Magnetic arc oscillation in mechanised orbital TIG welding

TWI Bulletin, November/December 1989

With a degree in chemical engineering and after two years' teaching experience, David Harvey joined The Welding Institute in 1986. He is now a Head of Section in the Arc Welding Department with responsibility for research in the gas-shielded processes.

Since joining TWI, David has looked at a number of difficult areas relating to MIG/MAG welding, ranging from materials problems such as porosity in aluminium alloys to process problems such as contact tip performance, welding wire condition and arc stability. More recently, he has been involved in the application of magnetic arc oscillation to the orbital TIG process and development of a backface penetration control system for TIG welding thin sheet materials.

An article in our July/August issue reported the successful application of magnetically-controlled arc oscillation to orbital TIG welding. Here David Harvey describes TWl's magnetic arc control unit, which has been operated with an orbital welding system with automatic arc voltage control.

However, the scope of the initial study was limited in two ways:

1. Full automation of the welding process could not be achieved because the orbital welding system required manual adjustment of the arc length as the head rotated around the pipe.
2. The work was carried out on a single, non-ferritic stainless steel.

This article describes an orbital welding system with automatic arc voltage control in conjunction with magnetically controlled arc oscillation, used to develop welding procedures for other commonly used pipe materials, including a low alloy ferritic steel, a C-Mn steel and a cupro-nickel alloy. Pipe sizes were selected which fell within the range of the modified welding head (36-80mm 0D).

Equipment

The intermediate function orbital welding system used during this experimental programme was chosen for three specific features ( Figure 1):

• The arc length could be automatically controlled by the arc voltage,
• The wire feed was continuous.
• The welding current could be continuous or pulsed.

The power supply was a thyristor-controlled transformer/rectifier with an output of 200A at a 60% duty cycle. The welding head was of the horseshoe type with clamp-on feet. In addition to the arc voltage control, the welding head had a number of features which were not used, including mechanical cross seam oscillation with end dwells.

Magnetic arc oscillation

The unit used in the earlier work was replaced with a magnetic arc oscillation control system, designed and built at TWI ( Figure 2). The original unit had uncalibrated field strength, end dwells and oscillation speed controls. Furthermore, the two thumbwheel switches for each parameter resulted in discontinuities when changing the settings during welding.

The new unit has the advantage of direct setting of the end dwells and speed of oscillation (or ramp). In addition, multi-turn potentiometers set the arc oscillation parameters, resulting in continuous linear scales. This unit comprises a 15V power supply, a waveform generator, a coil driver and a remote control pendant, with five fully independent multi-turn potentiometer controls ( Figure 3). 

The unit drives an electromagnet - also built at TWI - which comprises a shellac-coated copper wire coil wound around an insulated aluminium bobbin surrounding a soft iron core with a projecting pole piece.

Modifications

The electromagnet was connected to the torch arm, which was rack driven in response to the arc voltage control, so that its position in relation to the tungsten electrode was fixed ( Figure 4).

In contrast to the previous arrangement, the axis of the coil was mounted transverse to the welding direction, because of access difficulties. However, the pole piece remained parallel to the welding direction, and aligned with the electrode.

Materials

Four pipe materials, filler wires and inserts were used:

Specifications and detailed chemical analyses are shown in Table 1. Argon shielding gas was used for the three steels and argon with 5% hydrogen for the cupro-nickel. An argon back purge was employed for each material on all passes.

Table 1: Chemical analyses of pipe material, filler wire and consumable inserts.

Table 1: Chemical analyses of pipe material, filler wire and consumable inserts - continued.

Welding

Joint geometry

It was shown clearly in the earlier work that a narrow gap preparation can be used with magnetic arc oscillation without lack-of-sidewall fusion. The joint geometry used for the stainless steel and the cupro-nickel pipes, both of which employed an EB insert in the root pass, included a 5° bevel at a tangent to a 2mm radius, a 1.0mm root face and a 1.0mm land ( Figure 5). A similar geometry was adopted for the ferritic steels, but with a root procedure including wire fill, the land was increased to 1.5mm and the root face to 2.0mm ( Figure 5).

