In order to improve the ability to predict the manufacturability of complex surfaces in the design stage, the manufacturability evaluation indexes and models of complex part surfaces were studied. The factors affecting the manufacturability of the complex part surfaces were divided into two groups, the geometrical complexity (GC) and the technological complexity (TC). Considering the features in GC, the complex part surfaces were classified into several clusters according to distribution features of complex surfaces and slopes, which established the group patterns of the parts. Then the discrete and the reduction algorithms of rough set were used to calculate the weights of the feature attributes including a few attributes in TC. After that, the method of fuzzy pattern recognition and the calculation of surface similarity were used to evaluate the manufacturability of parts. A case study showed that the method could evaluate the manufacturability of the complex part surfaces accurately.