文献类型：期刊文献

中文题名：支持向量机决策树模型在气动弹性分析中的应用

英文题名：Application of Support Vector Machine Decision Tree Model in Aeroelasticity Analysis

作者：徐旺丁 张兵 王华毕 王云海

第一作者：徐旺丁

机构：[1]合肥工业大学机械工程学院;[2]贵州理工学院机械工程学院

第一机构：合肥工业大学机械工程学院,安徽合肥230009

年份：2017

卷号：47

期号：6

起止页码：36-40

中文期刊名：航空计算技术

外文期刊名：Aeronautical Computing Technique

收录：CSTPCD

基金：国家自然科学基金项目资助(11302065);贵州省科技厅项目:黔科合基础([2017]1408号)

语种：中文

中文关键词：气动弹性;降阶模型;支持向量机;决策树;颤振

外文关键词：aeroelasticity ; reduced order model ; support vector machine ; decision tree ; flutter

摘要：标准的支持向量机具有坚实的理论基础,而且在回归问题上泛化能力较高,但是回归速度比较慢,在很多情况下不能满足实际模式回归问题的要求。决策树效率高,能在较短时间对大型数据做出良好决策,每一次预测的最大计算次数不超过决策树的深度,但在处理特征关联性比较强的数据时表现得不是太好。针对CFD/CSD耦合计算气动弹性特性的精度和高效性问题,提出并构建了基于CFD技术的非定常气动力支持向量机决策树降阶模型,并应用于非定常气动弹性分析。选择二维NACA0012翼型进行颤振边界的预测以及选用NACA64A010翼型预测LCO特性,结果表明,降阶模型的计算结果接近CFD/CSD耦合计算的结果,同时提高了计算效率。
The standard support vector machine has high generalization ability for the regression problem with its solid theoretical foundations. Its major disadvantage is the relatively slow regression rate and the requirements of the actual mode regression problem. The decision tree is an efficient method which can make good decisions about large data in a short time. Its main advantage is that the maximum number of calculations for each prediction does not exceed the depth of the decision tree, but opposite it is not very good for the data with strong correlation. A new reduced order model of unsteady aerodynamic based on CFD,which called Support Vector Machine- Decision Tree, is proposed and applied to unsteady aerodynamic elasticity analysis to accelerate the CFD/CSD coupling calculation without losing accuracy. The simulations of the twodimensional NACA0012 airfoil flutter boundary and the NACA64A010 airfoil LCO characteristics were examined, the results indicated that the accuracy of the new reducedorder model is close to the CFD / CSD coupling calculation, and the computational efficiency is significant improved.