PROBLEM - The bobbin friction stir welding (BFSW) process has potential benefits for welding thin sheet aluminium alloy. The main benefits of friction stir welding over conventional thermal welding processes are minimisation of energy usage, no need for consumables, potential for good weld quality without porosity, no fumes, minimal adverse environmental effects (green), minimal waste (lean), and reduced threats to personal health and safety. The BFSW process has further advantages over conventional friction stir welding (CFSW) in the reduction of welding forces, faster welding, and less fixturing. It is especially attractive to industries that join thin sheet material, e.g. boat-building. The industrial need for this project arose from the desire to apply the technology at a ship manufacturing company, INCAT located in Hobart, Tasmania, Australia. However there are peculiar difficulties with the specific grade of material used in this industry, namely thin sheet aluminium Al6082-T6. Early efforts with a portable friction stir welding machine identified the process to have low repeatability and reproducibility, i.e. process-instability. There are a large number of process variables and situational factors that affect weld quality, and many of these are covert. This is also the reason for divergent recommendations in the literature for process settings. PURPOSE - The main purpose of this research was to identify covert variables and better understand their potentially adverse effects on weld quality. Therefore, this thesis investigated the hidden variables and their interactions. Developing this knowledge is a necessity for making reliable and repeatable welds for industrial application. APPROACH - An explorative approach that focused on the functional perspective was taken. An extensive empirical testing programme was undertaken to identify the variables and their effects. In the process a force platform and BFSW tools were designed and built. A variety of machine platforms were used, namely portable friction stir welding, manual milling machine and computer numerical control (CNC) milling machine. The trials were grouped into 14 test plans. These are tool shoulder gap, spindle and travel speed, tool features, machines, tool fixation, machinery, welding direction, plate size (width and dimension), support insulation, tool materials, substrate properties and fixation. For the welded plates besides visual inspection of the weld, current, force, and temperature were measured. The Fourier transform was used to analyse the frequency response of machines. Also the welded samples were tested to the maritime standards of Det Norske Veritas (DNV). A number of relationships of causality were identified whereby certain variables affected weld quality. A model was developed to represent the proposed causality using the IDEF0 systems engineering method. FINDINGS - From these trials six main variables have been identified. These are tool features, spindle speed, travel speed, shoulder gap compression, machine variability, tool and substrate fixation. A rigid system is required for a consistent weld results. Under this condition, full pin features (threads and flats) need to be used to balance the adverse effects of individual features. It has been shown that fabricated bobbin tools with sharp edges can cause cuts and digging thus this feature should be avoided. Additionally, the substrate should have continuous interaction with the tool so the shoulder interference needs to be fixed and well-controlled. It is found that the compression generated by the shoulder towards the substrate helps material grabbing for better tool-substrate interaction. It is also shown that tool entry causes ejection of material and hence an enduring mass deficit, which manifests as a characteristic tunnel defect. The new explanation of the formation, origin and location of this defect has been explained. Material transportation mechanisms within the weld have been elucidated. It is also found that the role of the travel speed is not only to control heat generation but also for replacing the deficit material. Additionally, heat supplied to the weld depends not only on thickness, but also the width of the plate. Different types of machine cause an interaction in the material flow through their controller strategies. Jerking motion can occur at a slow travel speed, which also alters the way material is being transported. The Fourier transform (FFT) has been used to identify the characteristics of good and bad BFSW welds. This has the potential to be expanded for real-time process control. IMPLICATIONS - Tool deflection and positioning, material flow and availability are identified as affecting weld quality through stated mechanisms. The impact is even more severe when involving thin-plate aluminium. For the industry to successfully adopt this technology the process typically needs tight control of shoulder gap, tool strength and stiffness, feature fabrication, substrate and tool fixation. Additionally spindle and travel speed need to be adjusted not only based on the type of materials and thickness, but also the width, type of machine and method of tool entry. ORIGINALITY - New data are presented, which lead to new insights into the welding mechanics, production settings, material transportation and weld defects for BFSW on thin sheet material. The conventional idea that the welding tool has a semi-steady interaction with the substrate is not supported. Instead the interaction is highly dynamic, and this materially affects the weld-quality, especially in the difficult-to-weld material under examination. Factors such as shoulder gap, tool and substrate fixation compliance and machine types emerge as variables that need to be given attention in the selection of process parameters. The causal relationships have been represented in a conceptual model using an IDEF0 system approach. This study has made several original contributions to the body of knowledge. First is the identification of previously hidden variables that effect weld formation for the fixed gap BFSW process. The second contribution is a new way of understanding the material transportation mechanics within the weld. This includes the flow around the pin in the plane of the weld, the vertical transportation of material up the pin, the formation of turbulent-like knit lines at the advancing side, and the formation of tunnel defects. Also included here is a new understanding of how material deficit arises at tool entry and exit, and from flash/chips, and how this contributes to the tunnel weld defect. In addition, new understandings of the role of feed rate have been identified. Related to the material transportation, the work has also identified the importance of an interference fit between the substrate and tool. A third contribution is the identification of the dynamic interaction between tool and substrate. This identifies the important role rigidity plays. Associated with this is the identification of frequency characteristics of the motors under load. The fourth contribution is identification of the specific process settings for the difficult-to-weld material of AL6082-T6. The fifth contribution is the development of a novel method of fabricating bobbin friction stir welding tools as embodied in a patent application.