• Author / Uploaded
• Onil

• Categories
• Soldadura
• Arco eléctrico
• Gota (Líquido)
• Corriente eléctrica
• Plasma (Física)

Citation preview

Home

Search

Collections

Journals

About

Contact us

My IOPscience

The physics of welding

This article has been downloaded from IOPscience. Please scroll down to see the full text article. 1984 Physics in Technology 15 73 (http://iopscience.iop.org/0305-4624/15/2/I05) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 200.239.65.101 The article was downloaded on 25/09/2012 at 19:03

Please note that terms and conditions apply.

Phys. Technol.. Vol. 15. 1984. Printed in Northern Ireland

THE PHYSICS OF WELDING J F Lancaster

Greater understanding of the physics of welding is leading to improved application and control of welding processes. Further gains in welding productivity could follow

Welding is an ancient art, and has been practised ever since man first learned to extract and refine iron. Until about the beginning of this century, the method of welding was the same as that used in Roman times, and still employed in the blacksmith's forge today. The two pieces of metal to be joined are heated and then hammered or pressed together, so as to squeeze out slag and oxide and allow the surfaces to bond together. This is forge welding, and is an example of a solid-phase welding process. The alternative technique is fusion welding. Here the edges of the two pieces of metal to be joined are melted and fused together. In order to melt the metal locally in this way an intense heat source is required, and it is largely in the provision and use of such energy sources that physical problems arise. In the case of a surface heat source the minimum rate of energy release per unit area q" required to maintain a molten weld pool of radius r is approximately

q" = AkT,/r where k is thermal conductivity, T, is melting temperature and A is a factor dependent on welding speed, weld size and thermal diffusivity. The weld pool size is limited by practical considerations: it must be manageable, on the one hand, and it must be large enough to fuse the edges of a weld preparation, on the other. Manual welds in steel are usually 10-20" in width giving a required power density of the order of lo7 W m-'. The electric arc, which is the most generally used heat source in fusion welding, generates such a power density. Figure 1 shows the energy densities for various types of welding heat source. The temperature distribution relative to a point heat source of power q on the surface of a semi-infinite medium moving at velocity U in the x direction is 030~624/84/020073+07$02.2501984 The Institute of Physlcs

73

1013

10 Ion 1'

t

Vaprisptior conductton and meltin (keyholing7

Radial mduction dominated with meliing Arc processes

Figure 1 Power density for various welding processes (from Lancaster 1984)

T = (qi2;rkr) exp [- u(r - x)/2cu] where 2 = x2 + y 2 + z2 and (Y is thermal diffusivity. When T = T, and z = 0 this equation gives the theoretical boundary of the molten weld pool. Measurements of weld pool width and length have been made over a wide range of heat input rates and for different materials and processes (Christensen et a1 1965). Assuming that the thermal properties of the liquid metal are the same as for the solid, the calculated width and length of the pool are, in almost every case, smaller than the measured values. This is consistent with a high effective thermal conductivity in the liquid metal, and indicates that (as would be expected) heat transfer in the weld pool is partly convectional. In general, however, the simple formulae for temperature distribution that have been derived assuming point or line sources of heat, and ignoring variations of thermal properties with temperature, agree reasonably well with the results obtained by experiment. The manner in which heat is transferred to the weld pool depends on the character of the welding process. Most arc welding is done with a consumable electrode, so that heat generated at both poles of the arc is absorbed by the workpiece. When a non-consumable electrode is used, and the workpiece forms the positive electrode of the arc, heat is generated by the condensation of electrons and by conduction and radiation from the arc plasma across the boundary layer at the metal surface. If the workpiece is the negative pole, then a non-thermionic cathode is formed. The mechan74

ism of such non-thermionic cathodes, which is described later. implies that there is substantial power generation in a thin layer (1 nm to 1 km) at the surface. Part of this power is absorbed in the evaporation of electrons and in ionisation, but some is transferred to the metal. Normally the heat input per ampere of arc current is greater at the positive electrode, where the electrons give up their heat of condensation, than at the negative electrode, where heat is absorbed by electron emission.

