Interpreting acoustic emissions to determine the weld depth during laser beam welding

Abstract

The interpretation of sensor system data is critical for monitoring industrial welding processes and providing reliable information about the condition of the weld seam. Previous investigations have shown that acoustic emissions of frequencies up to several kilohertz during laser beam welding are parameter-dependent and contain valuable information about the process. A microphone was employed to record the acoustic emissions produced when performing deep penetration laser beam welding of copper. Experiments were conducted in which the laser power and the feed rate were varied so as to obtain acoustic data comprising frequencies of up to 1 MHz. The signals were preprocessed and features were extracted using Fourier and wavelet analysis as well as speech analysis techniques. The relationship between the features extracted from the acoustic signal and the weld depth was modeled using Gaussian process regression. The results showed that acoustic emissions during laser beam welding can be used to predict the weld depth without having to rely on process parameters, i.e., the laser power and the feed rate. Overall, 17 features were extracted from acoustic signals, with the zero-crossing rate displaying the highest significance for determining the weld depth. These investigations open up new possibilities of robust quality assurance for laser beam welding applications based on acoustic emissions.