航空学报 > 2025, Vol. 46 Issue (14): 130910-130910   doi: 10.7527/S1000-6893.2024.30910

基于NSGA-Ⅲ-SAM算法的冲压发动机喷管性能预测

闵科, 蔡泽君, 张加乐, 朱呈祥()   

  1. 厦门大学 航空航天学院,厦门 361102
  • 收稿日期:2024-07-05 修回日期:2024-08-02 接受日期:2024-08-11 出版日期:2024-08-21 发布日期:2024-08-20
  • 通讯作者: 朱呈祥 E-mail:chengxiang.zhu@xmu.edu.cn
  • 基金资助:
    国家自然科学基金(U21B6003);国家自然科学基金(12202372);1912项目

Scramjet nozzle performance prediction based on NSGA-Ⅲ-SAM algorithm

Ke MIN, Zejun CAI, Jiale ZHANG, Chengxiang ZHU()   

  1. School of Aerospace Engineering,Xiamen University,Xiamen 361102,China
  • Received:2024-07-05 Revised:2024-08-02 Accepted:2024-08-11 Online:2024-08-21 Published:2024-08-20
  • Contact: Chengxiang ZHU E-mail:chengxiang.zhu@xmu.edu.cn
  • Supported by:
    National Natural Science Foundation of China(U21B6003);1912 Project

摘要:

喷管的排气特性直接影响冲压发动机的整体工作性能,如何实现喷管性能的有效预测,预防其发生急剧变化对于发动机的稳定工作十分关键。在不同飞行条件下对三维非对称喷管进行了数值模拟,搭建了不同马赫数和落压比下的喷管性能预测数据集。考虑到传统多目标优化算法易陷入局部最优的不足,提出了一种模拟退火式变异优化的非支配排序遗传算法(Non-dominated Sorting Genetic Algorithm-Ⅲ-Simulated Annealing Mutation, NSGA-Ⅲ-SAM),用于提取喷管壁面的最优压力特征。通过以最优压力特征数据作为输入,轴向推力系数、俯仰力矩系数和升力系数作为输出,建立了基于一维卷积神经网络(One Dimension-Convolutional Neural Network, 1D-CNN)的喷管性能参数预测模型,并利用马赫数4.5~6.0过膨胀状态数据对模型进行验证。结果表明,NSGA-Ⅲ-SAM算法所提取的最优压力位置能够使模型具备高精度、快速预测的性能,各性能参数平均绝对误差整体在0.5%范围内,最大绝对误差不超过0.8%,平均预测时间仅需0.6 ms左右。所构建的预测模型及方法能够为喷管性能监测及排气工况调节奠定可靠的技术基础。

关键词: 性能预测, 三维非对称喷管, NSGA-Ⅲ-SAM算法, 壁面最优压力特征, 1D-CNN模型

Abstract:

The exhaust characteristics of a nozzle directly affect the overall performance of the scramjet engine. It is crucial to effectively predict the nozzle performance to prevent drastic changes for the stable operation of the engine. Numerical simulations of three-dimensional asymmetric nozzles under different flight conditions were conducted to build a dataset for predicting nozzle performance at various Mach numbers and nozzle pressure ratios. Considering the limitations of traditional multi-objective optimization algorithms, a Non-dominated Sorting Genetic Algorithm-Ⅲ-Simulated Annealing Mutation (NSGA-Ⅲ-SAM) was proposed to extract the optimal wall pressure measurement points for the nozzle. By using the optimal pressure characteristic data as input and the axial thrust coefficient, pitching moment coefficient, and lift coefficient as outputs, a nozzle performance parameter prediction model based on the One Dimension-Convolutional Neural Network (1D-CNN) was established and validated by the data of over-expanded states at the Mach numbers from 4.5 to 6.0. The results show that the optimal pressure positions extracted by the NSGA-Ⅲ-SAM algorithm enable the model to have high-precision and rapid prediction performance, with the overall average absolute error of all performance parameters being within 0.5%, the maximum absolute error not exceeding 0.8%, and the average prediction time being only about 0.6 ms. The proposed prediction model and method provide a reliable technical foundation for monitoring nozzle performance and adjusting exhaust conditions.

Key words: performance prediction, three-dimensional asymmetric nozzle, NSGA-Ⅲ-SAM algorithm, wall-optimized pressure measurement point, 1D-CNN model

中图分类号: