Mechanical Response: Experiment and Modeling

The potential of the Friction Stir Welding (FSW) process is easily observed in the creation of defect free welds in almost all of the aerospace aluminum alloys. The success and applicability of the process, however, will depend on the performance of the welds compared to other joining processes. Experimental testing and numerical modeling are necessary for the determination of the mechanical response of friction stir welds, both on a global and local level, and vital to the development and optimization of the FSW process.

The goal of the experimental testing of friction stir welds is to obtain the data necessary to begin understanding the effects of the FSW process.

Figure 1 Tension tests are used to examine the mechanical response of friction stir welds on a global level. Transverse tension specimens are tested in a 100kN servo-hydraulic test machine equipped with a data acquisition system (Figure 1). Strain is measured using an extensometer with a 25.4mm gage length. The stress-strain response is compared to the base material to provide an overall measure of the weld "efficiency".

Figure 2 The friction stir weld is a heterogeneous material (Figure 2) however, and the global response provides no information about the local properties associated with characteristic weld regions (i.e. heat affected zone- HAZ, nugget or dynamically recrystallized zone- DRZ).

Local material behavior is determined using a combined Digital Image Correlation (DIC) / tensile test. DIC is an image analysis tool used for automated surface displacement measurements. The test involves recording digital images of the weld at specified global stress/strain levels. The "deformed" images are then correlated with the initial "un-deformed" image via the DIC software. A random speckle pattern (black and white spray paint) applied to the through-thickness surface provides the surface features for correlation. The resulting displacement fields are then post processed to determine full field strain data at each specified global stress/strain level.

Figure 3 The strain "history" may be used to identify the presence of defects, track strain localizations, and even identify failure locations (Figure 3). This information can be linked to the weld microstructure to isolate problem areas. In addition to a visual representation of local material behavior, the material properties may be determined. This is accomplished via the full field strain data and the assumption that the weld is a composite material undergoing nominally iso-stress loading, the stress-strain response at virtually any position in the weld may be determined (Figure 4). Once the stress-strain curve is known, a number of material properties are available. By constructing several stress-strain curves at various positions within the weld, local material properties may be plotted as a function of position to produce local property maps (Figure 5).

The goal of numerical modeling is to develop a 2-D model to accurately predict the post weld mechanical response. A working model will provide a tool, which can be used to examine the role of microstructural regions and defects in the mechanical response; provide detailed information on the local stress state; and ultimately serve as input to a 3-D structural response model.

Figure 4

Figure 5

Development of the 2-D Finite Element Model (FEM) is approached from the standpoint that the FSW is a composite material. Due to the observed variation in material properties from the experimental data, the number of materials used to describe the FSW may be large. However, for the preliminary model the materials are limited to Base metal, a single representative HAZ, and single representative DRZ. Future work will include the use of more materials [i.e. thermo-mechanically affected zone (TMZ), advancing and retreating side HAZ, etc..]. Model geometry and meshing are performed using MSC-Partran and the analysis is run using ABAQUS finite element software. The Base, HAZ and DRZ materials are defined using the experimental data. The distribution of materials within the model is based on the actual weld microstructure (see Figure 2). Results from the preliminary modeling efforts on a 2XXX series aluminum alloy are shown in Figure 6 along with the corresponding experimental results.

Figure 6 Once the model is able to accurately predict the mechanical response we will be able to gain some insight into the stress state in the area of failure, examine the effect of altering the amounts of the various weld materials (HAZ, DRZ, etc.) and even demonstrate the role of defects. This information in addition to the experimental data will be instrumental in the development and optimization of the Friction Stir Welding Process.

Student: William D. Lockwood
Advisor: Dr. Anthony P. Reynolds