Abstract: Arc welding robot is used for arc additive manufacturing, and the welding technology of additive manufacturing of thick-walled structural parts is studied. Based on the traditional stratification theory, the prediction and analysis of the molding dimension of the thick wall structure are carried out with the algorithm optimization. On the basis of this algorithm, the prediction model of the single weld shape size neural network is introduced to improve the precision of the preform model and the optimum selection of the actual welding parameters; In view of the collapse of the weld edge in the forming process of thick wall structures with inner holes and other features, the "boundary constraint" welding method was proposed and the interlayer welding trajectory was planned to improve the quality of the surface of the preform. Finally, the welding method has an illustrative physical component verification prediction algorithm and the accuracy of trajectory planning. The test results show that the structure is well formed and the size error is less than 1 mm.