Printing 3D objects with interlocking parts

https://doi.org/10.1016/j.cagd.2015.03.020Get rights and content

Abstract

Recent advances in 3D printing technologies bring wide range of applications from fast prototyping to product manufacturing. However, one intrinsic limitation of 3D printing is that we cannot fabricate a single object that is larger than the working volume of a 3D printer. To address this issue, we may partition the given object into 3D parts of manageable sizes for printing, and then assemble the object from the printed 3D parts. Rather than using connectors, glue, or skew, we propose to connect the printed 3D parts by 3D interlocking such that the assembled object can be not only repeatedly disassembled and reassembled, but also strongly connected by the parts' own geometry. To achieve these, we develop a voxelization-based approach to partition a given 3D model into 3D interlocking parts. To guarantee the generated 3D parts to be structurally sound and well-connected by 3D interlocking, we deform the local geometry of the 3D model to avoid voxel fragmentation, employ internal voxels to create initial interlocking parts, and analyze the local shape within voxels to guide the final parts construction. We demonstrate the effectiveness of our approach on 3D models with a variety of shapes, and realize some of them by 3D printing.

Introduction

3D printing is an additive manufacturing technology, facilitating convenient and rapid fabrication of physical objects of almost any shape. It has a wide range of practical applications, from fast product prototyping, product development, 3D visualization, to distributed manufacturing of larger-sized objects such as machine parts.

However, 3D printing has an intrinsic limitation: a 3D printer cannot directly print an object whose size is greater than the printer's working volume. This practical limitation has been pointed out recently by Luo et al. (2012), who proposed a solution to partition a given 3D object into parts for 3D printing and then assemble the printed parts together to reconstruct the object. Besides addressing the intrinsic limitation, this approach has several other advantages. First, it facilitates cost-effective maintenance since we only need to print a replacement part for a corresponding broken part rather than reprinting the entire object. Second, this approach is good for storage and transport since we can disassemble a large 3D object to save space and avoid breaking it during the transportation. Lastly, similar to Lego bricks, we could reassemble an existing 3D object, change some of its parts, and even reconfigure it for alternative designs and appearance.

To connect 3D printed parts, there are some common approaches, e.g., by male and female connectors (Luo et al., 2012, Lo et al., 2009) and by glue (Shapeways, 2014). For the case of male and female connectors, though they are generally practical, they may not provide sufficient structural strength to sustain the parts connections. Moreover, for these tiny printed features, they may be broken easily during the object assembly or the transportation. For the case of glue, though it could tightly join the printed 3D parts together, it is a permanent connection, discouraging object reassembly, cost-effective maintenance, and reconfiguration. In sharp contrast, we take a 3D interlocking approach (Xin et al., 2011, Song et al., 2012) to construct and connect printed 3D parts to form an object assembly. By this, we can overcome the above mentioned issues.

Connecting parts by 3D interlocking has several advantages: i) first, parts connections are achieved by the parts' own geometry without requiring the creation of extra tiny connectors such as male and female connectors; ii) the assembled object can be repeatedly disassembled and reassembled, facilitating cost-effective maintenance, storage, and transportation; iii) 3D interlocking is known to be strong, evidenced by its usage in long-standing architectural wooden structures; hence, this connection method allows us to achieve stronger 3D parts connections, which are in fact supported by inter-parts blockage with their geometry; and iv) lastly, it enables us to produce 3D parts with clean and smooth surface without hole drilling and protrusion.

In this paper, our goal is to partition a given 3D object into interlocking parts for 3D printing and object assembly. This is a highly challenging and unexplored problem, since we have to enforce not only the complicated 3D interlocking requirement (Xin et al., 2011, Song et al., 2012), but also the geometric and dimensional requirements on the printed 3D parts, as well as maintaining the 3D object appearance after its assembly. The requirements to be considered when we develop the computational method are listed as below:

  • 3D interlocking: the printed 3D parts should interlock one another, yet can be assembled and disassembled;

  • Printable parts: while being not too small, each printed 3D part should fit into the working volume of the target 3D printer;

  • Structural soundness: we should avoid thin and/or weak features on the 3D parts since such features could be easily broken during the object printing, assembly, or transportation;

  • Strong connection: we should achieve strong 3D parts connections by ensuring strong blockage among the 3D parts in the 3D interlocking assembly; and

  • Aesthetics: lastly, we should avoid having cutting seams that pass through salient regions on the assembled object surface since this affects the object appearance.

