You are on page 1of 8



M. Schnick, M. Dreher, J. Zschetzsche, U. Fuessel and A. Spille-Kohoff
GMA welding is one of the most frequently applied welding techniques in industry. Particularly the joining of aluminium, high alloyed steels or titanium requires a cover of shielding gas in order to provide a low PPM concentration of oxygen. The result of the welding process depends essentially on the chemical and thermophysical properties of the process gas used. Consequently, it is necessary to be able to describe and to analyse its ow with respect to various inuencing variables. However, it is very difcult to realize this during arc welding processes; a poor access is predominant due to the covered areas inside the welding torch and temperatures of up to 20 000 K cause the strong radiation of the arc and electromagnetic elds. This paper deals with experimental and numerical methods for visualization and quantication of process gas ows in arc welding and gives examples for their technical applications. Unlike previous work, the described methods consider the arc as a dynamic element which determinates the gas ow. Advanced Particle Image Velocimetry (PIV) and Schlieren measurement were used for characterization of the ow eld in the direct vicinity of the arc in GTA and GMA welding. Furthermore, a numerical model including magneto-hydrodynamics and turbulence models was used for a detailed visualization of the ow in the free jet and in the hidden interior of the torch. It is based on a commercial CFD code which allows to model complex 3-D geometries of torch and workpiece design. Mixing effects and turbulence model were validated by oxygen measurements in the gas shield.
IIW-Thesaurus keywords: Arc Physics; Arc Welding; Electric arcs; Flow; GMA welding; GMMA welding; GTA welding; Impurities; MAG welding; Measurement; Measuring instruments; MIG welding; Outdoor environment; Oxides; Oxygen; Photography; Research and development; Shielding gases; Simulating; Torches.


1 Introduction
GMA welding is used as a semi-automatic or automatic arc welding process in many applications. In this process the arc is burning between a continuously fed and consumable wire electrode and the workpiece. The shielding gas undertakes a lot of tasks, for example the cooling of the torch, the denition of the arc properties or the protection of the melt from oxidation. In particular, the joining of aluminium, high alloyed steels or titanium requires a cover of shielding gas in order to provide a low parts per million (PPM) concentration of oxygen; otherwise costly rework cannot be avoided. Furthermore, the shielding gas transports metal vapour, dust and fume into the working environment. In summary, the shielding gas ow is so important, that it is essential to analyse its characteristics in detail and in context with the other process components. Analyses of the gas ow in arc welding were done so far by numerical models [1-4]. CFD-simulation provides the possibility of visualization of the ow eld and its employing interactions with the arc process in high resolution, including information about the hot arc and plasma regions and the inside of the welding torch as well. However, the suitability of numerical predictions is mainly limited by the numerical and physical complexity of the model used.


Thus, a model is always a compromise of the geometrical complexity of the analysed welding torch design and the necessary inclusion of arc physics or transient behaviour. In order to obtain a link between the model and the physics observed diagnostic methods to analyse transient phenomena are needed. LDA (Laser Doppler Anemometry) and PIV (Particle Image Velocimetry) are particle based ow measurements [5-7]. Schlieren techniques are widely used to visualize density differences [8, 9]. Most of them have been used already with welding torches, but mostly without an arc in process. That is why the goal of their usage for the gas shield development or investigations in fume extraction of arc welding torches is to enhance these experimental setups for measurements during arc welding. However, numerical and ow measurement methods can visualize the ow characteristics and their causation and maybe allow to reduce the development effort and to increase gain of knowledge as well. Welding experiments, in step with actual practice, are inevitable to prove the efciency of developments. In this paper, the scope and restrictions of numerical and experimental methods for ow characterization at welding arcs will be discussed. Furthermore, it will be shown how they could complement one another.

Doc. IIW-2013, recommended for publication by Commission XII Arc Welding Processes and Production Systems.


Flow measurement and characterization

2.1 Particle Image Velocimetry in arc welding

The Particle Image Velocimetry (PIV) is a non-intrusive optical method, which enables ow investigations with relatively high spatial and temporal resolution by photographing a two-dimensional ow eld twice within a short, well-dened time interval. The ow describing information in the pictures comes from tracer particles, which follow the ow and are two times illuminated by a laser light section. The twin-pictures are taken by one or two cameras, which are synchronized with the laser and positioned orthogonal to the laser light plane. The ow map is calculated by cross correlation between the pixel groups of both pictures and by dividing the two-dimensional displacement of each traceable particle with the time delay between the pictures, (Figure 1). The delay in one picture pair (t) is between 100 and 500 s. The reload of the laser and the saving of both pictures take about 0.3 s. The classical PIV setup is not suited for ow measurements in light intensive arc or plasma processes because of overexposure of the pictures. To circumvent this problem, an extremely short exposure time of 67 ns was used and, additionally, all wavelengths beside the wavelength of the illuminating laser = 532 nm were ltered with a band gap lter of 3 nm half-intensity width. The tracer particles, which are necessary for PIV measurements, are subjected to a variety of requirements such as providing an intensive scattered or reected light for the pictures, following the gas ow immediately and showing appropriate stability in the ow. For the measurements in

arcs or plasma, only solid particles with high melting temperatures can be used, which have a much higher density than the carrier gas. To provide a sufcient capacity of following the ow, the particles should be as small as possible. Furthermore, they should not interact with the welding process by chemical reactions, volume expansion or deection by Lorentz forces. Magnesium oxide particles with 1 m maximum diameter, melting temperature of 2 640 C, density of 3.65 g/cm3 and an extended chemical stability at high temperatures were used. Only a small fraction of particles is melted in the arc and deposited as slag onto the welding bath. The majority of the particles pass the ow eld unmolten by the welding process. The correct dosage of particles into the shield gas ow requires a special apparatus to prevent the powder from bridging. A funnel formed storage vessel with the shield gas inlet at the small diameter, which is mounted on a vibration plate, was used as dosage unit. The shield gas ows upwards through the funnel and carries the particles over, while the vibrations prevent powder bridging. The magnesium oxide powder was dried before at 350 C for one hour. The calculated ow maps of a pulsed GMAW process are shown in Figure 2. The results demonstrate that the ow of the shielding gas is strongly inuenced by the arc. The ow map at the beginning of the pulse is totally different from that of the down slope at the end of the pulse or during the background current.

There are essential differences between the ow maps of single measurement and the average value of 50 single measurements that are triggered to the very same pulse phase time of every 30th pulse. A transient behaviour of the gas shield and large eddies were found in single measurements. Studies demonstrate that not only the design of welding torch and gas nozzle but also the gas ow, the average and peak current, the torch position and welding

Figure 1 Setup and data processing of a Particle Image Velocimetry (PIV)

N 01 02 2012 Vol. 56 WELDING IN THE WORLD

Peer-reviewed Section


Figure 2 PIV ow map of the gas shield by P-GMAW of aluminium at the middle and the end of the pulse, (left) single measurement and (right) averaged values

speed, as well as the wire and the workpiece materials, inuence the gas shield during arc welding.

interference-scheme is easy to picture at a white board or with a (high-speed) camera. The differences in density can come from: mixture of different gases, e.g. air, argon, helium or carbon dioxide, pressure gradients cause by eddies and turbulences and temperatures of the torch and the arc. The Toepler Z Schlieren technique is suitable because of the compact measurement setup, Figure 3. The light source, the mirrors and the aperture determine the results notably. The usage of coloured lters increase the sensitivity but also the blurring. The Schlieren technique is conventionally used without an arc; exceptions are TIG welding and very short short-arcs [10]. In general the self-radiation of the arc superposes the Schlieren lightning. Especially by high welding current

The used PIV measurement system enables ow characterization in the direct vicinity of the arc. There are no particles visible inside the arc. The ow should only have a low vector component orthogonal to the visualization plane. Otherwise, the particles are only visible in one of the two pictures and correlation fails. Beside the ow map it is possible to gather eddies from the rotation of pixel groups. Statements concerning diffusion and the oxygen concentration are not feasible.

2.2 Schlieren techniques

The Schlieren technique is also a non-intrusive measuring method. It is based on the differences in the density of a transparent media that cause changes of the refractive index and a deection of the light beams. The generated

Figure 3 Compact Toepler Z Schlieren measurement setup


Figure 4 Schlieren pictures of TIG Processes with various amounts of shielding gas [9] Comparison of Schlieren picture and simulation results (bottom)

and GMAW with metal vapour radiation cross fades the Schlieren. Figure 4 shows pictures of a TIG process (100 A) with different amounts of shielding gas (Ar50/He50). The increase of the shielding gas ow causes the change to a turbulent ow characteristic. The left picture with 10 l/min shows a laminar ow of the gas shield, only the Schlieren at the arc edge are visible. Additionally, strong Schlieren structures beside the arc are visible in the middle and right picture, which can be addressed to turbulences at the edge of the shielding gas jet and above the workpiece. The transition point from laminar to turbulent ow behaviour can be identied very easily. The advantages of the Schlieren techniques are the low costs and the easy handling of the measurement setup. The practicability of Schlieren for arc welding is limited by the emitted arc radiation. Furthermore, the density gradients are very small inside a laminar gas shield, thus these ows are hard to analyse by the Schlieren technique.

