Mesh self-intersection defects and image noise may prevent 3D model reconstruction and mesh formation of bones with comminuted fractures, making it impossible to assemble 3D-printed fragments perfectly. This study proposes an algorithm to remove overlapping meshes and to smooth fracture surfaces in order to fabricate well-assembled 3D-printed bone. 3D bone reconstruction, segmentation, and reduction were directly performed for three different classes of clinical fracture cases: pelvic 62-B1, 62-C2, and femur 31-A2.2. In contrast to the current Boolean operation, the proposed algorithm is not only capable of detecting overlapping meshes, but also recognizing the contact regions and detecting the boundary of each contact region. Hence, it was implemented in order to remove overlapping meshes and ensure that fragments fit together when physically assembled. Both gap distance and overlapping mesh errors during assembly of the 3D model from printed bone fragments were calculated and analyzed. Based on the comparison of results between the bone model before and after removing mesh defects, the RMS error is less than 0.33 mm and gap error is 3 mm, indicating that the proposed technique has high potential for eliminating mesh defects and providing a 3D-printed bone fracture model that is easy to assemble and disassemble.