Elsevier

Automatica

Volume 48, Issue 1, January 2012, Pages 233-238
Automatica

Brief paper
Adaptive interval model control of weld pool surface in pulsed gas metal arc welding

https://doi.org/10.1016/j.automatica.2011.09.052Get rights and content

Abstract

A skilled welder may produce quality welds with a desired weld penetration depth based on his/her observation on weld pool surface. In a recent study the authors found that the arc voltage change during the peak current period in pulsed gas metal arc welding is a relevant measurement of the weld pool surface that can estimate the penetration depth. A system is thus developed to measure the arc voltage and control its change as output using the base current period as input and the necessity to use relatively complex sensors such as imaging systems is avoided. Analysis shows that the resultant process being controlled is time-varying, noisy, and nonlinear. After simplification into a linear system, an adaptive interval model control system has been designed. Experimental results confirmed the effectiveness of the adaptive interval model control algorithm and the developed control system.

Introduction

Gas metal arc welding (GMAW) is one of the most widely used arc welding processes (Linnert, 1994). Its welding current is typically pulsed between a base and peak current to form pulsed GMAW (GMAW-P) in order to deliver heat and mass input in wide ranges while still transferring the liquid metal of melted wire into the weld pool in desired modes. To assure the mechanical strength, the depth of weld penetration or the penetration depth, i.e., how deeply the base metal has been melted, needs to be controlled at a desired level. Unfortunately, because of the lack of sensors that can measure this depth (that is not visible), all manufacturing conditions that may affect the relationship of welding parameters with this depth have to be controlled accurately at relatively high costs in order to assure that pre-determined welding parameters can produce the desired penetration depth. Although many methods have been proposed to obtain different kinds of indirect measurements (Bicknell et al., 1994, Graham and Ume, 1997, Xiao and den Ouden, 1993, Yudodibroto et al., 2004), they are typically complex and not suitable for the manufacturing environment.

Among possible indirect measurements, the weld pool surface appears to be most promising because skilled welders are capable of controlling the penetration depth based on their observation on the weld pool surface. Machine vision (Saeed and Zhang, 2007, Yoo and Lee, 2004, Zhang et al., 1996, Zhang et al., 2006) can be used to measure the weld pool surface but its application in a manufacturing environment is not most desirable because of its complexity and its needs for accurate calibration, complex and reliable image processing, and skills different from the present labor force.

This work aims at controlling the weld pool surface, as an indirect measurement, in GMAW-P to achieve the desired weld penetration depth using an easily implementable innovative method. This innovative method is based on our recent discovery documented in Wang, Zhang, and Wu (2010) and Zhang and Wang (2010) that the change in arc voltage during the peak current period possesses a simple relationship with the weld penetration depth. (This relationship is possible not only because the arc voltage is a direct measurement of arc length but also because the arc length increases as the weld pool surface becomes deeper.) Hence, an appropriate parameter in the waveform of the pulsed current can be selected as the system input and used to control the system output, i.e., the change in the arc voltage during the peak current period.

Section snippets

System description

Fig. 1 illustrates the waveforms of welding current and arc voltage. The current is pulsed between a base level and peak level with a fixed peak current period. The base current period is adjusted to change the heat input and weld penetration capability. The arc voltage continuously varies with the arc length and the change in the arc voltage measurement during a peak current period is defined as the difference between the maximal and minimal values of the voltage measurement during this period

System identification

Step response experiments with different magnitudes are conducted. The parameters are shown in Table 1, where Δt is the control/sample period, and tb(1) and tb(2) represent the input before and after the step change respectively. Other experimental conditions are: two 6.3 mm-thick mild steel (C1018) plates of 300 mm×25.4 mm were welded in a square butt joint with a 1.5 mm root opening at the flat position; the welding torch stayed stationary and the work-piece traveled with a welding tractor at

Pre-filter

A low-pass pre-filter is added to the system due to the noises: yf(k)=αyf(k1)+(1α)ỹ(k) where yf is the system output after filtering. Choosing different α produces different filtering results. When α=0.8, the filtering result is acceptable because it can effectively suppress the noises while the resultant control speed is still acceptable.

Eqs. (4), (5) result in yf(k)=αyf(k1)+b0u(k1) where b0=b(1α).

Predictive control algorithm

Large noises in the arc voltage signals affect modeling accuracy and a model predictive control algorithm which is insensitive to the modeling inaccuracy is thus introduced. The transfer function for system (6) is G(z)=b0z11αz1=b0z1+b0αz2++b0αj1zj+=h(1)z1+h(2)z2++h(j)zj+. Because the welding process is stable, it can be approximated using a finite impulse response (FIR) model with only the first N terms. The system can thus be expressed as yf(k)=h(1)u(k1)+h(2)u(k2)++h(N)u(kN).

Welding experiments

Welding experiments were done to verify the adaptive interval model control algorithm. All the experimental conditions including the material, peak current and duration, base current etc. are exactly the same as those for the step response experiments except for (a) the base current period is now determined by the adaptive interval model control algorithm and (b) different values are now used for the travel speed in different experiments. The resolution of the control variable is 1 ms because

Conclusion

  • (a)

    A control system has been developed to directly control the weld pool surface as characterized by the arc voltage change during the peak current period and to indirectly control the depth of the weld penetration for the complex but highly productive GMAW-P.

