采用SHAP的高压涡轮级高维设计空间数据挖掘

杨昭; 郭振东; 苏鹏飞; 汪祺能; 宋立明

doi:10.7652/xjtuxb202506015

您当前的位置：

首页 >

文章列表页 >

采用SHAP的高压涡轮级高维设计空间数据挖掘

—————————————————————— | 更新时间：2025-08-26

- 采用SHAP的高压涡轮级高维设计空间数据挖掘
- Data Mining for High-Dimensional Design Space of High Pressure Turbine Stage Using SHAP
- 西安交通大学学报 2025年59卷第6期页码：144-154
- 作者机构：
  
  1.西安交通大学能源与动力工程学院， 710049，西安
  2.东方汽轮机有限公司清洁高效透平动力装备全国重点实验室， 618000，四川德阳
- 作者简介：
  
  杨昭（2000—），女，硕士生；
  宋立明，男，教授，博士生导师。
- 基金信息：
  
  国家科技重大专项资助项目(2019-Ⅱ-0008-0028)
- DOI：10.7652/xjtuxb202506015
  中图分类号： TK471
- 收稿：2024-11-22，
  
  网络首发：2025-02-25，
  
  纸质出版：2025-06-10
- 稿件说明：
移动端阅览
杨昭, 郭振东, 苏鹏飞, 等. 采用SHAP的高压涡轮级高维设计空间数据挖掘[J]. 西安交通大学学报, 2025,59(6):144-154.

YANG Zhao, GUO Zhendong, SU Pengfei, et al. Data Mining for High-Dimensional Design Space of High Pressure Turbine Stage Using SHAP[J]. Journal of Xi’an Jiaotong University, 2025, 59(6): 144-154.
杨昭, 郭振东, 苏鹏飞, 等. 采用SHAP的高压涡轮级高维设计空间数据挖掘[J]. 西安交通大学学报, 2025,59(6):144-154. DOI： 10.7652/xjtuxb202506015.

YANG Zhao, GUO Zhendong, SU Pengfei, et al. Data Mining for High-Dimensional Design Space of High Pressure Turbine Stage Using SHAP[J]. Journal of Xi’an Jiaotong University, 2025, 59(6): 144-154. DOI： 10.7652/xjtuxb202506015.

摘要

为厘清涡轮级高维设计空间中各变量对级性能的影响，基于数据挖掘方法SHAP对GE-E3高压涡轮第一级的93个设计变量进行了知识挖掘。除常规数据挖掘工作中显著变量对总体性能的影响分析外，发展了显著变量对涡轮级沿叶高方向性能分布影响的分析方法，可视化表示了改善级总体性能的设计变量所影响的涡轮级的具体位置；同时，充分发挥SHAP局部解释的优势，在设计空间中选取典型样本进行归因分析，研究了各设计变量在样本性能指标变化中所发挥的作用。研究发现，对于高压涡轮级，影响级效率的显著变量包括有效出气角、静叶三维积叠参数、叶片吸力面前缘附近样条控制点等。基于GE-E3高压涡轮数据集进行数据挖掘归纳得到涡轮级设计准则：减小动叶中间截面有效出气角、增大静叶中间截面有效出气角，静叶三维积叠点周向顶部位置、周向中间位置向压力面偏移，静叶中间截面、动叶中间截面吸力面前缘附近控制点均向叶片变薄方向移动。遵循设计准则得到的最终设计使级效率提高了0.65%。

Abstract

In order to clarify the influence of various variables on the performance of turbine stage in high-dimensional design space

the knowledge mining of 93 design variables of GE-E3 high-pressure turbine first stage was carried out based on the data mining method SHAP. In addition to the analysis of the influence of significant variables on the overall performance in conventional data mining work

the analysis method of the influence of significant variables on the performance distribution along spanwise of turbine stages was developed

and the specific positions of turbine stage affected by the design variables to improve the overall performance of the stage were visualized. At the same time

SHAP’s advantage in local explanations was fully leveraged

attribution analysis on typical samples in the design space was conducted

and the role of each design variable in the change of performance indicators was studied. The results showed that for the high-pressure turbine stage

significant variables affecting stage efficiency included the effective output angle

three-dimensional stacking parameters of the stator

and the spline control points near the suction leading edge of blades. Based on GE-E3 high-pressure turbine dataset

data mining was conducted to obtain turbine stage design criteria: When the effective output angle of the middle section of the rotor is reduced and the effective output angle of the middle section of the stator is increased

the circumferential top and middle positions of the three-dimensional stack points of the stator will be shifted to the pressure surface

and the control points near the suction leading edge of the middle section of both the stator and rotor will move in the direction of blade thinning. The new design according to the design criteria has improved the stage efficiency by 0.65%.

