Welding is one of the most critical operations for the construction of reliable metal structures in everything from ships to reactor vessels. When welds fail, often the entire structure fails, and expectations on weld quality have never been higher.
Any process that uses a localised heat source, such as welding, is likely to result in some distortion. The welding process of very thick metal components is not inherently stable and is barely controllable without external forces.
In high-power laser beam welding, a small amount of metal in the region of the highest laser intensity vaporises. This penetration welding creates a vertical cavity in the workpiece that is known as a ‘keyhole’. In this process, the laser beam not only melts the metal, but also produces vapour.
The dissipating vapour exerts pressure on the molten metal and partially displaces it. The material, meanwhile, continues to melt. The result is a deep, narrow, vapour-filled hole, or keyhole, which is surrounded by molten metal.
As the laser beam advances along the weld joint, the keyhole moves with it through the workpiece. The molten metal flows around the keyhole and solidifies in its trail. This produces a deep, narrow weld with a uniform internal structure.
Well-known issues in deep-penetration welding of aluminium are the highly dynamic behaviour of the melt due to its low viscosity. Combined with high heat conductivity, the resulting weld pool is very wide.
The weld surface becomes unstable with the result being spattering and the ejection of droplets from the weld metal that results in underfills, undercuts, craters, blowholes or blowouts - all of which can have a detrimental effect on the weld’s mechanical properties.
If material is missing, there is often the need for post-treatment with arc welding to fill in the missing material or make the weld more visually appealing, which is an indicator of surface quality. In addition, a smooth weld surface becomes very important in places such as the food industry where rough surfaces could harbour bacteria populations.
One side effect of an uncontrolled welding process is droplets that accelerate from the weld bead. These droplets make the process ‘dirty’ and lead to a lack of material after the weld has cooled down. Second, the Marangoni effect leads to a non-uniform weld, which can be a reason for stresses and/or distortion of the workpiece.
Part of the weld pool is moving under surface tension and electromagnetic forces, thus inducing a non-uniform distribution of the material and different solidification rates in different parts of the weld pool. Once the bead solidifies, it is likely constituted by different materials because of non-uniform distribution and cooling times.
Decelerating the melt
Is it possible to counteract such effects? Under a grant from the German Research Foundation, the Federal Institute for Materials Research and Testing (BAM) is investigating various methods to control and reduce them. In this particular case, the BAM researchers* are applying a stationary magnetic field to the laser welding process.
With the help of Comsol Multiphysics, the team determined the distribution of the magnetic field required to improve the uniformity of the weld. A particular goal was to reduce the impact of the Marangoni effect. At the surface, there is a very high temperature at the point where the laser beam impacts the metal, and the temperature drops off rapidly with distance away from the weld.
The resulting high temperature gradients cause a flow of metal directed from the middle of the weld pool towards the outer boundary due to the temperature-dependent surface tension (the Marangoni effect). In order to create a perfect weld, this flow needs to be suppressed so that the energy goes into the depth of the pool rather than spreading out on the surface.
Consider that a perfect weld would have side walls that are parallel, with solidification taking place at all depths at the same time. An actual weld without the application of external forces has more of a wine glass shape, with a strong curvature of the solidification front. This leads to heavy stresses in the workpiece and relatively large distortions after it cools.
However, when a static magnetic field is applied perpendicular to the welding direction, the weld takes on a more homogeneous shape that starts to resemble a V, which is closer to the desired form. This ability to change the weld shape is due to the Hartmann effect. Specifically, for an electrically conducting liquid such as a molten metal, a magnetic field induces electric currents that create a Lorentz force field with a component directed against the original melt flow direction.
To model this effect, the BAM team simulated the heat transfer, fluid dynamics, and electromagnetic in 3D using Comsol’s CFD and AC/DC modules. First, the electromagnetic field is modelled to calculate the Lorentz forces; these results are then used as a volume force to calculate the velocity and pressure of the turbulent flow in the weld pool.
This allows the team to solve for the heat transfer where the velocity field is taken from the previous turbulent flow simulation. Temperature, of course, influences the material properties, so it was necessary to go back and recalculate the Lorentz forces, which also depend on the velocity of the flow. This looping continued until the simulation reached the desired accuracy in a steady state condition where the solution is self-consistent; that is, it satisfies all the physics involved.
To verify the model, actual welds, done with and without magnets, were cut and the macrosections polished. The simulation results were then superimposed, which show good agreement. This welding process is extremely complex and, thanks to the Comsol Multiphysics applications, the team managed to achieve accurate results.
According to the BAM researchers, Comsol’s advantages are to be found in a combination of easy handling, very comfortable geometry building and meshing, and the ability of using pre-defined multiphysics modules, while all the time having the option for manual tuning and case-dependent modification.
These include, for instance, temperature-dependent material properties coming from experimental data points or analytical expressions, using source terms for the velocity modelling in the solid phase, inclusion of gravity effects, and inclusion of latent heat of fusion. All of these can easily be taken into account for the calculation.
The team was also pleased with the software’s ability to make easily available quantities originated from all the physics. For instance, it took just one click to let the fluid flow physics know that the volume force acting in the weld pool was the Lorentz forces. This is just an example that can be extended to all the current and future multiphysics coupling that the team may need.
Thanks to the Comsol Multiphysics simulation, the BAM researchers were able to determine the underlying effects and understand how to counteract them. The next step is to learn how to put this knowledge into practice at a large scale. Those magnetic fields that improve the quality of this welding process have been identified, and the team plans to perform further experiments to redefine the whole welding process.
*The BAM researchers involved in the work described above were: Marcel Bachmann, Vjaceslav Avilov, Andrey Gumenyukand and Michael Rethmeier