A study of welding process modeling based on Support Vector Machines

This paper addresses the Support Vector Machines (SVM) method for developing the model of pulsed Gas Tungsten Arc Welding (GTAW). Modeling of the welding process is an important but difficult process in automatic welding because it is a multivariable, time-delay and nonlinear process. SVM is a tool based on statistical learning theory, widely used for prediction tasks on small sample data for its generalization capacity. In this paper, we analysis the characteristics of SVM for solving the modeling problem of pulsed GTAW and gives the main steps of modeling. Experiment results show that the SVM model is able to predict the GTAW process correctly and comparison of SVM method with neural network method shows that the SVM model is more precise.