A variable precision rough set based modeling method for pulsed GTAW

Modeling is an important step both for quality and shaping control of the arc welding process. Current modeling methods have made great advances in the field of arc welding, however they all posses certain limitations. It is due to these limitations that we created the variable precision rough set (VPRS) based modeling method. The VPRS modeling has been shown to be both a more efficient and reliable modeling method for the arc welding process due to its ability to account for the character of the welding media. The method was used to produce a dynamic predictive model for pulsed gas tungsten arc welding (GTAW). Results showed that the VPRS modeling method was able to sufficiently acquire knowledge during welding practices. In addition, comparison of VPRS model with classic rough set model and BP neural network model showed that VPRS model was more stable and could predict the unseen data better than classic RS model. Moreover, the VPRS model owns similar precision with neural network model, but has better understandability.