2.3 Measurement of the oxygen concentration in the arc

In the literature, the measurement of the oxygen concentration is documented, above all, for shielding and forming gases, but without an arc. However, these measurements have only limited value because of the intensive ow of the arc and the rise in the diffusion coefcient of the gases along with the temperature. The experimental determination of the oxygen content in the arc is based on the arc pressure measurement setup of [11]. A small pump is in place of the pressure sensor and it extracts a small partial ow through a borehole in the water-cooled copper anode which is analysed by a broadband lambda probe. The measurement is insensitive to magnetic elds and, because of the temperature regulation system integrated in the probe, to temperature uctuations as well. The measuring setup is shown on Figure 5.


Figure 5 Setup for the oxygen measurement

N 01 02 2012 Vol. 56 WELDING IN THE WORLD

Peer-reviewed Section


The measuring method is easy to apply for arc welding with non-consumable electrodes. Predictions were done for the gas shield of GMAW torches by using a tungsten electrode. However, the removed gas volume inuences the ow and the local oxygen concentration. This effect manipulates the results especially in regions of low ow velocity and high concentration gradients in the outer areas. Furthermore, low concentrations of nitrogen monoxide and hydrocarbons can fail the results because of the physical principle of a lambda probe.


2 j

(4) (5)

FL = j B

The arc attachment at the wire and the workpiece can be simplied by using a grid solution of 0.1-0.4 mm. However, the simplication of the arc attachment and the neglect of metal vapour inuence may cause limitations of the model. Contaminations of the gas shield by atmosphere gases results from turbulence and diffusive mixing. A number of different turbulence models, e.g. two equation models (k-epsilon, k-omega, SST), Reynolds stress models (BSL, SSG) and large eddy simulation (DES, SAS, LES), are available in ANSYS CFX. The shear stress transport (SST) model is an easy to apply two-equation-model that combines the advantages of the k-epsilon and the k-omega turbulence models [16]. Numerical studies of different turbulence models and comparisons to measured oxygen concentration distribution at the workpiece demonstrate similar results for SST and Reynolds models but a much lower numerical expense of SST model (Figure 6). The results of the 6-model predict lower values of oxygen as measured in the outer areas. The contamination of the shielding gas by the atmosphere due to turbulences is not represented correctly. Thus it is proven that the geometrical complexity of the torch is too much simplied in most published numerical investigations. Murphy [17] calculated ordinary diffusion coefcients of argon-air-mixtures. His results demonstrate that the diffuvsion coefcient at 10 000 K is approximately 500 times higher than the known coefcient at 300 K. Therefore, the diffusion coefcients DxAr, Air were implemented as functions of the gas temperature. The model can provide different transient boundary conditions. This could be pulsed current and gas ow or a dened change of the torch geometry by mesh motion.

3 Simulation of shielding gas ow

The commercial software ANSYS CFX can be used for the process simulation of arc welding. The software offers high solver stability, good parallelization and advanced physical models for turbulence, radiation and particle tracking as well. For arc simulation the software was enhanced by models of: MHD effects to implement electromagnetism [12] Sheath layers near to the electrodes [12-13] and Diffusion to include mixing and demixing [13-15]. ANSYS ICEM CFD was used for meshing the complex torch design by using a high quality hexahedral mesh, which is recommended for arc calculations with ANSYS CFX. As shown above, the arc determines the ow of shielding gas. Therefore, a simulation of the gas shield quality is insufcient without the inuence and dynamic of the arc. In addition to the momentum, mass, energy and turbulence conservation equations the electric current and the magnetic eld have to be calculated, Equations (1-3).


j = grad A = 0 j B = rot A

(1) (2) (3)

where j is the electric current density,

is the electric conductivity, is the electric potential,

A is the magnetic vector potential,

0 is the magnetic permeability and

B is the magnetic force. The calculation of the magnetic eld is done by solving for the magnetic vector potential. The model can be used for non-axially symmetric arcs as they happen caused by the setting angle of the torch or the environment in welding. Electromagnetism affects the ow by resistive heating QHeat, Equation (4) and by Lorentz force FL, Equation (5).

