Abstract: Cold metal transfer (CMT) lap welding was performed on 6061 aluminum alloy sheet (aluminum sheet) and DP590 steel sheet. The welding process parameters were optimized by orthogonal method. BP neural network was used to predict welding deformation amount of the aluminum/steel sheets. The predicted results were applied to the aluminum/steel sheets, and then the welding deformation amounts were measured. The results show that the optimal process parameters were listed as follows:wire feed speed of 3.6-3.9 m·min^-1, welding speed of 0.66-0.70 m·min^-1, arc length correction of 0-5% and aluminum sheet thickness of 1.5-2.0 mm. The maximum test force that the joint welded with the optimal process parameters can withstand was up to 3 400 N, and the maximum thickness of the metal compound transition layer in weld was about 7.43 mm. The prediction results for welding deformation amount by the BP neural network were in good agreement with the experimental results. After anti-deformation, the welding deformation of the aluminum/steel sheets decreased significantly, with deformation amount decreasing from 0.67 mm to 0.12 mm, indicating that the prediction method was valid.