk after the abrupt change become larger. The abrupt change of Uk also occurs beyond one cycle, as can be seen in Fig. 7. There is a time interval (ti (sp-ep)) between the start point (sp) and the end point (ep) of the abrupt change. A change in amplitude (U’k ) is also observed. The detailed data regarding the abrupt change of Uk are displayed in Table 2. 5.2. Penetration time The penetration time, namely the time that the weld spot achieved the full penetration (which equals to tep ), is shown in


S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527















Time/s Fig. 6. Original current signals.

Table 2 Data statistics for the sudden change of Uk . Experiment No.

tp /ms

tep /s(The number of cycles)

ti (sp-ep)/s(The number of cycles)

The amplitude of the abrupt change of Uk /V

1 2 3 4 5 6 7

950 450 283 200 150 117 93

17.16(13) 9.84(12) 6.72(13) 5.12(12) 4.36(11) 4.08(8) 3.68(8)

1(1) 1(2) 0.67(2) 0.75(2) 0.6(3) 0.83(3) 0.14(1)

1.50 2.12 2.10 3.00 1.82 3.05 2.84

Table 2. tep Becomes shorter with the decrease in the peak current duration and the increase in frequency and in fact, the theoretical average current (TAC) is reduced. TAC = Ip × D + Ib × (1 − D)


where, Ip and Ib represent the peak and the base current (in Amperes) and D represents the duty cycle for each experiment (in percentage). The actual average current (AAC) also decreases






where, i(t) represents the experimental current signals and t is time. Fig. 8 shows the TAC and AAC for the seven experiments. The theoretical heat input of welding, which is expressed as the square of the average current, also decreases. The theoretical heat input for Experiment No. 7 decreases by about 23% compared to that for Experiment No. 1. The time required to penetrate the workpiece in Experiment No. 7 is 21.5% of that required in Experiment No. 1. Therefore, increasing the impact times for large arc forces on the weld pool with increasing frequency is advantageous for penetration depth increment compared to the decrease of heat input in the experiments with a high heat input. 5.3. Feature signal determination As shown in Table 2, the time interval of the abrupt change of Uk , namely ti (sp-ep), also decreases with the increase of frequency for increasing impacts of arc force. Moreover, the time

interval in different experiments varies from 1 to 3 cycles. Hence, the amplitude of the abrupt change of Uk , namely U’k , can be calculated from the following equation. Uk = Uk − Uk−4


U’k is the feature signal for full penetration. If the value of U’k is over a threshold, full penetration is achieved and the torch subsequently moves to the next weld spot in the step-by-step GTAW-P process. Since the values of U’k (from Table 2) range from 1.50 V to 3.05 V, and U’k has an approximate incremental trend, 1.50 V should be defined as the minimum threshold for weld penetration control in the step-by-step GTAW-P process under the abovementioned welding conditions.The threshold determined (1.5 V) to indicate full penetration is specific to the particular welding procedure used in this work. The threshold will be different if the welding conditions, such as the material, thickness (even groove) and welding parameters, are varied. For example, the wetting conditions for the aluminum alloys are worse than the steel, which makes aluminum pool more difficult to fluctuate than steel pool, hence the threshold for the aluminum alloys is smaller than that for the steels at the same thickness. In the condition of Section 4 except the peak current (300 A herein), the threshold for Q345-B with a thickness of 10 mm is 1.10 V while that for 6061 aluminum alloy with a thickness of 10 mm is 0.95 V. However, the method using the abrupt change of Uk to detect the penetration will still work in other welding conditions with different thresholds. A database

S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527


Fig. 7. Experiment results. (a) Results for Experiment No. 1. (b) Results for Experiment No. 2. (c) Results for Experiment No. 3. (d) Results for Experiment No. 4. (e) Results for Experiment No. 5. (f) Results for Experiment No. 6. (g) Results for Experiment No. 7.


S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527


410 400


390 380 370 360 350 340




4 Frequency/Hz




Fig. 8. Theoretical average current and actual average current.

