Neural Slicer for Multi-Axis 3D Printing

Tao Liu,Tianyu Zhang,Yongxue Chen,Yuming Huang,Charlie C.L. Wang

2024-04-23

Abstract:We introduce a novel neural network-based computational pipeline as a representation-agnostic slicer for multi-axis 3D printing. This advanced slicer can work on models with diverse representations and intricate topology. The approach involves employing neural networks to establish a deformation mapping, defining a scalar field in the space surrounding an input model. Isosurfaces are subsequently extracted from this field to generate curved layers for 3D printing. Creating a differentiable pipeline enables us to optimize the mapping through loss functions directly defined on the field gradients as the local printing directions. New loss functions have been introduced to meet the manufacturing objectives of support-free and strength reinforcement. Our new computation pipeline relies less on the initial values of the field and can generate slicing results with significantly improved performance.

Computational Geometry

What problem does this paper attempt to address?

The paper proposes a novel computational pipeline called "Neural Slicer" for multi-axis 3D printing. Traditional 3D printing typically uses a planar layering approach, which has limitations in terms of support structure requirements, surface quality, and mechanical strength. Multi-axis 3D printing solves these issues by increasing degrees of freedom, but existing methods have limitations in representing and processing complex topology of input models.

The Neural Slicer aims to overcome these challenges by being able to handle models with diverse representations and complex structures. The core of the method is to use a neural network to establish a deformation mapping, defining a scalar field in the space surrounding the input model. This mapping is generated by optimizing the neural network directly with a loss function defined on the local printing direction, in order to meet manufacturing objectives such as support-free (SF) and strength enhancement (SR). The new method reduces dependency on initial field values and can generate significantly improved slicing results.

The paper mentions that existing S3-slicer methods can achieve multiple objectives but require high-quality tetrahedral meshes, and the optimization objectives are indirectly defined, which may lead to violations of manufacturing requirements. The Neural Slicer directly optimizes the scalar field using a neural network, reducing dependency and better adapting to complex geometry and topology.

In summary, the paper addresses the problem of developing a method that can handle 3D models with different representations and complex structures, while optimizing the printing layers to meet the manufacturing requirements of support-free and strength enhancement in multi-axis 3D printing, in order to improve the performance and outcomes of 3D printing.