基于序贯随机克里金模型的涡轮叶片力学性能预报  

Prediction of Mechanical Performance of Turbine Blade based on Sequential Stochastic Kriging Model

作　　者：张益敢 胡杰翔 刘华坪 董爱华 ZHANG Yi-gan;HU Jie-xiang;LIU Hua-ping;DONG Ai-hua(School of Aeronautics and Astronautics,Huazhong University of Science and Technology,Wuhan 430074,China;Harbin Electric Corporation,Harbin 150028,China)

机构地区：[1]华中科技大学航空航天学院,武汉430074 [2]哈尔滨电气集团有限公司,哈尔滨150028

出　　处：《汽轮机技术》2022年第4期241-245,266,共6页Turbine Technology

基　　金：中国博士后科学基金(2020M682396)。

摘　　要：针对基于近似模型的涡轮叶片设计过程中一次性建模方法未能充分利用样本点信息导致近似建模精度较低的弊端,提出了基于积分均方误差(Integral mean square error,IMSE)减小的序贯建模方法。以随机克里金模型(Stochastic Kriging,SK)作为近似模型考虑响应不确定性的影响,并通过添加使近似模型IMSE期望减小最大的样本点序贯更新近似模型。“克里金信任”(Kriging believer)策略被应用于序贯选取样本点优化问题的求解过程,以减小仿真成本。选用了一个数值算例和涡轮叶片力学性能预报工程算例验证所提出方法的性能,结果表明,所提出的方法与一次性建模方法相比,在相同仿真成本下能构建全局和局部精度更优的近似模型。To consider the disadvantage of the one-shot metamodeling method that it cannot make full use of the information from the sample points during the surrogate model-based turbine blade design process,a sequential metamodeling method based on the reduction of integral mean square error(IMSE)was proposed.The stochastic kriging(SK)model was selected as the surrogate model to consider the influence of the uncertainty of the responses.The SK model was updated sequentially by adding sample points that could maximize the expected reduction of IMSE.The"Kriging believer"strategy is applied to the optimization problem for selecting the sequential samples to reduce the simulation cost.A numerical example and a turbine blade engineering problem was selected to test the performance of the proposed method.Results show that compared with the one-shot metamodeling method,the constructed surrogate model with the proposed method has higher global and local accuracy under the same simulation cost.

关 键 词：涡轮叶片 随机克里金 序贯建模 响应不确定性 置信区间 

分 类 号：TH11[机械工程—机械设计及理论]