商用客机气动噪声大规模并行计算的建模、算法与软件

Aerodynamic noise problem is considered as a prerequisite for commercial aircraft airworthiness engineering. It is essential embodiment to improve core technical competitiveness of both home and abroad airliner markets. It is also an important guarantee for sustainable development of commercial aircraft in China. The project will deal with and overcome technical difficulties of commercial aircraft aerodynamic noise by the collaborations among high performance computing and applications, applied mathematics and aircraft designs. Based on the Boltzmann equation, Ffowcs Williams-Hawkings (FW-H ) equation and their coupling equations, highly precise and fast large-scale parallel algorithm of lattice Boltzmann method for directly solving aerodynamic noise are presented with parallel mesh generations and adaptive mesh refinements. Loop tiling techniques for exploitation of hierarchical memory in computing nodes and parallel optimization are intensively studied on the domestic Petaflop-scale supercomputer system. Numerical models and computational optimization program of commercial aircraft aerodynamic noise with large scale, high precision, scalability and efficiency will be fulfilled by using more than ten thousands of computing cores. Numerical results will be verified and validated by experimental data. An applied software system of computational commercial aircraft aerodynamic noise will be developed, applied to numerically simulate practical problems of commercial aircraft aerodynamic noise, and finally used to provide effective technical support for the development of commercial aircraft.

气动噪声问题是大型商用客机适航工程顺利推进的先决条件，是提高我民航客机市场竞争力核心技术的根本体现，更是我国大型商用客机可持续发展的重要保障。本项目联合高性能计算、应用数学和飞行器设计等领域研究人员，通过协同研究攻克客机气动噪声技术难点。本项目将基于Boltzmann方程，发展与FW-H方程耦合求解的大规模、高精度、快速并行求解算法、气动噪声的直接求解格子Boltzmann方法、网格快速并行生成及自适应网格加密技术，并面向P量级超级计算机系统开发相对应的计算节点层次存储架构和并行优化的循环分块方法，在国家超算中心P量级超级计算机上，实现大规模、高精度、快速高效的商用客机气动噪声数值计算和程序优化，计算核数达十余万核量级，并通过实验数据验证计算方法和确认计算结果。研制出适用于商用客机气动噪声计算的软件系统，并应用于商用客机气动噪声实际问题的数值模拟，最终为商用客机研制提供有效地技术支撑。

气动噪声问题是大型商用客机适航工程顺利推进的先决条件，本项目基于格子Boltzmann方法（LBM），引入多松弛时间和大涡模拟模型，建立了适用于高雷诺数情况下的大规模可扩展的LBM-MRT-LES并行算法，并研究了无反射的消声边界条件，保证了流场计算的精确性。在流场计算的基础上，结合FW-H声传播方程，建立了LBM-FW-H大规模流场-声场耦合计算模型。为计算有着不同复杂几何外形的物体，项目组开发了基于OpenMP和MPI的多层网格LBM算法，能够快速有效地生成LBM计算网格，并进一步地开发了在计算过程中根据物理量的变化动态调整多层网格的网格自适应加密技术，有效的减少了冗余网格量，提高资源利用效率。项目组面向国产超级计算机“神威·太湖之光”对LBM-FW-H耦合计算模型进行了并行优化，开发了主核和从核的两级并行模式、高效的并行I/O技术、LBM中代码的向量化优化以及减少共享Cache资源竞争的并行任务分配和调度策略，提高了算法的并行效率和可扩展性，13万核时可以保持94%的并行效率，百亿网格计算量时并行效率可以保持90%以上。应用LBM-FW-H耦合模型对经典气动标模进行流场和声场的计算，结果表明该耦合计算模型能够精确的计算近场及远场的噪声，项目组进一步应用该耦合模型对某民机全机模型进行了计算，在国内首次获得了真实商用飞机全机的噪声计算结果，并与中国商飞上海飞机设计研究院实验结果吻合。项目组构建了包含声场和流场的计算以及可视化的商用客机机体气动噪声数值计算平台，平台具有友好的流场及声场计算用户界面，并已经将平台用于商用客机的数值计算和噪声分析。