Root pass

Prepared pipe lengths of between 0.3 and 0.4m were aligned in V-blocks with the root faces abutting (the EB insert in the case of the stainless steel and the cupro-nickel) and joined with three equally-spaced tack welds. Pulsed welding current procedures were used in all cases ( Tables 2-4). The procedure developed for the root pass in the grade 440 steel pipe was found to be readily transferable to the type 620 steel; the differences in composition were seemingly too small to influence the weld pool behaviour.

Table 2: Optimised root pass parameters for 60mm OD x 5.7mm wall type 304 stainless steel pipe in the 5G position.

Table 3: Optimised root pass parameters for 38mm OD x 6.5mm wall CN107 cupro-nickel pipe in the 5G position.

Table 4: Optimised root pass parameters for 51mm OD x 6.7mm wall grade 440 steel and type 620 steel pipe in the 5G position.

Filler passes

Welding procedure development was kept as simple as possible by using a continuous welding current and continuously fed wire, so that simultaneous use of magnetic arc oscillation and arc voltage control could be assessed. The orbital welding system has fully automatic control of the welding program, but the magnetic arc oscillation unit, which was not interfaced to the control unit, was switched on only when the arc had been stabilised, ie after 1-2s.

The first filler pass was not attempted until the root pass had cooled to room temperature. Thereafter, welding was carried out without any deliberate interpass cooling periods. However, short breaks were required to permit cleaning of the weld, rewinding the cables and modifying the welding and magnetic parameters.

Type 304 steel

The welding procedure described by Harvey ^[1] was used as the basis for developing this procedure which also employed arc voltage control of the arc length. Once again, four filler passes were required to complete the weld ( Table 5).

Table 5: Optimised filler pass parameters for 60mm OD x 5.7mm wall type 304 stainless steel pipe in the 5G position, using magnetically controlled arc oscillation.

Several modifications were made to the procedure - in particular a slight increase in the travel speed - which reduced the time to complete a revolution from 145 to 135s and produced a more consistent weld bead profile.

CN107 cupro-nickel

Despite the use of a 'hotter' shielding gas (argon-5%H ₂), the cupro-nickel weld pool was appreciably less fluid than the stainless steel.

This meant that a stronger magnetic field was needed to wash the weld pool into the sidewalls. A four-pass procedure was required to fill this weld ( Table 6).

Table 6: Optimised filler pass parameters for 38mm OD x 6.5mm wall CN107 cupro-nickel pipe in the 5G position, using magnetically controlled arc oscillation.

Grade 440 steel and type 620 steel

The procedure for the stainless steel pipe was used as the basis for the filler passes. To avoid repenetration of the root, which was thinner than that produced with the insert in the stainless steel, it was necessary to reduce the welding current and wire feed speed during the first filler pass. The remaining three passes were somewhat heavier than those required for the stainless steel, because of the greater joint depth ( Table 7).

Table 7: Optimised filler pass parameters for 51mm OD x 6.7mm wall grade 440 steel and type 620 steel pipe in the 5G position, using magnetically controlled arc oscillation.

With the exception of the first pass, the magnetic parameters used were similar to those for stainless steel, reflecting the similarity in the weld pool fluidity. The welding procedure developed for the grade 440 steel was transferable to the type 620 steel.

Assessment

Four equally spaced bore protrusion measurements were taken with vernier calipers on each completed weld and assessed in accordance with BS 4677: 1984. For pipes of internal diameter 50mm or less the maximum permitted bore constriction is 3.0mm with a protrusion no greater than 1.5mm at any point on the circumference. All welds completed using the magnetic arc oscillation process met these requirements.

Transverse macrosections of each weld confirmed the radiographic results by demonstrating satisfactory root and sidewall fusion ( Figures 6-9). For comparison, macrosections were also taken from a weld made in type 304 stainless steel using magnetic arc oscillation, but without automatic arc voltage control ( Figure 10), and from a cupro-nickel stringer bead weld ( Figure 11).

Economic comparison

The Institute's WELDCOST computer program ^[2] was used to calculate the costs per metre for each welding procedure and, for comparison, a number of process variants ( Figure 12). The analysis takes into account the arc time, duty cycle, weld volume, costs of consumables, energy, labour, plant and plant maintenance, plant depreciation and interest rates.

Type 304 stainless steel

Magnetic arc oscillation with arc voltage control (WELDCOST 1/2 in Figure 12) reduced the weld cost and completion time to a little over half the respective figures for the stringer bead technique (1/0).

