Abstract: For welding seam tracking of multi-layer single pass welding by narrow gap rotating arc MAG,an intelligent modeling method based on rough sets (RS) theory for seam deviation was put forward. First,workpiece was designed and processed to mimic multi-layer single pass welding groove,and enough experimental data were acquired under different deviations. Second,according to the simulation results of arc length and variation of experimental data,current signal was divided into 12 intervals in each rotating cycle. The mean value of each interval and differences between left and right intervals were computed to build decision table. Then after discretization and reduction for decision table,the knowledge model of "IF…THEN" form was obtained. At last the model was validated and compared with BP net model. It showed that the both models had similar predictive capability,and RS model precision could meet actual needs. Furthermore,RS model had better comprehensibility, and was useful to find potential laws between seam deviations and welding electrical signals from experimental data. The research was helpful for controller design.