基于GPU计算架构的体素离散化复杂药柱燃面计算方法  

GPU computing architecture-based discrete voxelized burning surface calculation method for complex SRM grains

作　　者：刘旸[1] 鲍福廷[1] 蔡强[1] 胡海峰[1] 杜佳佳[1] 

机构地区：[1]西北工业大学航天学院,西安710072

出　　处：《固体火箭技术》2011年第1期18-22,共5页Journal of Solid Rocket Technology

基　　金：国家自然科学基金(51005179);博士后科学基金(20100471636)

摘　　要：为了高效而通用地对各种具有复杂结构的药柱进行燃面计算,文中基于图形处理器(GPU)高度并行化的计算构架,提出了一种三维离散化体素点阵表达的燃面计算方法。首先,将药柱所在的包围盒空间均匀地离散成三维空间体素点阵,通过对边界表达的任意药柱模型进行Z-Buffer测试确定计算边界;其次,采用并行的燃烧状态填充算法确定所有的体素点与初始燃面之间的距离关系;然后,针对任意指定的燃烧肉厚,利用GPU多核并行处理技术,实时地抽取所有药柱空间内的燃烧表面;最后,以一个具有复杂包覆层的不规则药柱作为算例,从计算精度和计算效率上验证了该燃面计算方法的有效性。To calculate burning surface of complex SRM grains in a more efficient and more generalized way,a 3D discrete voxels-represented burning surface calculation method was proposed.Based on highly paralleled graphics process units（GPU） computing architecture,the bounding box space of the grain was discretized into a 3D voxels matrix firstly,and Z-Buffer testing was used to confirm the calculation boundary of the grain.Then,paralleled burning state filling algorithm was adopted to calculate the distance fields between initial burning surface and voxels matrix.For a given web,paralleled GPU multi-core parallel processing technology was also used to extract the burning surface in real time.Finally,taking an irregular grain as the example,which had complex inhibitors,the calculation accuracy and computational efficiency verified the validity of this burning surface calculation method.

关 键 词：图形处理器 体素离散 燃面计算 

分 类 号：V430[航空宇航科学与技术—航空宇航推进理论与工程]