This was also 20% less than the cost of the equivalent mechanical welding head oscillation process (1/3), and about 8% less than the magnetically controlled process with manual adjustment of the arc length (1/1). The additional cost of the arc voltage control is balanced by a reduction in labour costs.

CNI07 cupro-nickel

The combination of the narrow gap joint geometry and magnetic arc oscillation with automatic arc voltage control (1/5) reduced the cost and the weld completion time to less than half the respective figures for the stringer bead process (1/4).

Grade 440 and type 620 steels

The costs of labour, power and equipment are very similar to those for type 304 stainless steel (1/2), and the marginally reduced costs for the grade 440 (1/6) and type 620 (1/7) steels reflect the lower prices of consumables.

Review

The unit was designed to simplify programming arc oscillation parameters. The values for the end dwells and the ramp can be set directly and the amplitude and offset of the magnetic field are linear controls. This is a significant advance on the arc oscillation control unit used in the previous study, where the settings had no direct or linear relationship to the control parameters.

It was not necessary to water cool the electromagnet for this study. It did not overheat as a result of the resistive heating of the wire coil or the radiated heat from the arc or the weld. Typically, temperatures between 40 and 60°C were recorded on the side of the electromagnet facing the arc on completion of each pass. However, we recommend that the electromagnet should be water-cooled for production welding, since the duty cycle is likely to be considerably higher. This could be achieved by tapping water off the torch cooling system, as described by Harvey. ^[1]

The main advantage of arc voltage control was that it made the process fully automatic. The type 304 stainless steel welds were similar to those completed in the previous study without arc voltage control. Macrosections of these welds ( Figures 6 and 10) confirming this observation are not surprising, since orbital TIG welding is governed by the weld pool behaviour in all positions, and the joint geometry and resultant surface tension forces are the same in both cases.

Initially, we encountered difficulties with the electrode stubbing into the weld pool. This was not related to use of magnetic arc oscillation, but occurred because the arc length was found to be sensitive to small differences in the arc voltage, eg an arc voltage 0.2-0.3V below that required reduced the arc length from an acceptable 1.5mm to less than 1.0mm.

We expected that, with the arc length changing in response to the magnetic oscillation, the arc voltage control system would adjust the electrode-to-workpiece distance to the extent that it might prove difficult to achieve a constant weave width. However, the arc voltage control system was not particularly sensitive to these rapid fluctuations.

We were also concerned that the narrowness of the preparation might result in the arc voltage sensing system using the sidewall as a reference, rather than the bottom of the joint. This was not a problem however, probably because the typical electrode-tip-to-sidewall distance of 2.5-3.0mm is sufficiently greater than the normal undeflected arc length of approximately 1.5mm.

We expected some difficulties with the grade 440 and type 620 steels. As they are both ferromagnetic, there might be bias of the magnetic field to one side of the joint. However, the residual magnetism found in the pipes after welding was generally less than 5gauss, which is considerably weaker than the minimum strength at the tip of the electrode of approximately 25gauss.

We did not carry out a detailed parametric study, but made a number of observations concerning selection of the magnetic parameters:

1. Since the first filler pass was deposited on to a cold root, it was essential to use a strong magnetic field to wash the weld pool into the joint radius;
2. The second and third passes were less sensitive to variation in the magnetic parameters, and the magnetic field strength required to wash the weld pool into the sidewall was lower than that needed on the first pass, since the pipe was hot by then;
3. To produce an optimum weld profile and minimum weld bead width, the magnetic parameters were more critical for the final pass.

The conclusions we drew from this study on the application of magnetically controlled arc oscillation to arc voltage controlled, orbital TIG welding were:

1. Magnetic arc oscillation and arc voltage control can be simultaneously operated, and after removing the need to adjust the arc length manually, the process becomes fully automatic. This provides a low cost alternative to mechanical welding head oscillation with automatic arc voltage control.
2. The magnetic arc control system designed and manufactured by TWI was suitable for orbital TIG welding. All the controls are linear, the time functions (end dwells, ramp) being directly set.
3. Magnetic arc oscillation with arc voltage control is a suitable technique for welding CN107 cupro-nickel, grade 440 steel and type 620 steel. There were no difficulties with magnetic field behaviour, although both steels are ferromagnetic.
4. The combination of automatic arc voltage control and magnetically controlled arc oscillation may be applied to narrow gap joint geometries.

References