显著变量识别与高温叶片多学科设计优化方法

Multi-Disciplinary Optimization Design of High Temperature Blade with Significant Variables Recognition

关键词:  高温叶片  多学科优化  显著变量识别  数据挖掘  共轭换热

Key words:  High temperature blade  Multi-disciplinary optimization  Significant variables recognition  Data mining  Conjugate heat transfer

摘要:

针对高温叶片气热多学科优化设计问题中设计变量过多造成的维数灾难问题，提出了基于数据挖掘技术的显著变量识别方法。采用显著变量识别方法剔除了对高温叶片Mark II气热性能影响小的设计变量，使设计变量个数从36个减少为15个。通过耦合共轭换热分析方法、三维叶片及冷却系统参数化方法以及自适应多目标差分进化算法，建立了高温叶片多学科多目标设计优化系统。基于显著变量识别方法获得的设计变量，完成了Mark II型叶片的气热性能多学科设计优化。优化获得了9个Pareto解，典型Pareto解的气热分析结果表明，优化后叶片的气热性能明显优于原始叶片，验证了基于数据挖掘技术的高温叶片多学科设计方法的有效性。

Abstract:

A significant variable detection method based on data mining technique is proposed to solve the problem of‘curse-of-dimensionality’caused by too many design variables in the aero-thermal multidisciplinary design optimization of high temperature blade. By employing the proposed significant variable detection method，design variables which have small effects on the aero-thermal performance of Mark II are eliminated，and the number of design variables for optimization process is decreased from 36 to 15. Applying Conjugate Heat Transfer analysis，3D blade profile and cooling system parameterization as well as Self-adaptive Multi-objective Differential Evolution algorithm (SMODE)，a multi-objective and multi-disciplinary optimization platform for high temperature blade was developed. After optimization，9 optimal Pareto solutions are obtained. The aero-thermal performance of typical optimal Pareto solutions is analyzed. It is indicated that the performance of optimal designs is significant better with respect to the reference design. Therefore，the effectiveness of the data mining based multidisciplinary design method for high temperature blade is demonstrated.