叶片气动优化仿真数据的数据挖掘应用研究

叶片气动优化设计过程中产生的海量过程仿真数据中隐含着丰富的领域设计知识，为了获取其中隐含的设计知识，将基于粗糙集的决策树数据挖掘方法应用到叶片气动过程仿真数据的知识挖掘中。以跨音速压气机转子叶片NASA Rotor37气动优化设计为例，利用K-Means聚类分析对仿真数据进行离散化处理，采用粗糙集属性重要性算法进行属性约简，采用决策树算法构建叶片优化设计知识决策树，挖掘出叶片优化设计变量弯扭、周向积迭线的控制点偏移量与目标函数总压损失系数之间隐含的设计规则。结果表明，基于粗糙集的决策树的数据挖掘技术为叶片气动优化设计领域知识获取提供了一条有效的新途径。

Applied research on simulation data of blade optimization designing based on data mining

A large amount of implicit discipline knowledge is embedded in the aerodynamic optimization simulation data. In order to acquire this knowledge, the data mining method of decision tree algorithm based on rough set is applied to extracting from the aerodynamic simulation data of the transonic rotor. Taking NASA Rotor37 as an example, using the improved K-Means algorithm for discretization of continuous attributes, the discretization attribute is reduced by using rough set theory. The decision tree is established by taking bending, circumferential stacking line control point offset as condition attributes, and the total pressure loss coefficient as decision attribute. The designing rule of the transonic rotor blade optimization designing is acquired. The results show that the method of knowledge extraction is feasible in the field of compressor optimization design.