generation. Until now, however, it has been too difficult to generate hexahedral meshes of complex CAD models automatically. As a result, a complex model that cannot be well meshed as a whole is usually decomposed into a number of components, each of which can be meshed automatically by certain algorithms, and then the hexahedral meshes of all the components are combined together as the resultant meshes of the whole CAD model. One problem regarding this method is that the interfaces between the hexahedral meshes of the components are usually non-conforming.Assembly meshing is a typical example. We usually mesh the assembly components, i.e., parts, separately, instead of the whole assembly together. Because each part of the assembly may be meshed by different FEA engineers, or using different meshing techniques, or taking different element density, the meshes of the parts often become non-conforming on the assembly interfaces.In addition to the assembly meshing problem, reusing of the existing hexahedral mesh models also encounters a problem regarding the non-conforming interfaces. Because the high-quality hexahedral mesh generation of complex CAD models is very difficult and time-consuming, the ability to reuse the existing hexahedral mesh models in the library whose size keeps expanding is becoming increasingly important. Considering a reuse-based scenario, for a complex CAD model, instead of meshing it from scratch, its hexahedral meshes are created by combining existing mesh models available by searching with CAD models. Obviously, the searched mesh models hardly conform on their interfaces.Currently, to obtain an integral solution based on the mesh model with non-conforming interfaces, it is necessary to impose constraints on the non-conforming interfaces, such as gap elements and multi-point constraints [2][3][4]. Nevertheless, when the mesh models are large and Abstract Mesh matching is an effective way to convert the non-conforming interfaces between two hexahedral meshes into conforming ones, which is very important for achieving high-quality finite element analysis. However, the existing mesh matching algorithm is neither efficient nor effective enough to handle complex interfaces and selfintersecting sheets. In this paper, the algorithm is improved in three aspects: (1) by introducing a more precise criteria for chord matching and the concept of partition chord set, complex interfaces with internal loops can be handled more effectively; (2) by proposing a new solution, self-intersecting sheet can be inflated and extracted locally; and (3) by putting forward a mesh quality evaluation method, the sheet extraction operation during mesh matching can be done more efficiently. Our improved mesh matching algorithm is fully automatic, and its effectiveness is demonstrated by several examples in different matching situations.