Abstract: The retrieval for existing 3D CAD models has a great potential in efficient reuse of product design information.In engineering applications,a lot of products often contain some quadratic or freeform surfaces.This paper describes a 3D surface model retrieval approach based on the shape distribution of distance and curvature.First,a sufficiently large number of random points on surface of 3D model are sampled.The curvature and the Euclidean distance between each sample point and the 3D model's centroid are calculated.Secondly,a distance-curvature planar grid taking ball shell radius centered by the 3D model's centroid and curvatures respectively as two coordinate axes is constructed to express the distance-curvature distribution by obtaining a statistic data of the sample points and their curvatures existed in each mesh.Thus,a 3D model can be represented by a distance-curvature distribution matrix.Finally,an algorithm to compute the similarity between two distance-curvature distribution matrices is proposed and it can give the similar coefficient for two compared 3D CAD models.Experiment results showed that this approach can efficiently evaluate the similar degree of 3D surface models,especially for those models with freeform surfaces.