The electric arc in welding: the cathode The tungsten inert gas (TIG) welding process employs a thoriated or zirconiated non-consumable tungsten electrode, and the arc operates inside an argon or. less frequently, a helium gas shield. Except for welding aluminium, the electrode forms the cathodic pole of the arc. Various cathode modes have been described, but in most cases welding is performed with a pointed electrode at the tip of which a well defined thermionic cathode spot is formed. The current density of the cathode does not vary much with electrode material (e.g. zirconium instead of tungsten) and is probably governed by conditions in the arc column. This is fortunate since with pure tungsten electrodes the tip of the electrode melts to form a small sphere. and at low currents the cathode spot wanders over the surface of the sphere, causing the arc column to wander and the welder to lose control. The addition of small amounts of zirconia or thoria to the tungsten reduces the work function. and allows thermionic emission at the required current density to occur at a lower temperature. For thoriated and zirconiated material this temperature is below the melting point, and the geometry of the electrode tip is maintained, thus stabilising the arc. Other metals that are encountered as electrodes in welding. such as steel and aluminium, boil at temperatures below that required for a sufficiently high thermionic current density level. The cathode that forms on these metals is not fixed in any one position, but moves rapidly over the surface in a random fashion. Understanding of the mechanism and behaviour of such non-thermionic cathodes has made a significant advance following recent work by Guile (1979). There are at least three types of non-thermionic cathode: the vapour type, which forms on unfilmed metal; the tunnelling type, which forms on metal having a thin oxide film (less than about 10 nm); and the switching type, which forms on thicker oxide films. Guile suggests that positive ions originating from the arc plasma condense on the oxide surface and set up a high electric field. In the case of thin films electrons may 'tunnel' through the film and generate an emitting site; for thicker

films a phenomenon known as switching makes the film locally conductive. Such mechanisms allow relatively large currents to flow in filamentary channels through the oxide. Individual emitting sites are 1nm to 1 pm in diameter, and have a lifetime of 1ns to 1 ps. Examination of the cathodic afea by the scanning electron microscope after arcing for a short duration such as 1 ps shows a pattern of craters, as in figure 2. Thus, the non-thermionic cathode operates by the formation and decay of numbers of small emitting sites, and removes the oxide film from the cathodic area. This effect is put to good use in the TIG welding of aluminium. Aluminium oxide persists as a solid film after the metal has melted, and may cause discontinuities in the completed weld. Welding with the workpiece negative results in the oxide being stripped by cathodic action, and the weld is free from oxide film. In practice alternating current is used to avoid overheating the electrode. In other cases the formation of a thick oxide film is employed to control the extent of movement of the cathode spot. Metal inert gas (MIG)welding, in which a consuniable bare wire electrode is protected by an inert gas shield, can normally be operated with electrode positive only. With electrode negative the cathode wanders up and down the electrode, making the process uncontrollable. Coating the electrode surface with a relatively thick oxide, however, confines the arc root to the tip of the rod, giving a symmetrical and controllable arc. In welding steel by the same process with electrode positive and a pure argon gas shield, the cathode spots may wander over the plate to an extent that makes the process unstable. Adding a few per cent oxygen to the argon forms a thicker oxide film on the metal surface and restricts movement of the cathode to an acceptable degree.

