Material Flow in Friction Stir Welding: Experiment and Fluid Mechanics based Process Model

Friction Stir Welding (FSW) is a recently developed welding technique which is especially well suited for joining high strength aluminum alloys. FSW is a solid state joining process producing high quality, defect-free joints. The shape of the tool promotes high hydrostatic pressure along the joint line, causing consolidation of the material plasticized due to heat generation. Although significant effort has been expended in putting FSW to use in full-scale production of such products as ferry boats, rocket fuel and oxidizer tanks, the operating mechanisms and in particular, the material flow during friction stir welding is not fully characterized. In the work described herein, the material flow is visualized using a Marker Insert Technique. Additionally, a FSW process model is developed based on the principles of fluid flow. At the current stage, the model is two-dimensional and steady state. Some details will be discussed later.

Marker Insert Technique

The material flow in FSW is visualized using a Marker Insert Technique. The idea behind this method is to place marker inserts in the weld path in a manner that the marker can be detected after welding. Therefore, this marker needs to have similar mechanical properties as the base material (AA2195-T8). Furthermore, the marker material should have significantly different metallographic etching characteristics. The marker material used in this research was 1.8 mm thick AA 5454-H32 sheet. It contains no copper whereas AA 2195 is rich in copper, which makes it easy to distinguish the two alloys when they are etched with Keller's reagent.

t hspace=15 align=

Schematic drawing of marker insert placement. The upper drawing is a top view, and the lower is a view from the side. The dark blue, green and orange markers are on the retreating side, the light blue, yellow and red markers are on the advancing side.

Inserts were placed on the advancing and retreating side, respectively, at three different heights: top, middle and bottom. Inserts at the same height were staggered so that mutual mixing could be eliminated. The length of the markers was such that some of the marker, remote from the weld line, extended beyond the TMZ and was not deformed by the welding process. Markers were placed by milling narrow slots into the faying surface and press-fitting the markers into the slots. For the markers that were placed in the middle third of the plate, a 2195 shim of similar size to the marker was press fit into the upper third of the plate above the marker. Marker flow was elucidated by milling off successive slices, 0.25 mm thick, from the top surface of the weld. After each cut the surface was etched with Keller's reagent and digital images of the region surrounding each marker were obtained. Due to the etching process the base material (which contains copper) turns dark whereas the 5454 stays shiny. The pixel positions of the 5454 markers were extracted from the digital images of each marker (a total of six) at each level (up to 32 per weld). Combining the data from all six markers, a three-dimensional flow visualization similar to a CT scan was obtained. The 3-D plot gives the final position of an originally thin layer that was perpendicular to the welding direction.

Representative Images of the Markers

The welds made at the University of South Carolina had some general material flow pattern in common. The flow is not symmetric about the weld centerline, which describes the original interface of the two plates. The bulk marker material was moved to a final position behind its original position and only a small amount of the material on the advancing side was moved to a final position in front of its original position. Note, that no material was transported further backwards than one pin diameter behind its original position.

Detected marker positions at the top (left image) and at the middle (right image) of the weld.

In the black and white images, the 5454 markers of advancing and retreating side are shown at two representative sections in the weld height; at the top on the left and at the middle at the right side. The pictures of advancing and retreating side markers have been combined to single images, in which the advancing side is the left hand side in both images The images show the final position of the markers in a cold weld, which was performed at low rotational and medium translational speed. Material from the retreating side was transported along with the rotating tool to the advancing side and parts of it even to a position in front of its original position. At the very top of the weld (left image), advancing side marker movement was only in welding direction. No advancing side marker was moved to a position behind its original position closed to the shoulder. However, the image of a marker at one height does not give any details about the vertical flow that is examined later in this paper. The right image shows advancing and retreating side markers at the middle of the weld thickness. Advancing side material in the middle of the weld thickness was mainly transported backwards, dissimilar to the material movement at top levels. The connection of advancing and retreating side behind the pin is not necessarily located at the centerline. Note that material transport backwards is smaller at the top than at the middle of the weld height.

Vertical Mixing and Full 3-D View

Since the material flow pattern changes throughout the weld height a look at a vertical plane in welding direction gives information about the vertical material movement in FSW.

Projection in welding direction. The Markers cover the total weld height and width. During welding the markers are mixed vertically (right image).

As can be seen, material inside the pin diameter (dashed lines) is pushed downwards on the advancing side and moved upwards on the retreating side. This phenomenon is referred as a vertical circular motion. It is initially caused by the movement at the top where retreating side material is transported to the advancing side due to the rotating tool shoulder. Therefore, advancing side material is pushed downwards in the weld when it passes the pin. On the retreating side the material is moving upwards since flow is restricted in x-y-z-direction by the cold base material, the surrounding material flow and the backing plate. Material at the bottom is involved in the vertical circular motion to a smaller extent.

3-D plot of the markers after welding. View from behind in welding direction. Tool rotation is clockwise.