  • (b)

    Because the system only requires measurements of arc voltage signals, it provides a practical solution for a relatively poor manufacturing environment. Also, because of the fundamental role of the GMAW-P in welding and the lack of acceptable

Zhijiang Wang is currently a lecturer in Tianjin University where he received his B.S. degree in Metal Material Engineering (Welding Major) at 2003. He received his Ph.D. and M.S. degree in Material Process Engineering (Welding Major) from the State Key Laboratory of Advanced Welding Production Technology at the Harbin Institute of Technology, China. Dr. Wang, as a visiting scholar, worked with Dr. YuMing Zhang at the University of Kentucky for more than two years since 2008. His research

References (12)

  • G.M. Graham et al.

    Automated system for laser ultrasonic sensing of weld penetration

    Mechatronics

    (1997)
  • A. Bicknell et al.

    Arc voltage sensor for monitoring of penetration in TIG welds

    IEE Proceedings A: Science, Measurement and Technology

    (1994)
  • G.E. Linnert
    (1994)
  • G. Saeed et al.

    Weld pool surface depth measurement using calibrated camera and structured-light

    Measurement Science and Technology

    (2007)
  • Z. Wang et al.

    Measurement and estimation of weld pool surface depth and weld penetration in pulsed gas metal arc welding

    Welding Journal

    (2010)
  • Y.H. Xiao et al.

    Weld pool oscillation during GTA welding of mild steel

    Welding Journal

    (1993)
There are more references available in the full text version of this article.

Cited by (28)

  • Study on the weld pool oscillation behavior during pulsed gas metal arc welding

    2022, Journal of Manufacturing Processes
    Citation Excerpt :

    Although certain results have been achieved, there is still a large gap from the practical application. In addition to the above means, due to the quantitative relationship between the weld pool surface oscillation and melting penetration with clear physical meaning, so that the weld pool oscillation method [10–12] in the actual gas tungsten arc welding is widely used and achieved a better control effect. Li et al. [12] established a continuous welding critical penetration control system based on the oscillation frequency characteristics of the melt pool in different penetration states during the GTAW welding process.

  • Sensing and characterization of backside weld geometry in surface tension transfer welding of X65 pipeline

    2022, Journal of Manufacturing Processes
    Citation Excerpt :

    Welding speed is one of the welding parameters that must be controlled in the pipeline welding, which mainly affects the heat input and weld pool volume. Skilled welders can observe the weld pool dynamics to control weld penetration by adjusting welding speed [20]. To guarantee the weld penetration and to study the influences of welding speed, WFS, Ip, and Ib were set constants at 3.00 m/min, 350 A, and 65 A, respectively.

  • Feedback control of variable width in gas metal arc-based additive manufacturing

    2022, Journal of Manufacturing Processes
    Citation Excerpt :

    A highly promising approach to achieve reliable and consistent deposition is implementing in-situ monitoring and feedback control which are still in their infancy in WAAM. Electrical parameters [18] and vision sensing [19] have been extensively applied in arc-based processes from the existing publications. Electrical parameter sensing responsible for collecting arc voltage and current curves is usually employed to characterize the metal transfer and arc stability.

View all citing articles on Scopus

Zhijiang Wang is currently a lecturer in Tianjin University where he received his B.S. degree in Metal Material Engineering (Welding Major) at 2003. He received his Ph.D. and M.S. degree in Material Process Engineering (Welding Major) from the State Key Laboratory of Advanced Welding Production Technology at the Harbin Institute of Technology, China. Dr. Wang, as a visiting scholar, worked with Dr. YuMing Zhang at the University of Kentucky for more than two years since 2008. His research interests include innovative welding processes, monitoring and control of welding processes, and welding automation. His email is [email protected].

YuMing Zhang has been with the University of Kentucky since 1991 where he is currently a Professor and the James R. Boyd Professor in Electrical Engineering. He was a faculty member from 1984 to 1991 in the State Key Laboratory for Advanced Welding Production Technology at the Harbin Institute of Technology, China where he received his Ph.D. degree in Welding Major, and M.S. and B.S. degrees in Control Major. Dr. Zhang is a Fellow of the American Welding Society (AWS), a senior member of the IEEE and the SME, and a member of the ASME. He received The Donald Julius Groen Prize from The Institution of Mechanical Engineers, United Kingdom; The A.F. Davis Silver Medal award from the American Welding Society; the Adams Memorial Membership Award from the American Welding Society; and the 15th IFAC Triennial World Congress Best Poster Paper Prize and Application Paper Honorable Mention from the International Federation of Automatic Control.

Lin Wu has been with the Harbin Institute of Technology, Harbin, China since 1959 where he has been a junior lecturer, lecturer, associate professor, and professor in welding engineering. Professor Wu served the Institute as the founding director of the Center for Robotics Research, the founding director of the State Key Laboratory for Advanced Welding Production Technology, a Vice President, and the Chairman of the Board. In addition, he was one of the initial members in the Expert Commission for Automation in China’s State High-Tech Development Plan (the 863 Program), the President of the Chinese Welding Society, and a Vice President of the International Institute of Welding (IIW). Professor Wu has been recognized by the prestigious “China Welding Lifetime Achievement Award”. His contributions to welding automation, welding robots, automatic weld quality inspection, adaptive control of weld penetration, welding robot vision, intelligent welding, welding remote-control, intelligent robots, spot welding robots, and special robots have brought him 18 national and provincial prizes, 13 books and over 400 papers.

This work is funded by the National Science Foundation (Arlington, VA, USA) under grant CMMI-0726123 entitled “Measurement and Control of Dynamic Weld Pool Surface in Gas Metal Arc Welding”. This paper was not presented at any IFAC meeting. This paper was recommended for publication in revised form by Associate Editor Pedro Albertos under the direction of Editor Toshiharu Sugie.

1

Tel.: +1 859 323 3262; fax: +1 859 323 1035.

View full text