The arc column The column or gaseous portion of the electric arc is characterised by two features: high temperature, such that the gas is sufficiently ionised to be conductive; and high flow velocity, the direction of which is, under welding conditions, from electrode to workpiece. The temperature is maintained by ohmic heating, which balances losses by conduction, convection and radiation. The proportion of energy lost from a TIG arc by radiation increases with current and at 100A is about 20% of the total column energy. The relative importance of conduction and convection may be assessed from the Peclet number Pe

Pe = puLC,/k where p is density, U is flow velocity, L is a typical dimension, C,, is specific heat and k thermal

Figure 2 Mild steel cathode with 2.5 nm oxide film. Surface damage caused by a 4.5 A arc of 30 ns duration. Magnification X3000 (photograph courtesy A E Guile)

conductivity. For the TIG arc at atmospheric pressure Pe is about 10, at which value convection dominates. At low pressures Pe may fall below 1, where heat flow is primarily by conduction and the arc column becomes spherical in form, whilst at high pressures Pe increases above the atmospheric value. Most of the measurements of temperature distribution have been for the argon-shielded TIG arc, illustrated in figure 3. The visible boundary represents an isotherm, probably about 1 x lo4 K. The measured temperature level varies quite significantly from one observer to another; earlier investigators found temperatures of about 2 x lo4 K near the cathode, whereas others obtained values of 1 X 104K in the same location. Arcs between iron electrodes have a column temperature of about 6 X lo3 K, presumably because of the higher conductivity of iron vapour at lower temperatures.

Mass flow in the arc column Mass flow in the arc plasma may result from chemical reactions, such as the breakdown of an electrode coating, or it may be externally imposed, as in plasma welding. The plasma torch is similar to that used in TIG welding, but with a constricted nozzle so as to direct a jet of hot plasma on to the metal surface. The primary interest here, however, lies with the electromagnetically induced jets that are observed in TIG and MIG welding. In virtually all arc welding operations the current flow is between a point-like electrode and an approximately flat plate. The current streamlines therefore spread outwards from the electrode. Because of this configuration, the interaction of the current and its self-induced magnetic field results in forces that induce flow from the electrode towards the plate. The flow is jet-like and the axial velocity is of the order of hundreds of metres a second. Increasing the ambient pressure (as may occur in 75

underwater welding) causes the jet to become more intense, and vice versa. Calculations of mass and heat flow have been attempted for simple cases like the TIG arc. A complete analysis requires the simultaneous solution of the equations for conservation of mass, energy, momentum and electric charge, together with Ohm’s law and Maxwell’s equations for magnetic fields. Such an analysis is possible using numerical methods, and promising results have been obtained (Lancaster 1984). Axial flow in the welding arc column is desirable in two ways. In using coated electrodes the gas flow produced by decomposition of the coating protects the molten metal from contamination by atmospheric nitrogen and oxygen. In TIG and MIG welding the electromagnetically induced flow gives the arc the quality of ‘stiffness’; it may be directed as required and is resistant to deflection by external forces such as stray magnetic fields. The flow is converted to a stagnation pressure where it impinges on the weld pool. This pressure generates the ‘arc force’, which may have useful effects in ensuring good penetration of the molten weld into the workpiece. On the other hand, if the flow, and the resulting arc force, are too high, instabilities may occur and in extreme cases the molten metal may be blown out of its proper location. Metal transfer

In welding with a consumable electrode, the electrode is at one and tlie same time a conductor for the arc current, thereby providing a heat source, and a source of liquid filler metal for the joint to be welded. It is essential for good welding that the major part of this liquid metal is transferred to the weld pool, and not dispersed as spatter over the surrounding plate There are two ways in which a smooth transfer may be effected. Where the flight path of droplets detaching from the electrode tip is erratic, as in gas metal arc ( G M A ) ~welding with a CO2 shield, the arc is kept short and the drop contacts the weld pool before it detaches, causing a short circuit until the liquid metal is drawn into the pool by surface tension and electromagnetic forces. When the flight of such droplets is directed in line with the electrode axis, on the other hand, it is possible to operate in a ‘free flight’ mode. This is the case with argon-shielded gas metal arc welding. We will here be primarily concerned with the free flight mode of metal transfer.

t GMA welding is similar to MIG, but the shielding gas may be wholly or in part chemically active. Likewise gas tungsten arc (GTA) may employ a chemically active gas shield. 76

