Bobbin tool friction stir welding (BT-FSW) is a special variant of solid-state conventional friction stir welding (CFSW). It has the same welding principle as CFSW but it has been identified that it can weld thicker materials with full penetration in a single pass. Thus, it is the latest technology to weld thick aluminum alloys since joining thick aluminum alloys with fusion welding, riveting, or adhesive joining is not easy due to the complex inherent properties of aluminum such property includes low melting temperature, a wide range of solidification temperature, and higher oxygen affinity. However, Bobbin tool friction stir welding experiences problems of weld initiation, void defect, and lack of scientifically identified optimal process parameters. Thus, this research aims to investigate the effect of welding temperature and optimize process parameters (tool rotation speed, tool traverse speed, tool pin diameter, and dwell time) of BT-FSW to enhance the mechanical strength (hardness and tensile property) of 10 mm thick aluminum alloy 6061-T6 with a butt joint configuration. In this research, the BT-FSW operation was performed by reconfiguring a vertical CNC milling machine and using a tool which was locally produced by using a special attachment of grinding tool on a lathe machine and then an experimental approach with an L9 orthogonal array was used to investigate tensile strength and hardness of BT-FSW. Literature review and preliminary tests were used to identify process parameters and corresponding levels. The Rockwell hardness and ultimate tensile strength result were modeled with the corresponding process parameters using an artificial neural network (ANN) applying optimized neural network architecture and process parameters percentage contribution (importance) was identified. Then this model was taken by genetic algorithm (GA) to determine the combination of process parameters that yields an optimal hardness and tensile strength. Moreover, a coupled Eulerian-Lagrangian (CEL) thermomechanical model was used to determine maximum temperature and investigate the thermal simulation of BT-FSW. The feasible optimal process parameter of a combined artificial neural network and a genetic algorithm (ANN-GA) was identified. Finally, three confirmatory tests were investigated, and they agree with the ANN-GA optimal result with only 2.137% average error. Thus, it has been concluded that the experimental results are within the range of acceptance. Keywords: Bobbin tool friction stir welding, combined artificial neural network and genetic algorithm, aluminum alloy 6061-T6, coupled Eulerian-Lagrangian.