摘要
为了提高航空发动机刷式密封结构仿真的时效性与适用性,通过CFD计算获得了20个2级刷式密封结构的泄漏量和级间压比训练样本,对设计的神经网络进行训练和数据泛化预测,获得50组上下游刷丝排数和保护间隙组合下的泄漏率、级间不平衡性数据,并讨论了三者的影响及刷丝排数和保护间隙之间的耦合作用。结果表明:7次即可完成神经网络的训练。通过预测数据发现,仅下游参数变化时,随着刷丝排数的增加,保护间隙对级间压比的影响减小,随着保护间隙的减小,刷丝排数对泄漏率的影响也减小;而仅上游参数变化时,参数对级间压比的耦合性不明显。通过泛化数据进行参数优选,泄漏率可达30.688 N•m^(3)/h,基本消除级间不平衡,且密封性能较好。所建立的神经网络适用于预测具有明显耦合性的2级刷式密封结构流动特性。
In order to improve the timeliness and applicability of aeroengine brush seal structure simulation,twenty training samples of leakage and inter-stage pressure ratio of two-stage brush seal structure were obtained by CFD calculation.The designed neural network was trained and predicted by data generalization,and fifty groups of leakage rate and inter-stage imbalance data under the combination of upstream and downstream brush wire rows and protection gap were obtained.The influence of the three as well as the coupling effect be⁃tween brush wire rows and protection gap were discussed.The results show that the training of neural network can be completed in seven times.Through the prediction data,it is found that when the downstream parameters change only,the influence of the protection gap on the inter-stage pressure ratio decreases with the increase of the brush wire rows,and the influence of the brush wire rows on the leakage rate al⁃so decreases with the decrease of the protection gap.When the upstream parameters change only,the coupling of the parameters to the in⁃ter-stage pressure ratio is not obvious.Through parameter optimization based on generalized data,the leakage rate can reach 30.688 N·m^(3)/h,the inter-stage imbalance can be basically eliminated,and the sealing performance is good.The neural network can be used to predict the flow characteristics of two-stage brush seals with obvious coupling.
作者
阎师
胡芳
黄首清
刘守文
YAN Shi;HU Fang;HUANG Shou-qing;LIU Shou-wen(China Academy of Space Technology,Beijing 100094,China;Beijing Key Laboratory of Environmental Reliability Testing Technology for Aerospace Mechanical and Electrical Products,Beijing 100094,China;Beijing Institute of Spacecraft Evironment Engineering,Beijing 100094,China)
出处
《航空发动机》
北大核心
2022年第3期70-75,共6页
Aeroengine
基金
国家自然科学基金(52075043)资助。
关键词
2级刷式密封
神经网络
泄漏
级间不平衡
耦合性
two-stage brush seals
neural network
leakage
inter-stage imbalance
coupling
作者简介
阎师(1981),男,硕士,工程师,从事军工核心能力建设、固定资产管理等工作。E-mail:13716785244@139.com。;通讯作者:胡芳(1988),女,硕士,工程师,从事机电产品环境与可靠性试验研究工作。E-mail:53816969@qq.com。
