期刊文献+

预训练模型驱动的固体火箭发动机总体优化设计

Solid rocket motor optimization design driven by pre-trained models
原文传递
导出
摘要 优化设计是固体火箭发动机设计的重要手段。针对现有固体火箭发动机优化设计方法模型调用次数多、优化效率低的问题,提出预训练模型驱动的固体火箭发动机优化设计方法。基于已有设计案例,提出固体火箭发动机几何-性能参数映射方法,将已有仿真结果转化为目标需求下的训练样本,由此构建预训练模型。基于预训练模型进行初始样本点筛选和模型响应预测,结合多可信度代理模型技术,在较少仿真模型调用次数下定位问题全局最优,提升了固体火箭发动机总体优化效率。对翼柱形装药固体火箭发动机进行优化设计,结果表明,基于200个已有设计案例建立预训练模型,只需调用30次高精度仿真模型即可完成优化设计,相比于传统直接优化设计的64次模型调用可大幅减少仿真模型调用次数,提高了设计效率,验证了本方法的高效性和有效性。 Optimization design is an important tool for solid rocket motor design.In view of the problems that the existing solid rocket motor optimization design method has many model calls and low optimization effi-ciency,a pre-trained model-driven solid rocket motor optimization design method is proposed.Based on the ex-isting design cases,the geometry-performance parameter mapping method of solid rocket motor is proposed to transform the existing simulation results into training samples under the target requirements,and thus the pre-training model is constructed.Based on the pre-training model,the selection of initial sample point and predic-tion of model response are performed,and combined with the multi-credibility proxy model technology,the glob-al optimum of the problem is located with fewer simulation model calls,which improves the overall optimization efficiency of the finocyl grain for solid rocket motor.The results show that the optimized design can be completed with only 30 calls to the high-precision simulation model with the pre-training model based on 200 existing de-sign cases,which significantly reduces the number of calls to the simulation model and improves the design effi-ciency compared with the 64 calls to the traditional direct optimization design,and verifies the efficiency and ef-fectiveness of this method.
作者 高经纬 马帅超 彭博 武泽平 张为华 GAO Jingwei;MA Shuaichao;PENG Bo;WU Zeping;ZHANG Weihua(College of Aerospace Science and Engineering,National University of Defense Technology,Changsha 410073,China)
出处 《推进技术》 EI CAS CSCD 北大核心 2024年第7期16-24,共9页 Journal of Propulsion Technology
基金 国家自然科学基金(52005502,52375278)。
关键词 固体火箭发动机 优化设计 代理模型 多可信度模型 案例映射 Solid rocket motor Optimization design Surrogate model Multi-fidelity model Case mapping
  • 相关文献

参考文献6

二级参考文献152

共引文献94

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部