Figure 3 Argon-shielded arc with tungsten cathode. The arc column is typically bell-shaped (photograph courtesy The Welding Institute)

Details of the transfer process are visible in high speed motion pictures of the arc, which may if necessary by correlated with oscillographic records of arc current and voltage. Transfer from coated electrodes has been examined by radiography, and in this way the movement of metal may be distinguished from that of the slag. When the electrodes are fully deoxidised, a depression forms at the root of the arc, distorting the drop at the electrode tip and eventually resulting in detachment either by short circuit or by the pinching-off of droplets. In electrodes that are not fully deoxidised a bubble of CO forms inside the drop; eventually the bubble bursts and a spray of fine drops is projected towards the electrode. Transfer from steel electrodes in gas metal arc welding with argon, argon-oxygen or argon-COz shielding is altogether more regular in character. At low currents (below 200A with a 1.2” diameter wire) the drops form as oblate spheroids, elongated in line with the axis of the electrode. These detach with an initial velocity and acceleration at fairly regular intervals. Above 200A a conical tip appears at the end of the electrode, and droplets form and detach - again in a regular fashion - from the tip of the cone. At still higher currents (about 250 A) the conical tip transforms into a relatively long cylinder of liquid metal from the end of which a stream of fine drops is projected (streaming transfer). Further increase of current causes the cylinder to transform into a rotating spiral (rotating transfer). Applying a longitudinal magnetic field to streaming transfer causes the transition to a rotating spiral to occur at a lower current (Lancaster 1984).

Metals of higher thermal and electrical conductivity, such as aluminium and copper, do not show the same transitions of metal transfer mode as steel. Metal is detached in the form of drops, as for steel below 200A. Typical figures for the rate of drop detachment at 200 A are 10 dropsis for steel, 20dropsls for copper and 170drops/s for aluminium. The regular behaviour of transferring drops in gas metal arc welding has encouraged various investigators to attempt a quantitative analysis of the phenomenon. One method of approach has been to assess the forces to which the drop at the electrode tip is subject. Those tending to detach the drop are gravity (assuming downward welding). the drag force due to the shielding gas flow, and (usually) the electromagnetic force, whilst surface tension acts in the opposite sense. These forces have been measured for steel in the range 0-220 A . The electromagnetic force on the drop at the tip of a cylindrical electrode may be calculated assuming that there is no internal flow. The magnitude and direction of the force so calculated depends on the relationship between the diameter of the electrode and that of the anode spot. If the diameter of the anode spot is smaller than that of the electrode, then the force acts towards the electrode: when it is larger it acts in the opposite direction. It was found that at low currents, corresponding to a small anode spot size, the electromagnetic force did indeed act towards the electrode, whilst at higher values up to 1 6 0 A there was good agreement between calculated and measured values (Waszink and Graat 1983). The dynamics of metal transfer have also been explored using the linear approximation employed by Lord Rayleigh (1879) for investigating the stability of a liquid cylinder or jet. This analysis was extended to the case of a cylinder carrying an electric current by Murty (1961. see also Alfven and Falthammar 1963). Various modes of instability are possible for the current-carrying cylinder. The simplest case is the pinch, varicose or sausage-type instability. when the surface of the cylinder is deformed so that its longitudinal section has a sinusoidal form. causing it eventually to disperse into drops. Analysis of the electromagnetic, surface tension and inertia forces associated with this mode of deformation shows that there is a critical wavelength above which the system is unstable. It has been suggested that the drop at the tip of the electrode in G M A welding will grow until its length is about equal to the critical wavelength, after which it becomes unstable and may be pinched off (Lancaster 1984). The analysis also yields a time constant from which the initial velocity. acceleration and detach-

