With the advance of industrial modernization,the development of robot-based industrial automation manufacturing technology has become a trend. The welding system of MOTOMAN-UP6 arc welding robot and 2R rotation-tilt positioner is designed according to additive manufacturing（3D printing） technique. The ideal welding parameters affect the formation of a standard cylinder. Therefore,the influence of welding current and welding speed on the size of bead is studied by using the BP neural network algorithm,and the ideal experimental welding parameters are predicted. Through a large number of experiments,a relatively standard cylinder is obtained.Comparing with rapid forming methods for welding robots without a positioner,this method is easier,and the deformation of the welding layer is smaller,the surface of the bead is smooth.