Figure 6 Oxygen concentration at the workpiece without an arc Comparison between different turbulence models and oxygen measurements


cause increased values of oxygen inside the arc, see Pictures 4 and 6. The predicted oxygen concentrations demonstrate that in case of a small nozzle diameter and a high nozzle-to-workpiece distance the diffusion causes a strong contamination in the arc process. Furthermore the numerical investigations predict a decrease of oxygen concentration in the arc and at the workpiece when the arc is in high-arc current phase. This statement was proven by oxygen measurements.

4 Conclusion
Figure 7 Flow characteristic in range of the gas distributor of a GMAW torch (10 l/min Ar)

This option in ANSYS CFX can be used for an optimization of the nozzle design or for a realization of an oscillating motion of the welding torch. The simulation of the gas ow inside the torch can be done without an arc model. The gas distributer can cause a transient ow with a lot of eddies and turbulences (Figure 7). Especially by small boreholes and a high gas ow the transient ow and turbulences inuence the gas shield below the gas nozzle and in the range of the arc. Figure 8 demonstrates the inuence of the arc on the shielding gas ow. The shielding gas is drawn into the arc and causes a smaller recirculation zone below the contact tube and the gas shield. This effect is more signicant at high welding current and long stick-out length. Additionally high welding current causes a high radial velocity above the workpiece that scatters welding fume. Figure 9 visualizes the effects of too small a shielding gas volume. The increased oxygen concentration is caused by the high diffusion coefcient at high temperatures inside the arc. Figure 10 demonstrates the transient behaviour of an AC TIG process. The calculations show that especially at the end of the pulse or during reversion of polarity the ow velocity inside the arc decreases and the diffusion effects

An effective gas shield is indispensable to guarantee the process stability of welding and to decrease the costs of subsequent works. The development of modern welding torches realizing a better gas shield by the same or reduced gas consumption necessitates modern methods of ow and mixing visualization during the arc welding process and in the hidden regions inside a welding torch. However, these methods would also be very useful for the development and optimization of fume extracting welding torches. The PIV and Schlieren techniques can be used for the visualization of gas ow during arc welding. The advantage of PIV is the high resolution especially by observing pulsed current arc welding and the applicability on GMAW. The Schlieren technique causes much lower costs but is difcult to use at high welding current or in GMAW. Oxygen measurement can be done at a watercooled copper anode using a pump for the separation of a small gas portion towards a broadband lambda probe. This method is unfortunately usable only for TIG or PTA welding. The classication of the gas shield of GMAW torches makes it necessary to substitute the wire by a tungsten electrode (EP). The CFD simulation of the shielding gas ow enables the visualization of the gas ow and gas concentration at the arc region and inside the torch. Based on the available models the inuence and the dynamics of the arc can be


Argon 12 l/min.

a) without arc

b) with a 200 A arc

a) 6 l/min

b) 18 l/min

Figure 8 Inuence of the arc on the proportion of oxygen in the gas ow

Figure 9 Inuence of the ow rate on the proportion of oxygen in the gas ow

N 01 02 2012 Vol. 56 WELDING IN THE WORLD

Peer-reviewed Section



Figure 10 Transient behaviour of arc, plasma and shielding gas ow as well as air molar fraction during AC-TIG process

considered. The numerical simulation allows a high resolution and detailed predictions. Furthermore, design changes during development process can be directly taken over in grid generation and numerical studies. This allows a decrease of development time and costs. The discussed ow visualization methods and oxygen measurement can be used for the validation of the model. The model is used for modelling the gas shield of TIG and GMAW processes neglecting the weld pool depression and the transient effects of wire feed and melting. Analyses of their inuence could be content of future works. However, welding experiments are still necessary to prove the efciency of developments.

[1] Cooper P., Godbole A. and Norrish J.: Modelling and simulation of gas ows in arc welding Implications for shielding efciency and fume extraction, IIW Doc. XII1932-07, 2007. [2] Speiseder M. and Lang A.: Optimierung des MIGSchweiprozesses mit Hilfe der numerischen Simulation und PIV-Messung, Optimization of MIG-welding by numerical simulation and PIV, Dresdner Fgetechnisches Kolloquium, Technische Universitt Dresden, 2006 (in German). [3] Lowke J.J. and Tanaka M.: LTE-diffusion approximation for arc calculations, Journal of Physics D: Applied Physics, 2006, vol. 39, no. 16, pp. 3634-3643.

