A wear characterization study was performed to determine the useful lifetime of polycrystalline cubic boron nitride (PCBN) tooling for the friction stir welding (FSW) of stainless steel samples in support of a nuclear repair welding research and development program. In situ and ex situ laser profilometry were utilized as primary methods of monitoring tool geometry degradation, and volumetric defects were detected through both nondestructive and destructive techniques, as repeated welds of a standard sample configuration were produced. These combined methods of characterization allowed for the successful correlation of defect formation with tool condition. Additionally, the spectral content of weld forces was examined to search for indications of evolving material flow conditions, caused by significant tool wear, that would result in the formation of defects; this analysis established the basis for a system that would automatically detect these conditions. To demonstrate this type of system, an artificial neural network was trained and evaluated, and a 95.2% classification rate of defined defect states in validation was achieved. This performance constituted a successful demonstration of in-process monitoring of tool wear and weld quality in FSW of a high melting temperature, high hardness material, with implications for remote monitoring capabilities in the specific application of nuclear repair welding.