WELD PENETRATION CONTROL DURING GTA WELDING USING

WELD POOL OSCILLATION SENSING

M.J.M. Hermans", B.Y. Yudodibroto* and G. den Ouden*

ABSTRACT

In this paper attention is given to on-line control of weld penetration during GTA welding by means of weld pool oscillation frequency measurements.

Weld pool oscillations can be triggered by applying short current pulses. Due to changes in arc length as a result of the oscillation of the weld pool, the frequency of the oscillation can be measured from the arc voltage signal. Different modes of oscillation can be distinghuised. For each mode the oscillation frequency depends on the weld pool geometry, the density of the liquid metal and the surface tension. It appears that a dramatic change in oscillation frequency occurs when a transition takes place from partial to full penetration or vice versa. This phenomenon is used as a sensing tool for penetration control.

By applying a feedback loop in the measuring and control system, a situation can be obtained where full penetration can be maintained by welding current adjustment.

Experiments were carried out both in bead-on-plate welding as in orbital tube welding, which show that good results can be obtained. It appears that shielding gas composition plays an important role as it influences the electric field strength in the arc and thus the detectability of the changes in arc voltage/length.

KEYWORDS

GTA welding, sensing and control, weld penetration control, weld pool oscillation

INTRODUCTION

Over the last decade a large amount of research has been carried out in the field of sensing and control during welding. Development in sensor systems for automation of welding processes is essential to increase productivity and to detect defects in an early stage of production. A large variety of tactile and non-tactile sensing systems have been developed for seam positioning, seam tracking, process control, penetration control and detection of weld pool dimensions [refs.1-10]. Especially through-the-arc-sensing is an elegant and relatively simple method, as it does not involve additional equipment.

This paper deals with penetration sensing and control during GTAW based on the occurrence of oscillations in the weld pool. The oscillation frequency of the weld pool is calculated from variations in the voltage. As the mode and the frequency of the oscillation are related with the

* Delft University of Technology, Joining Technology, Delft, The Netherlands

state of penetration, this information can be used for penetration control. The benefits of the developed feedback system incorporate adjustment of the current in order to prevent overpenetration due to accumulation of heat or underpenetration, for example as a result of castto-cast variations, and makes additional NDT superfluous.

In order to obtain oscillations in the weld pool, a short current pulse superimposed on the welding current is applied, which increases the arc pressure temporarily. Due to the variation of the arc length and thus the arc voltage, the frequency of the oscillation can be calculated from the collected voltage data. It must be noted that as the sensor is based on measuring arc voltage variations, the detectability in the case of GTAW with helium gas shielding is better than with argon gas shielding, as the electric field strength in the arc column is higher (Ar: E = ~1 V/mm, He: E=~2 V/mm).

A weld pool can oscillate in different modes depending on the external circumstances. The dominant modes in the case of a partially penetrated weld pool are schematically depicted in Figure 1. Mode 1 represents an up-and-down movement of the weld pool, whereas mode 2 represents a front-to-back movement. As these modes are surface oscillations, the oscillation frequencies are relatively high. They can be expressed by the following equations [2]:

In these equations y represents the surface tension, p the density of the liquid metal in the weld pool and D the (equivalent) diameter of the weld pool. It can be seen that the oscillation frequency decreases with increasing weld pool width and that mode 2 oscillation exhibits a lower oscillation frequency. A transition from mode 1 to mode 2 oscillation occurs when increasing trigger duration, trigger current and travel speed. Especially travel speed has a strong influence, as with increasing travel speed the weld pool becomes elongated and the position of the impulse of the arc pressure shifts towards the front part of the weld pool.

In the case of full penetration the weld pool oscillates as a membrane, see Figure 2. This bulk type of oscillation has a relatively low frequency, which can be expressed by:

with H the thickness of the workpiece. Higher order modes of oscillation in the full penetration situation are to be expected in case of a more eccentric position of the arc with respect to the centre of the weld pool [Ref. 11]. However, the amplitude of these higher order oscillation modes is expected to be relatively small.

The principle of the weld pool oscillation based penetration sensing is that the oscillation frequency of the weld pool changes abruptly during the transition from a partially to a fully penetrated weld pool. By applying a simple feedback control system, the welding current can be adjusted in order to maintain correct penetration over the entire weld length. Overpenetration and underpenetration can both be detected and automatically repaired by adjusting the welding

current.

