The objective of this research was to evaluate the capabilities of ALEGRA, a Sandia National Labs hydrocode, and Forge3, a Transvalor S.A. product, to accurately model the Friction Stir Welding Process. ALEGRA and Forge3 are discussed in light of the inherent challenges of modeling Friction Stir Welding, and a rotational boundary condition is added to ALEGRA. Results are presented from Friction Stir Welding modeling outputs from both ALEGRA and Forge3. ALEGRA is shown to be incapable of modeling the Friction Stir Welding process, in large part due to its focus on shock propagation, which causes extremely small time steps. Forge3 is shown capable of modeling of the FSW plunge process in a transient manner, but overestimates the temperature profiles 90% to 100% in comparison to experimentally measured values. It appears that the adiabatic boundary condition is the source of much of the error. It is recommended that future work focus on improving estimates of the boundary conditions utilized in the Forge3 model.

A Lagrangian finite-element model of the plunge phase of the friction stir welding process was developed to better understand the plunge. The effects of both modeling and experimental parameters were explored. Experimental friction stir plunges were made in AA 7075-T6 at a plunge rate of 0.724 mm/s with spindle speeds ranging from 400 to 800 rpm. Comparable plunges were modeled in Forge2005. Various simulation parameters were explored to assess the effect on temperature prediction. These included the heat transfer coefficient between the tool and workpiece (from 0 to 2000 W/m-K), mesh size (node counts from 1,200 to 8,000), and material model (five different constitutive relationships). Simulated and measured workpiece temperatures were compared to evaluate model quality. As spindle speed increases, there is a statistically significant increase in measured temperature. However, over the range of spindle speeds studied, this difference is only about 10% of the measured temperature increase. Both the model and the simulation show a similar influence of spindle speed on temperature. The tool-workpiece heat transfer coefficient has a minor influence (<25% temperature change) on simulated peak temperature. Mesh size has a moderate influence (<40% temperature change) on simulated peak temperature, but a mesh size of 3000 nodes is sufficient. The material model has a high influence (>60% temperature change) on simulated peak temperature. Overall, the simulated temperature rise error was reduced from 300% to 50%. It is believed that this can be best improved in the future by developing improved material models.