ment time of drops may be estimated as a function of current for various electrode diameters and materials. The values so obtained are consistent with experimental results (Lancaster 1984). In the presence of a longitudinal magnetic field a higher unstable mode may appear. This is the kink unstability, when the liquid cylinder collapses into an expanding spiral. Figure 4 shows metal transfer in the case of high current plasma-MIG welding. In this process an arc is maintained between an auxiliary tungsten electrode and the workpiece and the consumable electrode, which also carries a current, passes through the plasma so formed. In this instance the kink instability has developed first, but the pinch mode is visible at the end of the electrode, which eventually disperses into drops. In the normal MIG or GMA process the appearance of rotating transfer is similar to that shown in figure 4. In neither case is there an imposed magnetic field, but of course the spiral formation will generate its own longitudinal field. Thus the character of metal transfer in GMA welding is consistent, at least qualitatively, with the theory of instability of liquid cylinderst. In submerged arc welding drops are directed towards the weld pool in a different manner. This process uses a bare wire electrode and the arc and weld pool are protected from the atmosphere by a powdered flux. Flux melts around the arc, forming an expanding bubble that periodically bursts and then reforms. Drops are detached from the electrode tip in random directions, but those that fly outwards are trapped by the bubble of molten flux and therby directed into the weld pool.

Flow in the weld pool If current enters a hemisphere of liquid in a symmetrical manner from the plane surface, the electromagnetically induced flow should be toroidal in form. Such toroidal flow is rarely, if ever, observed in weld pools. The closest approach to an ideal geometry is in TIG welding at low current, when the weld pool is almost hemispherical. This type of weld pool tends to rotate. The rotation may take the form of a double circulation or, more t Recent work (including that of Murty 1961) on the stability of cylindrical systems relates mainly to problems in cosmology and nuclear fusion devices (Alfven and Falthammar 1963. Chandrasekhar 1961). It may be of interest theretore to note a possible application in welding. The quantities concerned are of course rather different. In GMA welding the time for development of an electromagnetic pinch instability is of the order s; in cosmology Chandrasekhar (1961) instances the case of a cylinder 250parsecs (7.7 x 10I8m) in diameter and of density 2 x kg r C 3 , for which the characteristic time of break-up due to gravitational instability is 10'years.

77

and other organic substances. Attempts to generate flow due to surface tension gradients in mercury were however unsuccessful, except under high vacuum. It is thought that under normal atmospheric exposure the metal surface becomes contaminated with surface-active material, and that stresses generated by temperature gradients are nullified by a redistribution of the surface-active agents.

High energy density welding

Figure 4 Metal transfer in high-current plasma-MIG welding. A rotating spiral of liquid metal eventually disperses into drops (photograph courtesy Philips, Eindhoven)

commonly, the pool rotates as a whole. The direction of rotation may be dhanged by changing the location of the earth return, and it would appear that the magnetic field due to assymmetric current flow in the workpiece is sufficient to cause rotation. In low current TIG welding the weld pool surface is flat or slightly raised, but in many welding operations a depression forms in the liquid metal below the electrode, due to the stagnation pressure generated by gas flow or to the impingement of liquid drops or both. Metal that is melted at the front of the weld pool is accelerated through the restricted cross-section around the depression. Thus there is a circulation along the bottom of the weld pool from front to back and along the surface from back to front (figure 5). This circulation convects heat backwards along the weld axis, and causes the weld pool to be more elongated than would be calculated assuming isotropic thermal properties. For weld pools generally the Peclet number lies within the range 10-5 X lo3, so that heat flow is predominantly convectional. Some investigators have suggested that flow may be induced in a weld pool by surface tension gradients. Such flows, which are toroidal, have been demonstrated in liquid pools of paraffin wax 78

