基于机器学习建模的液体火箭发动机喷管内型面优化设计  

Profile design and optimization of liquid rocket engine nozzle based on machine learning

作　　者：李晨沛 周晨初 高玉闪[1] 胡海峰[1] Li Chenpei;Zhou Chenchu;Gao Yushan;Hu Haifeng(Xi′an Aerospace Propulsion Institute,Xi′an 710100,China)

机构地区：[1]西安航天动力研究所,陕西西安710100

出　　处：《网络安全与数据治理》2024年第2期42-48,共7页CYBER SECURITY AND DATA GOVERNANCE

摘　　要：喷管是液体火箭发动机产生推力的重要部件。喷管型面的结构将直接影响燃烧所产生的燃气在喷管中的流动情况,进而对发动机的性能产生影响。采用B样条曲线对抛物面型线进行参数化,计算样本集的流体动力学(Computational Fluid Dynamics,CFD)流场,以比冲为优化变量对喷管性能进行评估。研究表明,基于代理模型优化得到的喷管内型面结构与特征线法计算结果基本一致,比冲计算结果相当,最大误差为0.28%。通过代理模型和网格变形方法,可实现液体火箭发动机喷管内型面优化设计,提高优化效率。The nozzle is an important part of the liquid rocket engine to provide the thrust.The structure of the nozzle profile could directly affect the flow of combustion gas in the nozzle,and then impact on the performance of the engine.In this paper,B-spline curve is used to construct the paraboloid profile of the nozzle.Based on the Computational Fluid Dynamics(CFD)flow field of sample set,the nozzle performance is evaluated with specific impulse as the optimal variable.The results show that the optimized nozzle profile obtained by the surrogate model is consistent with that by the characteristic line method,and the maximum error is 0.28%.In this work,the internal profile design and optimization is realized via the surrogate model and mesh auto-deformation method,and the optimization efficiency is improved.

关 键 词：内型面 比冲 机器学习 网格变形 

分 类 号：V434.2[航空宇航科学与技术—航空宇航推进理论与工程]