Note that state-of-the-art methods (Xin et al., 2011, Song et al., 2012) for creating 3D interlocking structures are insufficient to handle the problem since they mainly focus on the 3D interlocking requirement without considering the 3D printing issues such as structural soundness and object appearance.

To achieve the above requirements, we develop a novel voxelization-based approach to construct interlocking 3D parts from a given 3D model. Our technical contributions are three-fold. First, we develop a new framework that can create interlocking 3D parts from 3D models of general shape, where novel ideas include voxelizing the 3D model with local shape analysis, employing internal voxels to create initial interlocking parts, and attaching boundary voxels to initial parts while retaining the 3D interlocking. Second, we propose to deform the input model surface subject to the voxelization, which helps to avoid fragmented and disconnected shape features on the generated 3D parts. Lastly, we propose the shape and saliency connection graphs encoding local shape information to guide the parts geometric construction for achieving the parts structural soundness and aesthetics requirements. We have demonstrated the effectiveness of our approach on 3D models with a variety of shapes, and fabricated some of them by 3D printing to validate the parts connection capability, as well as their structural soundness.

Section snippets

Related work

3D fabrication

A number of research works in computer graphics have studied different aspects of 3D fabrication, e.g., deformation behavior (Skouras et al., 2013), mechanical characters (Coros et al., 2013, Ceylan et al., 2013), articulated models (Bächer et al., 2012, Calì et al., 2012), spinnable fabrication (Bächer et al., 2014), 3D shape balancing (Prévost et al., 2013), and printing material reduction (Wang et al., 2013, Lu et al., 2014).

Specifically, some of them focus on structural issues in 3D

Overview

In this section, we first discuss the challenges of partitioning a given 3D shape into interlocking 3D parts, and then give an overview of our approach.

Challenges of 3D shape partitioning

For a given 3D shape, there exist different ways of partitioning it into parts, for example, plane – (Luo et al., 2012), curvature – (Hao et al., 2011), and voxelization-based (Medellín et al., 2007, Zhou et al., 2014) approaches. This work employs the voxelization-based partitioning approach to create interlocking parts since i) cubical voxel

Voxelization and shape analysis

This section presents the various preparation works we have before constructing the geometry of interlocking 3D parts from an input 3D model. This includes voxelization, local shape deformation, and the generation of shape connection graph and saliency connection graph.

Voxelization method

Given the input watertight 3D model, we first voxelize it by (Nooruddin and Turk, 2003), which casts parallel rays through the model and uses parity count to classify voxels as interior or exterior. We improve this method by

Generating 3D interlocking parts

Given the various inputs we prepared in Section 4, there are four major steps in generating 3D interlocking parts:

Step 1: Generating initial 3D interlocking parts

First, taking the voxels in the internal volume as input, we employ (Song et al., 2012) to construct an initial set of interlocking 3D parts, see Fig. 2(c). Here we give a short description of the randomized algorithm in (Song et al., 2012), which takes a general voxelized shape as input and iteratively extracts puzzle pieces from it to create interlocking puzzles. A formal model is

Results

Our method enables us to create 3D interlocking parts from object models of various shapes, see Fig. 9. All the generated 3D parts are well-connected by interlocking upon the parts assembly, see Fig. 10 for three more examples.