This paper was supported by the AiF (KF 15.871). The described results are part of the AiF-DFG-cluster project: arc welding physics and application development. We also acknowledge the contributions of Prof. J. Schein, Dr. G. Wilhelm, E. Siewert and M. Haessler by prototyping the Schlieren measurement setup.

[4] Lowke J.J., Morrow R. and Haidar, J.: A simplied unied theory of arcs and their electrodes. Journal of Physics D: Applied Physics, 1997, vol. 30, no. 14, pp. 2033-2042. [5] Zschetzsche J.: Diagnostik von Schutzgasschweiprozessen, Diagnostic of GMAW-processes, Dissertation Technische Universitt Dresden, Verlag der Wissenschaften, Dresden, 2007 (in German).


[6] Zosel J., Guth U., Doerner K. and Zschetzsche J.: Strmungsmessung an Lichtbgen und Plasmen zur Materialbearbeitung mit LDA und PIV, Flow measurements in arcs and machining plasmas by LDA and PIV, In D. Dopheide: Lasermethoden in der Strmungsmesstechnik, ISBN 3-00-011903-5, Braunschweig, 2007 (in German). [7] Schnick M., Fssel U., Zschetzsche J., Guth U. and Zosel J.: Measurement and simulation of plasma and shielding gas ows in welding arcs, 58th IIW Annual Assembly and International Conference of the International Institute of Welding, Prague, Welding in the World, 2005, vol. 49, Special Issue, p. 413. [8] Settles G.S.: Schlieren & Shadowgraph Techniques, Springer Verlag, 2006. [9] Haessler M., Fssel U., Siewert E. and Schnick M.: Visualisierung von Strmungen am Lichtbogen durch Schlierenmesstechnik, Visualization of gas ows in the arc region by Schlieren technique, Technische Universitt Dresden, Dresden, 2009 (in German). [10] Mller F.: Visualization of gas ows for research and industry use, In: BIAS Bulletin, 2009, vol. 01, no. 2. [11] Lin M.L. and Eagar T. W.: Pressures produced by gas tungsten arcs, Metallurgical Transactions B, September 1986, vol. 17, pp. 601-607. [12] Spille-Kohof A.: Lichtbogenprozesssimulation mit ANSYS CFX, Numerical simulation of arc welding processes, Dresdner Fgetechnisches Kolloquium, Selbstverlag der Technischen Universitt Dresden, Dresden, 2009, ISBN 978-3-86780-115-7 (in German). [13] Schnick M., Fssel U. and Spille-Kohoff A.: Numerical investigations of the inuence of design parameters, gas composition and electric current in Plasma Arc Welding (PAW), Doc. IIW-1197, Welding in the World, 2010, vol. 54, no. 3/4, pp. R87-R96.

[14] Schnick M., Fssel U., Hssler M., Hertel M., SpilleKohoff A.: Effects of metal vapor in GMA welding Numerical investigations in diffusion effects of metal vapor and its inuences on arc behavior, Proceedings of 8th international Symposium of the Japanese Welding Society and IIW Intermediate Meeting, Kyoto, 2009. [15] Spille-Kohoff A.: Implementierung von Entmischung in ANSYS CFX, Implementation of demixing, CFX Berlin Software GmbH, Berlin, 2007 (in German). [16] Menter F.R.: Two-equation eddy-viscosity turbulence models for engineering applications, AIAA-Journal, 1994, vol. 32, no. 8, pp. 1598-1605. [17] Murphy A.B.: Transport coefcients of air, argon-air, nitrogen-air, and oxygen-air plasmas, Plasma Chemistry and Plasma Processing, 1995, vol. 15, no. 2, pp. 279307.


About the authors

Dr.-Ing. Michael SCHNICK (, Dipl.-Ing. Michael DREHER (, Dr.-Ing. Jrg ZSCHETZSCHE (joerg. and Prof. Dr.-Ing. habil. Uwe FSSEL (uwe.fuessel@ are all with Technische Universitt Dresden, Dresden (Germany). Dr. rer. nat. Andreas SPILLE-KOHOFF ( is with CFX Berlin Software GmbH, Berlin (Germany).

N 01 02 2012 Vol. 56 WELDING IN THE WORLD

Peer-reviewed Section