Abstract: Aiming at the characteristics of Aeronautical thin-walled part (ATWP) such as thin wall and low rigidity,to optimize the fixture locating layout and improve the fixture design efficiency,a prediction model for ATWP fixture locating layout design and optimization was developed.The training and testing data sets were generated by running only a few times of Finite Element Analysis (FEA) on the design sites standing for different fixture locating layouts selected through Latin Hypercube Design (LHD).The locating layout optimization prediction model based on Support Vector Regression (SVR) was built by learning from the limited training sample set to approximate the implicit function relationship between fixture locating layout and concerned ATWP strain energy,and meanwhile was compared with Radial Basis Function Neural Network (RBFNN) in terms of prediction accuracy on the same testing sample set.An aircraft fuselage skin case was studied to verify the proposed method.

Key words: aeronautical thin-walled part, fixture design, locating layout, prediction model, support vector regression, radial basis function neural network

摘要： 针对航空薄壁件厚度小、刚性弱和制造变形易超差等特点,为了优化夹具定位点布局并减小有限元计算成本、提高夹具设计效率,提出一种面向变形控制的曲面薄壁件夹具定位布局优化预测模型。基于“N-2-1”定位原理,以夹具定位点位置作为设计变量,以整体应变能作为定位布局设计的评价指标,采用拉丁超立方设计和有限元分析计算整体应变能并生成有限样本集;通过学习样本集构建基于支持向量回归机的定位布局优化预测模型,得到定位布局方案与整体应变能之间的非线性函数映射关系,并与径向基神经网络进行预测精度对比。以某一飞机的机身薄壁件作为应用实例,验证了所提预测模型的正确性和有效性。

关键词: 航空薄壁件, 夹具设计, 定位布局, 预测模型, 支持向量回归机, 径向基神经网络