LAUSR

Abstract The detection and control of the internal porosity have attracted increasing attention in the arc welding of aluminum alloys. In this paper, a spectrometer was utilized to realize the in-situ detection and processing of arc spectra. By performing improved locally linear embedding on the spectral band of interest, the mapping vectors characterizing porosity defects and the intrinsic structure of the spectra were obtained. To improve the identification accuracy of porosity defects, the ratio feature of specific spectral lines acquired by empirical mode decomposition was used as an evaluation index as well. Based on the above features, the rapid and high accuracy prediction of porosity defects was realized by extreme learning machine. A controller based on the Fuzzy-PID was designed to regulate and control the porosity defects evaluated by spectral features in real time. The closed-loop control experiments showed that the Fuzzy-PID control system achieved the porosity control under various welding assembly conditions by adjusting the welding current.