航空学报 > 2018, Vol. 39 Issue (5): 121736-121736   doi: 10.7527/S1000-6893.2017.21736

基于代理模型的高效全局低音爆优化设计方法

乔建领, 韩忠华, 宋文萍   

  1. 西北工业大学 航空学院 翼型叶栅空气动力学国家级重点实验室, 西安 710072
  • 收稿日期:2017-09-11 修回日期:2017-11-08 出版日期:2018-05-15 发布日期:2017-11-08
  • 通讯作者: 韩忠华,E-mail:hanzh@nwpu.edu.cn E-mail:hanzh@nwpu.edu.cn
  • 基金资助:
    国家自然科学基金(11772261);航空科学基金(2016ZA53011)

An efficient surrogate-based global optimization for low sonic boom design

QIAO Jianling, HAN Zhonghua, SONG Wenping   

  1. National Key Laboratory of Science and Technology on Aerodynamic Design and Research, School of Aeronautics, Northwestern Polytechnical University, Xi'an 710072, China
  • Received:2017-09-11 Revised:2017-11-08 Online:2018-05-15 Published:2017-11-08
  • Supported by:
    National Natural Science Foundation of China (11772261); Aeronautical Science Foundation of China (2016ZA53011)

摘要: 研究发展高效实用的低音爆优化设计方法,对于新一代低音爆超声速客机的研制具有重要的理论意义和应用价值。目前国内外发展的低音爆优化方法主要包括遗传算法(GA)和基于Adjoint的梯度优化。遗传算法虽然具有较强的全局优化能力,但其优化效率较低,无法很好满足实际应用的需要;而梯度优化虽然优化效率高,但易陷入局部最优。将最新发展的代理优化算法与音爆预测方法相结合,发展了一种具有全局优化能力的高效低音爆优化设计方法。首先,概述了所采用的线性音爆预测方法,并用NASA超声速圆锥体模型进行验证,表明其计算效率高、预测精度可满足飞行器初步设计的需要。其次,对所采用的代理优化(SBO)方法进行了概述,包括试验设计、代理模型建模、优化加点准则和收敛标准等。再次,运用所发展的方法开展了NASA多段圆锥体模型的低音爆优化设计算例研究,并与遗传算法和梯度优化的结果进行了比较,表明其优化效率比遗传算法提高了2个量级以上,且优化结果优于梯度方法。最后,将所发展的方法应用于AIAA音爆预测大会提供的翼身组合体外形(69°后掠三角翼)的低音爆优化设计,将远场音爆N型波峰值减少了27.4%,表明所发展的方法在复杂外形低音爆优化设计中具有很好的应用潜力。

关键词: 音爆, 超声速客机, 优化设计, 代理模型, 高效全局优化

Abstract: It is of great significance to develop efficient numerical optimization methods for low boom design of future supersonic transport aircrafts. To this end, researchers have developed the methods of combining sonic boom prediction with Genetic Algorithm (GA), gradient-based optimization using an Adjoint approach, etc. However, GA has suffered from the prohibitive computational cost for high-dimensional design optimization, and gradient-based optimization can become trapped into a local optimum. This paper proposes to use efficient surrogate-based global optimization towards more effective low sonic boom design. First, the fundamentals of the Whitham theory are introduced, and a comparison of the predicted pressure signals with experimental data shows that the theory is efficient and reasonably accurate for preliminary design of a supersonic transport aircraft. Second, the framework of Surrogate-Based Optimization (SBO) is introduced, including the key elements such as design of experiments, surrogate modeling, infill-sampling criteria and convergence criteria, etc. Third, the proposed methodology of low sonic boom design optimization using SBO is verified by a benchmark sonic boom model of the NASA stepped cone. The comparative study shows that the efficiency of the proposed method is two-orders higher than that of GA, and the optimization results are apparently better than that obtained by the gradient-based method. Finally, a wing-body configuration (69° sweepback delta wing body) taken from the first sonic boom prediction workshop of AIAA is optimized by using the proposed method, and 27.4% reduction of overpressure of the far-field N-wave is achieved. This test demonstrates the great potential of applying the surrogate-based optimization to low boom design of more complex configurations.

Key words: sonic boom, supersonic transport, design optimization, surrogate model, efficient global optimization

中图分类号: