DEVELOPMENT OF ROTATING GMA WELDING SYSTEM AND

ITS APPLICATION TO ARC SENSORS

C.-H. Kim* and S.-J. Na*

ABSTRACT

The sensitivity of rotating arc sensor is related with the dynamic wire melting due to the insufficient self-regulation of the arc length. This paper presents the dynamic simulation of wire melting by using the variable space network method and by modeling the heat flux from the molten end of the wire into the electrode. A new type of arc rotation mechanism with a hallowshaft motor was devised to implement a high-speed rotating arc and used to develop the arc

sensor.

KEYWORDS

GMA welding, Arc Sensor, Dynamic Wire Melting, Rotating Arc

INTRODUCTION

Through-arc-sensing is widely used for automatic seam tracking because of its many advantages such as the possibilities for real time control, no auxiliary parts around the welding torch, no need for maintenance and low cost. The arc sensing method uses the electrical arc as a sensor and is based on signal variations as a function of the CTWD (contact-tip-to-workpiece distance). Therefore it is generally necessary to weave or rotate the welding torch to intentionally stimulate the differences in the CTWD. With the conventional torch weaving method the upper limit of oscillation frequency is about 4-5Hz owing to the mechanical restraint. The arc rotation method enables a high-speed rotation of arc over several tens of Hz. In Japan and Germany, arc rotation mechanisms that rotate the electrode nozzle by an external motor have been developed (Ref. 1-2). However, accessibility to the joint location may be limited by the external motor system attached to the welding torch.

Self-regulation in GMAW occurs due to using a constant voltage power source. The arc length remains approximately constant for variations in the CTWD, because the time constant of the self-regulating process is shorter than the torch weaving rate in conventional GMA welding (Ref. 3). However, the self-regulation of the arc length is not fully performed in high-speed rotating GMAW owing to a rapid movement of the welding torch. Consequently the rotating arc sensor operates in a dynamic state and the sensitivity of the sensor is greater than that of the conventional weaving arc sensor.

In a static state, the electrode melting model was confirmed experimentally by Lesnewich in 1958, and was proved theoretically by Halmoy in 1979 and experimentally by many researchers.

* Dept. of Mech. Eng., Korea Advanced Institute of Science and Technology (KAIST), 373-1 Kusung Dong, Yusong Gu, Taejeon, Korea (South)

Recently, dynamic wire melting models have been developed using an energy balance approach which uses an 'action integral' and neglecting the heat conduction along the electrode wire (Ref. 4). However, Kim et al. addressed that the Peclet number is not high enough to neglect the thermal conduction (Ref. 5). They modeled the heat transferred to the melting tip of the solid electrode from two sources: the heat transferred from the molten end of the wire, and the heat directly delivered to the solid electrode by electron condensation.

In this study, a dynamic model of the GMA welding process and an arc rotation mechanism were developed for high-speed rotating arc welding. The mathematical model includes :

dynamic model that predicts the electrode melting by considering the heat conduction in the electrode and the heat transfer by condensing the electrons directly from the welding arc to the electrode

heat transfer model that describes the heat from the molten end of the electrode wire to its solid ends

simplified weld pool model under assumption that the arc rotates very rapidly

The developed arc rotation mechanism was used to improve the weld quality and to develop an automatic seam tracking sensor. It could improve the accessibility to the joint and adaptability to the conventional welding torch system.

ARC ROTATION MECHANISM

A schematic diagram and a photograph of the developed arc rotating mechanism are shown in Fig. 1. The mechanism includes a hollow-shaft motor, an eccentric tip and 3 carbon brushes. The electrode wire is deflected circularly by an eccentric tip that is rotated by the hollow shaft of rotating motor and galvanized through the carbon brushes. This mechanism can be installed inside the electrode nozzle and connected directly to a conventional torch system. The weight and size of the nozzle can be reduced considerably by using an adequately small motor.

The arc positions are defined in Fig. 2(a) and the images of 20 Hz rotating arc taken by a highspeed camera at various arc positions are shown in Fig. 2(b). The arc length seems to be shorter at C, than at Cf, because the molten pool is formed at the position of Cr.

Figure 3 shows arc heating at the electrode and the equivalent electrical circuit of conventional GMA welding system having a power source with constant-voltage characteristics. For a dynamic state the one-dimensional conduction equation in electrode wire is given by

and the dynamic wire melting model is given by the following expression.

(1)

-k(T)

ᏧᎢ Qm(t) = (w-dl/dt)▲Hp(T)

S

(2)

In Eq. 2, Qm can be calculated from the temperature distribution of the electrode and Waszink's experimental results (Ref. 6). From Fig. 4, Qm/I shows a linear relation with the Peclet number for spray and globular transfer mode. The charcteristic length of the Peclet number is S/Le,where S is the cross-sectional area of wire and L, is the wire extension length. S is in proportion to the volumetric melting rate and Le to the Joule heating inside the electrode.

Equation 1 is a kind of one-phase Stefan problem, and for low Peclet number the conduction heat transfer along the electrode axis is not negligible. To analyze this one-phase melting phenomenon, the Murray-Landis method (the variable space network method) is used. For each iteration step, the electrode is divided into the same number of equally sized space increments, which expand or shrink in time as shown in Fig. 5 (Ref.7).

Figure 4: Heat flux entering the solid wire from the molten end

The governing equation for the whole loop of the welding circuit can be written as

where the elecrical resistance in the electrode(Re) is calculated as follows.

(3)

R ̧ = f• P.(T (2)) dz

S

(4)

The welding current signals were simulated by using eq. 1-4 and compared with experimental ones as shown in Fig. 6. To minimize the effect of weld pool shape, the amplitudes of welding current were measured for bead-on-plate welding with sinusoidal variation of torch height.

Figure 6: Sensitivity of arc sensor for bead-on-plate welding with torch height oscillation

In rotating arc welding, the CTWD is considerably affected by the weld pool shape. Especially for rotating along the rear half-circle, the arc moves above the previously formed pool and calculation of the CTWD considering the molten pool height at arc position is crucial. Analyses of dynamic 3-D GMA weld pool are regarded as of major importance but are so sophisticated and time-consuming that it is beyond the scope of this paper.

Under the assumption that the arc rotates very rapidly, a simplified 3-D quasi-steady model of molten pool was adopted in this study. The previous work has demonstrated that the conduction model, without consideration of fluid flow, can predict the shape of weld pool boundary in the vicinity of arc with reasonable accuracy (Ref. 8). A quasi-steady conduction model was used for calculation of the CTWD during high-speed arc rotation. Welding current waveforms were simulated by solving the governing equations (1-4) and the simplified weld pool model. For example, experimental and simulated results under 25 Hz rotation speed are shown in Fig. 7. Although the experimental welding signals contain high-frequency fluctuations, their basic waveforms can be predicted reasonably well by the simulations.