A Parallel Slicing Algorithm for Solid Freeform Fabrication Processes
A Parallel Slicing Algorithm for Solid Freeform Fabrication Processes
复制标题
固体自由形状制造过程的并行切片算法
DOI：
--
复制
发表时间：
1992
期刊：
影响因子：
--
通讯作者：
C. C. Jara
中科院分区：
文献类型：
--
作者：
C. Kirschman;C. C. Jara
切片可以占60%以上的时间来准备用于在天体光刻机上建造的材料。为了缩短准备时间，提出了一种并行切片算法。该算法在使用2、4、8、16和32个处理器的Butterly GP1000上运行，由于对内存的要求很高，观察到了超线性加速比。与单处理器相比，该并行算法在16个并发处理器上可以减少高达92%的切片时间。简介立体光刻是当今可用的几种不同的固体自由电子制造(SFF)技术之一(Ashley，1991)。最常用的立体光刻机(SLA)是来自3D Systems(加利福尼亚州巴伦西亚)的SLA 250。由于它的流行，本文描述的最初软件开发是针对这台机器的；然而，人们相信这些概念可以扩展到任何一种技术。由3D系统公司()详细介绍的零件的准备工作从计算机辅助设计系统开始。将创建实体模型来表示零件。然后将该零件转换为立体平版印刷字体(STL)文件。该文件包含一组定义零件曲面的三角形以及指向这些三角形的向外的法线。接下来，对该分面表示进行预处理，以便在SLA中构建。首先支撑零件，然后由另一台计算机对其进行切片。然后，包含这些层的文件被发送到第三台计算机，以便在SLA中进行合并、准备和重建。这些部件自下而上逐层建造，厚度从0.0025英寸到0.030英寸不等。这项技术的用户最大的抱怨之一是建造前的准备时间太长。尤其是切片，可能占到CAD系统和SLA之间60%或更多的时间。正因为如此，正在努力减少切片时间。本文讨论了采用并行体系结构来提高切片速度的技术。
Slicing can accountfor more than 60% ofthe time to prepare apartfor building on astereolithographic apparatus. To improve the preparation time, a parallel slicing algorithm was developed. The algorithm was run on a Butterfly GP1000 using 2,4,8, 16, and 32 processors and superlinear speedup was observed due to high memory requirements. The parallel algorithm can reduce slice times by up to 92% on 16 concurrent processors as compared to a single processor. INTRODUCTION Stereolithography is one of several different Solid Freefonn Fabrication (SFF) technologies available today (Ashley, 1991). The most common stereolithography apparatus (SLA) is the SLA 250 from 3D Systems (Valencia, CA). Because of its popularity, the initial software development described in this paper was aimed at this machine; however, it is believed that the concepts can be extended to any of the technologies. The preparation of a part, detailed by 3D Systems (1989), begins in a CAD system. A solid model is created to represent the part. The part is then converted to a stereolithographic fonnat (stl) file. This file contains a set of triangles which define the surface of the part and the outward pointing nonnals to these triangles. Next, this faceted representation is preprocessed for building in the SLA. First the part is supported, and then it is sliced by another computer. Then the files containing the layers are sent to a third computer for merging, preparation, and reconstruction in the SLA. The parts are built layerby-layer from the bottom up in layers 0.0025 to 0.030 inch thick. One of the biggest complaints of the users of this technology is that the preparation time before building is too long. Slicing in particular can account for 60% or more of the time between the CAD system and the SLA. Because of this, efforts are underway to reduce the slicing time. This paper discusses techniques for improving the slicing speed by employing parallel architectures.