The electron beam and laser welding processes are both capable of producing very narrow, deeppenetration welds, such as that illustrated in figure 6. This capability is particularly attractive for the welding of machined components such as gears and aero-engine components, since the volume occupied by the weld is much smaller than for normal fusion welding, and distortion is correspondingly reduced. Such deep penetration welds are made by producing a cavity, known as a 'keyhole', and traversing this along the joint. In electron beam and laser welding this cavity is maintained by the vaporisation of metal. The internal pressure so generated is balanced by the stress due to surface tension y in the film of liquid metal surrounding the cavity p = yfr where r is the radius of the keyhole. The tendency for a cylindrical cavity of this type to collapse inwards is counteracted by increased evaporation at the point of collapse, so that a stable configuration can be maintained. As in conventional welding processes the width of the completed weld is determined by practical considerations. A very narrow weld may require excessive accuracy in preparation and positioning of the joint, whilst if the weld is too wide it may show protrusion at the root and sink-away at the surface. Typical weld widths lie within the range 0.75-3". Now in heat flow generated by a line Figure 5 Flow pattern in a submerged arc welding pool. For clarity the electrode and arc have been omitted (from Lancaster 1984)

source, such as the deep penetration electron beam, at least half the heat is absorbed into the solid metal, the remainder being used to melt the weld metal. Assuming this minimum value, the total power of the line source is

q = 2wdvp C,T, where w is weld width, d is weld depth, v is velocity, p density and C, specific heat. This heat must be generated within the keyhole, which may reasonably be expected to have a radius about one-half that of the weld pool, i.e. w14. Then the surface power density must be at least

q” = 16q/nw2 which for a 5 mm weld leads to a minimum power density of 1.5 X lo1’ Wm-’. This is indeed at the bottom end of the spectrum for deep penetration electron beam welding. Plasma welding may be operated in the keyhole mode, but the power density is lower than for electron beam welding, and the cavity is maintained by pressure from the plasma jet. Vaporisation of the,metal does not occur to any significant extent in plasma welding. Both electron beam and plasma welding torches may be operated as surface heat sources, as in electric arc welding, and for certain applications this may be advantageous. But the keyhole mode is the most important, since it makes possible novel joint configurations, reduces distortion and permits the welding of some’ of the more difficult of metals and alloys.

Future developments At present the emphasis in development work is on the improved application and control of existing processes, rather than the introduction of new methods. The major use of arc fusion welding is in the construction Of process plant, bridges, steel buildings, ships, machine frames and the like, and the processes most commonly employed here are manual welding with coated electrodes, submerged arc welding and gas metal arc welding, whilst TIG or GTA welding is used for thinner sections and special materials. There is a requirement for improved quality, and at the Same time for improved productivity through automation and the use of robots. To this end numbers Of Current investigators are working on diagnostic techniques, with the object of providing feedback control and improving the consistency of the welding operation. The use of electron beam (EB) and laser welding continues to be limited by high capital cost and, in the case of EB welding, by the need, in most cases, to evacuate the welding chamber. Nevertheless, the inherent advantages of a deep penetration keyhole

Figure 6 Typical deep-penetration weld produced by the electron beam process (photograph courtesy The Welding znstitute)

weld are such that much effort goes into methods for applying the EB process to the welding of thick sections. Success in this activity could lead to a major advance in welding productivity, always provided that the limitations of the vacuum chamber can be overcome.

References A,fvCnH and Fllthammar C-G 1963 Cosmical Electroe dynamics (Oxford: Oxford University Press) Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic Stability (New York: Dover) Christensen N, Davies V de L and Gjermundsen K 1965 ‘Distribution of temperature in arc welding’ Brit. we1ding J . U 54-74 Guile A E 1979 ‘Processes at arc cathode roots on non-refractory metals having films of their own oxide’ in Arc Physics and Weld Pool Behaviour (Cambridge: The Welding Institute) Lancaster J F (ed) 1984 The Physics of Welding (to be published Oxford: Pergamon Press for the International Institute of Welding) Murty G S 1961 ‘Instability of a conducting fluid cylinder in the presence of an axial current’ Ark. F. Fys. 19 483 Rayleigh Lord 1879 ‘On the instability of jets’ Proc. Land, Math. soc, 1o 4-13 Waszink J H and Graat L H 1983 GExperimental investigation of the forces acting on a drop of weld metal’ Welding J. 62 108-S-16-s J

79