飞机载荷谱实测数据并行统计处理算法

Parallel statistics processing algorithm of aircraft load spectrum testing data

针对飞机载荷谱数以万亿计的实测数据,为了提高统计处理运行效率,提出了飞机载荷谱实测数据处理的并行算法。本文对载荷谱实测数据处理模型进行了多级并行化分析,在此基础上论述了粗粒度、中粒度、细粒度级并行处理方式,建立了两种并行处理算法——基于机型数据流的粗粒度与中粒度并行数据处理算法和基于某起落数据流的中粒度与细粒度并行数据处理算法。在小规模对称多处理器(Symmetrical multi-processors,SMP)集群运算平台下进行比较测试表明,可大幅地提高载荷谱数据处理运行效率,在8核运算环境下,最高能获得5.82的加速比,为飞机载荷谱实测数据处理研究领域进行大规模科学计算和提高数据处理效率提供了新的技术途径。

The parallel algorithm was presented for enhancing the statistics processing efficiency of thousands of billions aircraft load spectrum testing data.The data processing was multi-level parallel analyzed,and the coarse-grained,medium-grained,fine-grained level parallel processing modes were formulated.Two parallel algorithms were proposed ultimately,one is coarse-grained and medium-grained parallel algorithm based on aircraft type data flow,and the other is medium grain and fine-grain parallel algorithm based on flight data flow.Comparison test based on the small-scale SMP（Symmetrical Multi-processors） cluster shows that the algorithms can improve statistical processing efficiency greatly.A highest speedup of 5.82 in 8 cores can be achieved.The algorithms present a creative method for the complex scientific computation and enhance the computing efficiency in real testing data processing of load spectrum.