Nondestructive, in-process inspection of inertia friction welding: An investigation into a new sensing technique

This paper investigates the capabilities of a new sensor for in-process monitoring of quality during friction welding. The non-contact sensor is composed of microphones that are mounted in an aluminum ring which surrounds the weld joint. The sensor collects the acoustical energy (in the form of sound pressure) that is emitted during the plastic deformation and phase transformations (if applicable) in friction welding processes. The focus in this preliminary investigation is to search for and identify features within the acoustical emission that are indicative of bond quality. Bar-to-bar inertia friction welding (one form of friction welding) of copper to 304L stainless steel is used in this proof-of-concept study. This material combination exhibits only marginal weldability and is ideally suited for validating the capabilities of this new sensing technique. A probabilistic neural network is employed in this work to analyze the acoustical emission's frequency spectrum in an attempt to classify acceptable, conditional, and unacceptable welds. Our preliminary findings indicate that quality-based process features do exist within the frequency spectrum of the acoustical signature. The results from this analysis are presented. Future work in improving the sensing and interpretation of the data is discussed in an effort to develop a robust method of quality-based, in-process monitoring of friction welds.