Investigation of the Directional Characteristics of the Emitted Airborne Sound by Friction Stir Welding for Online Process Monitoring

Abstract

AbstractCurrently, destructive or non-destructive testing methods are used to verify the weld seam quality subsequent to the manufacturing process. Therefore, pre-processes such as visible or mechanical testing require additional efforts, which can lead to expensive reworking or rejection of the components. The acoustic process characterization for Friction Stir Welding (FSW) applications permits a comparatively new approach of process monitoring to detect weld seam irregularities by the characterization of the emitted noise in the audible frequency range (airborne sound signal). In previous publications, the acoustic detection of weld seam irregularities was mostly based on structure-borne sound sensors. Although a correlation between weld defects and audio signals has been demonstrated, there are process-related deficits in the use of structure-borne sound sensors. These include a fixed installation position and limited applicability for large-scale components such as battery cases. In contrast airborne sound sensors (microphones) can be mounted directly in the area of the joining process and thus influences of component size, joining materials, and weld seam geometry can be reduced. However, the use of airborne sound sensors for FSW applications requires preparatory considerations on the sensor position towards the joining process (sidely, in front of or behind the processing tool). Therefore, in this study an approach will be presented to evaluate the directional characteristic of the airborne sound emitted by the FSW process. First, the positioning of the microphone for the various welding directions were investigated. This was done to determine a suitable microphone orientation during the process. Then, the general determination of audio signals from the FSW process will be considered and compared to the process force feedback. Further, it was demonstrated that acoustic analysis can be used for detection of weld seam irregularities such as flash formation on 5 mm AA 5754 H111 sheets. All experiments were performed with a robotized FSW setup that was modified by a self-developed acoustic measuring device.KeywordsMachine learningFriction stir weldingWeld seam irregularitiesProcess forcesAcoustic emissionsAudio signalPolar plotsProcess monitoringReal time monitoring