基于椭圆基神经网络模型的排汽缸优化设计
Optimization Design of Exhaust Hood Based on Elliptical Basis Functions Neural Network Model
彭姝璇 1付经伦2
作者信息
- 1. 中科南京未来能源系统研究院燃气轮机数字化中心,南京 211135
- 2. 中科南京未来能源系统研究院燃气轮机数字化中心,南京 211135;中科院工程热物理研究所先进燃气轮机实验室,北京 100190;中国科学院大学,北京 100049;中国科学院大学南京学院,南京 211135;中国科学院先进能源动力重点实验室,北京 100190
- 折叠
摘要
本文以最优拉丁超立方试验设计方法构建样本库,采用椭圆基神经网络构建近似模型,结合多岛遗传算法搭建全自动优化设计系统.以排汽缸静压恢复系数最大化作为优化目标,采用此系统对某型号汽轮机机组排汽缸导流环型线进行气动优化设计.结果表明:各设计变量中导流环起始角度对排汽缸性能影响最显著.优化后排汽缸扩压器、排汽蜗壳性能均有改善,排汽缸静压恢复系数提高7.91％,总压损失系数下降5.98％.在其他运行工况,优化结构性能仍优于原型,且在THA工况下性能改善显著,排汽缸静压恢复系数提高13.45％,总压损失系数下降10.44％,表明该优化系统对汽轮机低压排汽缸气动设计有效可行.
Abstract
The optimal Latin hypercube experimental design method was used to build the sample database,the approximate model based on the elliptical basis functions neural network model and the global optimization of multi-island genetic algorithm were employed to develop a fully automatic optimization design system.Taking the maximization of static pressure recovery coefficient as the optimization goal,the system was applied to optimize the exhaust hood diffuser guide ring profile in the aerodynamic design process.The results show that the initial angle of the guide ring has the most significant impact on the performance of the exhaust hood.The performances of the diffuser and the collector after optimization are improved,the static pressure recovery coefficient of the exhaust hood is increased by 7.91％,and the total pressure loss coefficient is reduced by 5.98％.The improvement is the most significant for the THA conditions.Under the condition of THA,the enhancement in the static pressure recovery coefficient is 13.45％and the decrement in the total pressure loss coefficient is 10.44％.It means that the optimization system is effective and feasible for the aerodynamic design of the steam turbine low-pressure exhaust hood.
关键词
椭圆基神经网络/排汽缸/导流环/静压恢复系数/总压损失系数Key words
elliptical basis functions neural network/exhaust hood/diffuser guide ring/static pressure recovery coefficient/total pressure loss coefficient引用本文复制引用
基金项目
国家自然科学基金项目(51776201)
国家重大科技专项(J2019-Ⅱ-0017-0038)
出版年
2024
NETL
NSTL
维普
万方数据