Fig. 9. Typical status for full penetration.

for thresholds will be in need if the method is used in different conditions. 5.4. Image processing The corresponding images with the feature signals are extracted from the video. As the images are sampled synchronously with the electrical signals, their joint analysis for weld penetration sensing can be readily performed. The image processing is based on Wien’s displacement law, which describes a direct correlation between “max /␮m” and “T/K”, max × T = C


where, C is a constant, max is the maximum wavelength of light that steel emits, and T is the temperature of steel. Hence, steel temperature can be assessed from the color of light (i.e., the wavelength of light). Yellowish-white or bright white indicates that the

metal temperature reaches between 1500 ◦ C and 1600 ◦ C, which is above the melting point of steel. Area with yellowish-white, namely which have color (R = 255, G > 245, B > 120) but not white, were taken as the weld pool boundary and were in semi-solid status; while area with the bright white color (R = 255, G = 255; B = 255) were taken as backside weld pool and were in the liquid status. Fig. 9 illustrates the typical status for full penetration. Image processing suggests that all the backside weld images in Fig. 7 correspond to full penetration as yellowish-white or bright white areas can be observed in the center of the images, but with slight differences due to smog, light intensity variation, and so on. Fig. 10 shows the typical propagation of backside weld in singleweld-spot welding, which illustrates the relationship between the arc voltage signals and the weld penetration status. As shown in Fig. 10, for the start of the abrupt change, it starts to have bright white color surrounding by yellowish-white at the center of backside weld (which shows sign to melt) due to the emission

S. Zhang et al. / Journal of Materials Processing Technology 229 (2016) 520–527


from partial penetration to full penetration. If U’k is over a threshold (1.50 V herein), a fully penetrated weld spot is obtained in the pulsed gas tungsten arc welding. Taking U’k as the feature signal, a full penetration control system for step-by-step pulsed gas tungsten arc welding can be easily established at manufacturing sites. References

Fig. 10. Typical propagation of backside weld in the single-weld-spot welding.

of backside weld; for the end of the abrupt change, the backside molten pool shows bright white in a circular area. 6. Conclusions The fluctuation amplitude of the arc voltage in the peak duration (Uk ) reflects the weld penetration status in the single-weld-spot GTAW-P process. An abrupt change of Uk , namely U’k , occurs in every single-weld-spot process, which indicates the transition

Chen, H.B., Lv, F.L., Lin, T., Chen, S.B., 2009. Closed-loop control of robotic arc welding system with full-penetration monitoring. J. Intell. Robot. Syst. 56, 565–578. Fan, H.B., Ravala, N.K., Wikle III, H.C., Chin, B.A., 2003. Low-cost infrared sensing system for monitoring the welding process in the presence of plate inclination angle. J. Mater. Process. Technol. 140, 668–675. Guu, A.C., Rokhlin, S.I., 1989. Computerized radiographic weld penetration control with feedback on weld pool depression. Mater. Eval. 10, 1204–1210. Ma, X.J., Zhang, Y.M., 2009. Reflection of illumination laser from gas metal arc weld pool surface. Meas. Sci. Technol. 20 (11), 1–8. Mnich, C., Al-Bayat, F., Debrunner, C., 2004. In situ weld pool measurement using stereovision. In: Proceedings of 2004 JUSFA,2004 Japan-USA Symposium on Flexible Automation, ASME, pp. 1–2. Saeed, G., Lou, M., Zhang, Y.M., 2004. Computation of 3D weld pool surface from the slope field and point tracking of laser beams. Meas. Sci. Technol. 15, 389–403. Song, H.S., Zhang, Y.M., 2007. Three-dimensional reconstruction of specular surface for a gas tungsten arc weld pool. Meas. Sci. Technol. 18, 3751–3767. Tao, Y., Wang, W., Sun, B.S., 2014. Nondestructive online detection of welding defects in track crane boom using acoustic emission technique. Adv. Mech. Eng. 6, Article ID. 505464. Vasudevan, M., Chandrasekhar, N., Maduraimuthu, V., Bhaduri, A.K., Raj, B., 2011. Real-time monitoring of weld pool during GTAW using infra-red thermography and analysis of infra-red thermal images. Weld. World 55 (7–8), 83–89. Zhao, P.C., Wu, C.S., Zhang, Y.M., 2004. Numerical simulation of the dynamic characteristics of weld pool geometry with step-changes of welding parameters. Modell. Simul. Mater. Sci. Eng. 12, 765–780. Zhang, Y.M., Song, H.S., Saeed, G., 2006. Observation of a dynamic specular weld pool surface. Measur. Sci. Technol. 17, L9–L12.