Parallel and scalable block system generation
Parallel and scalable block system generation
复制标题
DOI：
10.1016/j.compgeo.2017.05.001
复制
发表时间：
2017-09
影响因子：
5.3
通讯作者：
Michael Gardner;Jack Kolb;N. Sitar
中科院分区：
文献类型：
--
作者：
Michael Gardner;Jack Kolb;N. Sitar
在许多不同的分析中，生成破裂岩体的真实表示是第一步。需要将实地观测转化为3-D模型，作为这些分析的输入。块系统可以包含数十万到数百万个大小和形状不同的块；生成这些大型模型的计算成本非常高，并且需要大量的计算资源。通过利用大数据分析和云计算方面的进步，我们开发了一个开源程序SparkRock，它可以并行生成块系统。该应用程序运行在ApacheSpark上，这使得它可以在本地、计算集群或云上运行。块生成基于Boon等人介绍的细分和线性规划优化。(2015).SparkRocks自动维护并行进程之间的负载平衡，并且可以在云上向上扩展，而无需对底层实施进行任何更改，使其能够在几分钟内生成包含数百万块的真实世界规模的块系统。
Generating a realistic representation of a fractured rock mass is a first step in many different analyses. Field observations need to be translated into a 3-D model that will serve as the input for these analyses. The block systems can contain hundreds of thousands to millions of blocks of varying sizes and shapes; generating these large models is very computationally expensive and requires significant computing resources.By taking advantage of the advances made in big data analytics and Cloud Computing, we have a developed an open-source program—SparkRocks—that generates block systems in parallel. The application runs on Apache Spark which enables it to run locally, on a compute cluster or the Cloud. The block generation is based on a subdivision and linear programming optimization as introduced by Boon et al. (2015).SparkRocksautomatically maintains load balance among parallel processes and can be scaled up on the Cloud without having to make any changes to the underlying implementation, enabling it to generate real-world scale block systems containing millions of blocks in minutes.