Implementation and performance

We implemented our method in C++ and ran our experiments on a desktop computer with a 3.4 GHz CPU and 8GB memory. For a given 3D model, users can specify the number of parts to be generated. Although our method allows the creation of a large number of 3D parts from a

Conclusion

This paper presents a novel voxelization-based method to partition a given 3D object model of general shape into interlocking 3D parts such that the 3D object can be printed with smaller parts that are connected by 3D interlocking. We first voxelize the input 3D model and analyze the local shape within each voxel, where a local deformation strategy is developed to avoid voxel fragmentation in the voxelization. Second, we differentiate internal and boundary voxels according to the local shape

Acknowledgements

The authors would like to thank Yael Friedman, a jewelry designer, for her various helpful comments on designing the 22-parts interlocking ring. The project is partially supported by the National Natural Science Foundation of China (61403357, 61222206), the Fundamental Research Funds for the Central Universities (WK0110000044), One Hundred Talent Project of the Chinese Academy of Sciences, and MOE AcRF Tier2 funding (MOE2011-T2-2-041 (ARC 5/12)), Singapore.

References (37)

  • M. Attene

    Direct repair of self-intersecting meshes

    Graph. Models

    (2014)
  • K. Hildebrand et al.

    Orthogonal slicing for additive manufacturing

    Comput. Graph. (SMI)

    (2013)
  • 3D Systems
  • P. Alliez et al.

    Recent advances in remeshing of surfaces

  • M. Bächer et al.

    Fabricating articulated characters from skinned meshes

    ACM Trans. Graph. (SIGGRAPH)

    (2012)
  • M. Bächer et al.

    Spin-It: optimizing moment of inertia for spinnable objects

    ACM Trans. Graph. (SIGGRAPH)

    (2014)
  • M. Botsch et al.

    Polygon Mesh Processing

    (2010)
  • J. Calì et al.

    3D-printing of non-assembly, articulated models

    ACM Trans. Graph. (SIGGRAPH Asia)

    (2012)
  • D. Ceylan et al.

    Designing and fabricating mechanical automata from mocap sequences

    ACM Trans. Graph. (SIGGRAPH Asia)

    (2013)
  • D. Chen et al.

    Computing and fabricating multiplanar models

    Comput. Graph. Forum (EuroGraphics)

    (2013)
  • S.T. Coffin

    The Puzzling World of Polyhedral Dissections

    (1990)
  • S. Coros et al.

    Computational design of mechanical characters

    ACM Trans. Graph. (SIGGRAPH)

    (2013)
  • J. Hao et al.

    An efficient curvature-based partitioning of large-scale stl models

    Rapid Prototyping J.

    (2011)
  • K. Hildebrand et al.

    crdbrd: shape fabrication by sliding planar slices

    Comput. Graph. Forum (EuroGraphics)

    (2012)
  • M. Lau et al.

    Converting 3D furniture models to fabricatable parts and connectors

    ACM Trans. Graph. (SIGGRAPH)

    (2011)
  • C.H. Lee et al.

    Mesh saliency

    ACM Trans. Graph. (SIGGRAPH)

    (2005)
  • K.-Y. Lo et al.

    3D polyomino puzzle

    ACM Trans. Graph. (SIGGRAPH Asia)

    (2009)
  • L. Lu et al.

    Build-to-last: strength to weight 3D printed objects

    ACM Trans. Graph. (SIGGRAPH)

    (2014)
  • Cited by (89)

    • A computational design approach for multi-material 4D printing based on interlocking blocks assembly

      2022, Additive Manufacturing
      Citation Excerpt :

      From this, it is then possible to address multi-scale modeling [60–62] with different voxel shapes and structures [63], whether solid, lattice, or even metamaterials [64–69]. The proposal extends prior work on interlocking assemblies [52,55,56,70] by considering as initial input a computed materials distribution in a 4D object via a dedicated CAD plugin [21,26]. The structure of the paper is as follows.

    • Synthesis of 3D jigsaw puzzles over freeform 2-manifolds

      2022, Computers and Graphics (Pergamon)
      Citation Excerpt :

      The state-of-the-art techniques currently available on a wide range of fabrication capabilities are reported in Bermano et al. [7] and Bickel et al. [8].) The majority of these recent results are in the form of volumetric dissections [9,10], often with interlocking structures [11–15] or even with transformable structures [16–18]. Sequin [10] used dissection puzzles in a graduate course at UC Berkeley for training the spatial understanding and modeling of solid shapes.

    View all citing articles on Scopus
    View full text