The 3-D plots show a continuos marker after welding. At upper levels material transport is due to the shoulder rotation where material is moved from the retreating side to the advancing side behind the pin. Parts of the material from the top of the advancing side is transported in the welding direction to a final position in front of its original. The backward flow of material from the top of the weld is different from the backward flow of material from lower levels. Material originally from the top did not move as far backwards as material from middle heights. Instead, it is dragged along with the translational motion of the pin sometimes even to positions in front (but still below) its original position. The minimum deformation against the welding direction at the top of the weld occurs where the vertical circular material transport interacts with the material flow at the top caused by the rotating shoulder. The two perpendicular circular motions cause the observed flow pattern. The marker insert technique gives insight in the thermo-mechanically affected zone by reconstruction of the piecewise cut-off welded zone. As a result, full three-dimensional plots of the deformed markers provide a good qualitative characterization of the material flow of friction stir welds. However, this technique does not reconstruct the actual flow path of the material to its final position. The material transport in FSW is a result of the two tool motions - the translational and rotational tool movement. Most material is moved around the pin to final positions behind its original position. No material is transported across the centerline at the leading edge of the pin. Material at the top of the weld is moved with the tool rotation from the retreating to the advancing side. At the top of the advancing side material is transported in welding direction. Both major motions cause vertical mixing within the pin diameter.

Fluid Mechanics based FSW Process Model

In the work described herein, friction stir welding is simulated with the commercially available Computational Fluid Dynamics (CFD) tool FLUENT. As a first approach, the process is modeled as a two-dimensional, steady state, laminar fluid flow past a rotating cylinder. A flowing fluid with a rotating tool is equivalent to a stationary fluid with a rotating and traversing tool. The former case, however, is steady state. The cylinder, which rotates at a constant angular velocity, represents only the pin of the FSW tool. The base material, aluminum alloy 2195-T8, is treated as fluid at elevated temperatures. A non-newtonian power-law viscosity function, which is calculated based on experimental flow stress data, describes the material flow properties. The viscosity is temperature and strain rate dependent. Other material properties, like the thermal conductivity and the specific heat, are temperature dependent functions, only. Heat is generated in the fluid due to viscous dissipation. Furthermore, a constant temperature is assigned to the rotating pin to simplify the problem setup. The 2-D problem description neglects the shoulder of the tool and therefore, the material flow as well as the energy input do not completely represent FSW. Hence, the temperature distribution in the 2-D simulation is different from real FSW. Cases with different rotational pin velocities and pin temperatures were studied. However, only the simulation of the cold weld is shown here. Main focus was on the flow pattern, which is compared to the experimentally observed material flow. It turned out that the simple CFD model is able to predict the material flow in FSW, at least for a cold weld, which produces a low amount of vertical mixing.

Predicted Material Flow in FSW

The stream function, which is constant along a streamline, is a relationship between the streamlines and the statement of conservation of mass. Streamlines are lines that are tangent to the velocity vectors and they are equivalent to the path lines in steady-state fluid flow. In the present case, straight horizontal streamlines describe material, which flows at a constant uniform velocity. In other words, the material shows the same behavior as a rigid material without internal deformation.

Left: Streamlines around the rotating pin; Right: Final Position after passing the pin

The streamlines in the flow domain suggest that material, originally within the pin diameter, is transported around the pin only in rotation direction of the pin. No material within the pin diameter is passing the pin on the advancing side whereas all the material is transported around the retreating side. Furthermore, streamlines downstream end up at the same distance to the centerline as they were upstream. Since the flow is laminar the streamlines are not crossing each other. The material flow path past the tool, predicted by the model, is not intuitive. Knowing only the final position of the markers in the weld the flow around both sides of the pin is conceivable as well. However, only if the material within the pin diameter flows around one side, e.g. the retreating side, the experimentally observed flow pattern can be matched. To compare the material flow prediction with the experimentally observed flow pattern, the final position in the weld must be determined. The final position in the weld describes the position of particles relative to each other after welding, which were vertically lined up in the in the path of the tool before welding. If particles are released in the flow field upstream at one line in transverse direction, the relative position of those particles to each other downstream at any time describes the final position in the weld. Both, the original position upstream as well as the final position downstream are plotted. The simulation shows that all of the material within the pin diameter is transported to a position behind its original. The last plot shows the marker at their final position at middle height in a cold weld (blue is from the advancing side and red from the retreating side) as well as the predicted final position in the 2-D simulation. The rotation direction of the pin (black circle) is clockwise. The markers are shown in their original and final position. The 2-D simulation is able to predict the material flow in a cold weld at middle heights. Since the amount of vertical mixing is very small in a cold weld, the material flow is nearly two-dimensional in a sufficient vertical distance to the tool shoulder.

Comparison of predicted (black squares) and experimentally (red and blue marker) observed flow pattern in FSW. The markers of advancing (blue) and retreating (red) side are displayed in the undeformed and deformed stage.

The 2-D model cannot predict the effect of the shoulder on the flow field and therefore, the simulation approach is considered only as a step towards a three-dimensional model, representing the real tool geometry. Other fields of studies are the tool-fluid interaction as well as the heat and work input.


Future Work

Experiment Simulation

Student: Dipl.-Ing. Tilman U. Seidel
Advisor: Dr. Anthony P. Reynolds