Porosity detection in pulsed GTA welding of 5A06 Al alloy through spectral analysis

The influence of line-broadening and zero shift of the sensor on identifying spectral lines was eliminated by an improved K-medoids clustering algorithm after performing principal component analysis on the spectral band of interest. The intensity ratio of the acquired spectral lines was calculated and dealt with by empirical mode decomposition to predict the existence of porosity based on the method given by our previous study. Furthermore, the effect of sheet gap on porosity formation was investigated by the proposed method. The analysis results were verified by X-ray detection, which demonstrated the distribution of porosity as well. The results showed that an appropriate gap could contribute significantly to reducing the porosity. The spectral sensor system and analysis method proposed in this study may help decrease the testing costs and increase productivity, while maintaining high weld quality.