In order to satisfy the increasing fabrication quality requirements of micro curved thin wall structures, deformation, burrs height and dimensional errors in micro-milling processes of curved thin walls are studied. At first, the finite element analysis of curved thin wall deformation is carried out, it shows that complex elastic-plastic deformation and recutting occur in each micro-milling layer. Appropriate radial depth of cut, smaller spindle speed and axial depth of cut are beneficial to reduce deformation. Then, the influence of key process parameters in micro-milling as the speed of spindle n, feed engagement fz, radial depth of cut ae and axial depth of cut ap on milling force (Fn), burrs height (H) and dimensional errors (△w) are analyzed using the response surface method (RSM). To facilitate the quantitative analyses, mathematical models of burrs height and dimensional errors are established considering the interactions and quadratic effects of key parameters. It shows that the interactions among key parameters cannot be ignored. Therefore, cutting parameters should be delicately selected. The multi-objective particle swarm optimization (MOPSO) experiments are carried out based on the established prediction model. The results show that the combination of the optimized parameters identified by this method can effectively reduce the burr height and the dimensional error. Finally, under the guidance of this method, high-quality curved thin wall parts are fabricated.