关键词

Keywords

references

BENINI E . Three-dimensional multi-objective design optimization of a transonic compressor rotor [J ] . Journal of Propulsion and Power , 2004 , 20 ( 3 ): 559 - 565 .

汪祺能 , 宋立明 , 郭振东 , 等 . 采用动态交互作用分析的叶轮机械优化算法研究 [J ] . 西安交通大学学报 , 2023 , 57 ( 7 ): 139 - 150 .

WANG Qineng , SONG Liming , GUO Zhendong , et al . Study on turbomachinery optimization algorithm based on dynamic interaction analysis [J ] . Journal of Xi'an Jiaotong University , 2023 , 57 ( 7 ): 139 - 150 .

VIANA F A C , SIMPSON T W , BALABANOV V , et al . Special section on multidisciplinary design optimization: metamodeling in multidisciplinary design optimization: how far have we really come? [J ] . AIAA Journal , 2014 , 52 ( 4 ): 670 - 690 .

WANG Qi , YANG Li , RAO Yu . Establishment of a generalizable model on a small-scale dataset to predict the surface pressure distribution of gas turbine blades [J ] . Energy , 2021 , 214 : 118878 .

JU Yaping , QIN Ruihong , KIPOUROS T , et al . A high-dimensional design optimisation method for centrifugal impellers [J ] . Proceedings of the Institution of Mechanical Engineers: Part A Journal of Power and Energy , 2016 , 230 ( 3 ): 272 - 288 .

LI Zhihui , ZHENG Xinqian . Review of design optimization methods for turbomachinery aerodynamics [J ] . Progress in Aerospace Sciences , 2017 , 93 : 1 - 23 .

BAERT L , DUMONT C , BEAUTHIER C , et al . Multidisciplinary design of a low-noise propeller: part Ⅱ efficient aero-acoustic-mechanical design methodology exploiting surrogate models in an adaptive design space [C ] // ASME Turbo Expo 2020: Turbomachinery Technical Conference and Exposition . New York, NY, USA : ASME , 2020 : V02DT38A023 .

CHIBA K , JEONG S , OBAYASHI S , et al . Data mining for multidisciplinary design space of regional-jet wing [C ] // 2005 IEEE Congress on Evolutionary Computation . Piscataway, NJ, USA : IEEE , 2005 : 2333 - 2340 .

IULIANO E , QUAGLIARELLA D . Proper orthogonal decomposition, surrogate modelling and evolutionary optimization in aerodynamic design [J ] . Computers & Fluids , 2013 , 84 : 327 - 350 .

段焰辉 , 吴文华 , 范召林 , 等 . 基于本征正交分解的气动优化设计外形数据挖掘 [J ] . 物理学报 , 2017 , 66 ( 22 ): 220203 .

DUAN Yanhui , WU Wenhua , FAN Zhaolin , et al . Proper orthogonal decomposition-based data mining of aerodynamic shape for design optimization [J ] . Acta Physica Sinica , 2017 , 66 ( 22 ): 220203 .

OYAMA A , NONOMURA T , FUJII K . Data mining of Pareto-optimal transonic airfoil shapes using proper orthogonal decomposition [J ] . Journal of Aircraft , 2010 , 47 ( 5 ): 1756 - 1762 .

SHI Yayun , LAN Xiayu , WANG Yiwen , et al . Data mining-based design space exploration and optimization for tandem airfoils [J ] . International Journal of Aerospace Engineering , 2023 , 2023 ( 1 ): 3513507 .

