The optimization of process parameters of Friction stir welding have not gone in to the details of microstructural changes with variations in the parameters. The single consumable in the FSW process is the tool. The reported studies for the optimization of parameters for obtaining better results often suggested complex shape features for the tool which obviously increases the process cost. Gathering a suitable combination of process parameters with simple tool geometry, hence would have contributed much to the research in the area of FSW process. The thesis elucidates the experimental efforts to propose an optimum combination of parameters with simple tool geometry for FSW at higher linear speeds. Two precipitation hardenable aluminium alloys were selected as the materials for study: AA2024 and AA6061. Both are light weight and possess high strength to iv weight ratio. AA2024 alloy is popular in the aviation manufacturing and AA6061 is a pioneer material in the structural applications. For the precipitation hardened alloys the concentration and distribution of the strengthening precipitates is most influential for the strength of the alloys. Hence the distribution and dissolution of the strengthening particles under the influence of the process parameters are crucial in the study of friction stir welds. Taguchi analysis which is very useful for the identification of the control factors to obtain optimum results for the process was used to design the experiments and further analysis. The effect of process parameters on the microstructural changes in the weld region and on the defect formation was also investigated. The experiments were conducted with, perhaps the highest welding speed reported, so far. Through the analysis, optimum combinations of the parameters were suggested for each material at a high speed friction stir window. An analytical model was proposed from the basic theory and parameters for suggesting the process parameters and tool parameters for various aluminium alloys. The computed results of the model were validated by comparing with the reported results.