To apply the penetration sensing method based on weld pool oscillations, two different approaches can be followed.

The first approach (single frequency approach) can be used in the case of constant current welding [Ref. 8]. The basic principle of this method is to predefine and maintain an optimal oscillation frequency (control frequency) for a correct fully penetrated weld pool. During welding the oscillation frequency is measured and welding parameters are adjusted in such a way that the requirements of the full penetration control frequency are met. The advantage of this method is the better detectability of the mode 3 oscillation compared to the partially penetrated oscillation modes (mode 1 and 2) as the amplitude and thus the variations in arc voltage are relatively high. This approach is therefore suitable for small weld pool top width, i.e. relatively high welding speed conditions, and for welding under argon gas shielding. Selecting the appropriate control frequency is the key to success of this approach.

The second approach (double frequency approach) can be applied in the case of pulsed GTAW [Ref. 7]. Basically, the procedure is similar to the single frequency approach. The oscillation in the weld pool is triggered by a short current pulse and the frequency is measured during both the pulse and the base period. Correct penetration is achieved when during the pulse period an oscillation frequency is measured corresponding to the oscillation frequency of a fully penetrated weld pool, while during the base period the oscillation frequency corresponds with the frequency of a partially penetrated weld pool. In this approach mode 1 oscillation should be monitored, as the difference between the magnitude of the mode 2 oscillation frequency and the mode 3 oscillation frequency is small. Therefore the travel speed has to be relatively low. The focus of this paper will be on the single frequency approach.

Momentarily, research is carried out concerning the influence of metal transport on weld pool oscillations during GTAW with filler wire.

EXPERIMENTAL

A large number of experiments are carried out on mild steel and stainless steel in the form of plate and tube of different dimensions. The physical properties of the metals are listed in Table 1. For welding a transistorized power source was used in constant current mode. Trigger pulses were superimposed on the welding current. As shielding gas both Ar and Ar-5%H2 were used. The welding parameters are listed in Table 2.

Labview (National Instruments) was used to control the power source and shielding gas supply, for applying the trigger pulses, for data acquisition and for feedback control. Figure 3 shows a data flow chart. The input parameters are: initial base current and trigger current, trigger duration, trigger frequency and control frequency. During welding, voltage data was collected with an acquisition rate of 20.48 kHz. During every cycle 4896 data points are captured, see Figure 4. From this set of data the first 400 data points (including data before the trigger pulse and the data of the trigger pulse) and the next 400 data points (just after the trigger pulse) are omitted for data analysis. Thus, a sub-set of 4096 voltage data points remain for determining the oscillation behaviour. The DC component is removed from the signal. This data was then filtered with a band pass filter and on-line analysed by means of Fast Fourier Transformation. The calculated oscillation frequency was compared with the selected control frequency. The outcome of this comparison was then used for current feedback.

The response rate in current adjustment depends on the difference in the measured oscillation frequency and the required control frequency. If necessary, current adjustments were carried out after each cycle. As an additional safety in the system the amplitude of the peak frequency is also taken into account. In the case that the amplitude is too low, for instance when generation of a proper oscillation fails, the signal is disregarded by the feedback system.

The raw and analysed data is stored on disk. Therefore, at any position in the weld, the data of the oscillation behaviour is traceable and provides information about penetration quality.

Data acquisition and analysis for partial and full penetration Oscillation frequency measurements were carried out under various experimental conditions. Figure 5 shows the signals obtained in the case of a partially penetrated weld pool during beadon-plate welding of mild steel. Figure 5a depicts the arc voltage as function of time after triggering when the weld pool is partially penetrated. A good oscillation signal can be directly identified by the occurrence of damping. This signal is processed by FFT, which is displayed in Figure 5b. The oscillation frequency of a partially penetrated weld pool is situated at the higher regions of the frequency domain and can be associated with surface oscillations.

The situation in which full penetration occurs is presented in Figure 6. In this case, the relatively low frequency, characteristic for bulk oscillation is observed.

The measured oscillation frequency, in both situations, is in good agreement with the calculated oscillation frequency (equation 1 and 3).