Monitoring of 304 austenitic stainless-steel laser-MIG hybrid welding process based on EMD-SVM

Abstract

Real-time monitoring of laser-metal inert gas (MIG) hybrid welding plays a significant role in modern automatic production. In this paper, an innovative prediction model is proposed for laser-MIG hybrid welding process monitoring and weld defect detection. Specifically, the status of laser-MIG hybrid welding was monitored using a high-speed imaging system that could observe the visual information from the top and bottom of the weldment simultaneously. Then, the morphological features were extracted by image process algorithm. Furthermore, the empirical mode decomposition (EMD) was employed to analyze the time series of visual features, and the decomposed component intrinsic mode functions (IMFs) and residue were inputted to the support vector machine (SVM) classification model. Finally, the EMD-SVM model was employed to evaluate the welding status of four weld formations, including sound weld, root humping, incomplete penetration and burn through. Experimental results demonstrate the superiority of the proposed method in terms of accuracy and robustness compared with traditional methods such as the backpropagation (BP) neural network, SVM, wavelet packet decomposition (WPD)-BP, WPD-SVM, and EMD-BP. This proposed method provides a novel and accurate approach to perform laser-MIG welding process monitoring and online defects detection.