基于多代理模型透平叶栅端壁流热复合优化设计

Aero-Thermal Coupled Optimization of Turbine Cascade Endwall Based on Multi-Surrogate Model

基于叶栅端壁压力分布特性,采用控制点关联函数,构建10个设计变量的NURBS参数曲面,控制二次流和降低端壁传热系数。应用最优拉丁超立方实验设计,建立设计样本,并进行CFD性能评估,以端壁面平均Nu数和总压损失系数作为双重目标函数,构建多个代理模型的组合代理模型,用多目标遗传算法优化获得Pareto前沿。结果表明:在一定优化计算成本上,基于组合代理模型进行多轮优化,能够获得更满意的最优解。获得的最优端壁构型中,端壁平均Nu数可降低4.4%,叶栅出口总压损失可减小2.3%。

In accordance with the surface pressure distribution,endwall is contoured to reduce secondary flow and wall heat transfer with a NURBS surface of 10 design variables connected by trigonometric functions.Design samples are built by the optimal Latin Hypercube method and CFD evaluated.Their total pressure loss and averaged Nusselt number on the wall are the double minimum ob.jectives.Ensemble surrogate models are constructed to search the Pareto frontier of optimization.Results show that the optimal solution is more satisfactory from a multi-round search.Among available solutions,it is achieved with a reduction of 4.4% of endwall averaged Nusselt number and2.3% of total pressure loss.