This work deals with two crucial issues related to the numerical simulation of Friction Stir Welding (FSW) processes. The first issue is the identification of microstructural parameters for precipitates dissolution in precipitation hardenable aluminium alloys. Here the main goal is to get the effective activation energy and master curve using Neural Networks (NN). Based on the model of kinetics of dissolution of precipitates for hardenable aluminium alloys given by Myhr & Grong (1991), a new parametrization of the master curve is proposed. The novel methodology, which has been applied to different aluminium alloys, such as the AA 2014 T6, AA 6005 A T6 and AA 7449 T79, tends to avoid both the overestimation and underestimation of the relative volume faction of precipitates at later and early stages, respectively, shown by the Myhr & Grong (1991) and Shercliff et al. (2005) models. Numerical results for the effective activation energy and the new parametrized master curve are given. A good comparison with experimental results available is shown. The second issue addressed in this work is the process parameters optimization in FSW. Here the main goal is to identify and to optimize the key material process parameters, on the basis of a suitable measure of the weld quality. The optimization of the process parameters is performed using Genetic Algorithms (GA). The advancing and rotating tool velocities have been selected as the key process parameters to be optimized. The measure of the quality of the weld has been defined in terms of the maximum hardness drop at the cross section under the tool. The goal is to find optimum values of the advancing and rotating speed such that the maximum hardness drop at the cross section under the pin (measured either locally or as an averaged value) is minimized. The computational model has been applied to find optimum advancing and rotating velocities for the FSW of an AA 7449 T79.