SONG Liming , GUO Zhendong , LI Jun , et al . Research on metamodel-based global design optimization and data mining methods [J ] . Journal of Engineering for Gas Turbines and Power , 2016 , 138 ( 9 ): 092604 .

ADJEI R A , FAN Chengwei , WANG Weizhe , et al . Multidisciplinary design optimization for performance improvement of an axial flow fan using free-form deformation [J ] . Journal of Turbomachinery , 2021 , 143 ( 1 ): 011003 .

郭振东 , 王杰 , 陈云 , 等 . 融合多源数据挖掘信息的非轴对称端壁/叶身联合成型设计空间知识挖掘 [J ] . 西安交通大学学报 , 2024 , 58 ( 1 ): 30 - 41 .

GUO Zhendong , WANG Jie , CHEN Yun , et al . Exploration of the design space for non-axisymmetric endwall/blade through integration of multi-source data mining techniques [J ] . Journal of Xi'an Jiaotong University , 2024 , 58 ( 1 ): 30 - 41 .

QIN Ruihong , JU Yaping , SPENCE S , et al . Metamodel-driven data mining model to support three-dimensional design of centrifugal compressor stage [J ] . Journal of Turbomachinery , 2021 , 143 ( 12 ): 121013 .

BAERT L , CHÉRIÈRE E , SAINVITU C , et al . Aerodynamic optimization of the low-pressure turbine module: exploiting surrogate models in a high-dimensional design space [J ] . Journal of Turbomachinery , 2020 , 142 ( 3 ): 031005 .

LUNDBERG S M , ERION G , CHEN H , et al . From local explanations to global understanding with explainable AI for trees [J ] . Nature Machine Intelligence , 2020 , 2 ( 1 ): 56 - 67 .

LUNDBERG S M , LEE S I . A unified approach to interpreting model predictions [C ] // Proceedings of the 31st International Conference on Neural Information Processing Systems . Red Hook, NY, USA : Curran Associates Inc. , 2017 : 4768 - 4777 .

PALAR P S , ZUHAL L R , SHIMOYAMA K , et al . Shapley additive explanations for knowledge discovery via surrogate models [C ] // AIAA SCITECH 2023 Forum . Reston, VA, USA : AIAA , 2023 : AIAA 2023-0332 .

SHEN Yang , HUANG Wei , YAN Li , et al . An automatic visible explainer of geometric knowledge for aeroshape design optimization based on SHAP [J ] . Aerospace Science and Technology , 2022 , 131 ( Part A ): 107993 .

ZHANG Xiawen , JU Yaping , LI Zhen , et al . Metamodel-interpreted data mining for stability and efficiency enhancement of multistage axial-flow compressors [J ] . Journal of Turbomachinery , 2023 , 145 ( 4 ): 041001 .

PALAR P S , DWIANTO Y B , ZUHAL L R , et al . Multi-objective design space exploration using explainable surrogate models [J ] . Structural and Multidisciplinary Optimization , 2024 , 67 ( 3 ): 38 .

PALAR P S , SHIMOYAMA K , OBAYASHI S . Polynomial-based Shapley additive explanations for design exploration in turbomachinery [C ] // AIAA SCITECH 2024 Forum . Reston, VA, USA : AIAA , 2024 : AIAA 2024-1226 .

WANG Junying , HE Xiao , WANG Baotong , et al . Shapley additive explanations of multigeometrical variable coupling effect in transonic compressor [J ] . Journal of Engineering for Gas Turbines and Power , 2022 , 144 ( 4 ): 041015 .

WANG Qing , GUO Zhendong , SONG Liming , et al . A panoramic aerodynamic performance prediction method for turbomachinery cascades using transformer-enhanced neural operator [J/OL ] . Chinese Journal of Aeronautics [ 2024-12-20 ] . https://doi.org/10.1016/j.cja.2025.103473 https://doi.org/10.1016/j.cja.2025.103473 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于免疫记忆克隆的特征选择

高效的用户访问预测新算法

基于视觉原理的分类算法

应用相似度测量的图离群点检测方法

融合多源数据挖掘信息的非轴对称端壁/叶身联合成型设计空间知识挖掘