基于BB-递归核函数SVR算法的U型折弯件模型参数优化研究

Parameter optimization of U-bend model based on SVR algorithm with BB-recursive Kernel function

影响U型折弯件回弹的因素众多,工件尺寸、力学性能、负载条件、材料各向异性等相互耦合,表现出高度复杂的非线性,从而导致回弹预测结果的不确定性。本研究以板料折弯件回弹后的张开角(α)为目标函数,构建一个递归核函数支持向量回归(SVR)模型,并部署到分支界限法(BB)中,从而筛选出维度为4的最优的特征变量参数子集,其决定系数(R^(2))为0.982147,均方误差(MSE)为0.00433,模型预测精度相对较高。算法优化得到的折弯件参数为:厚度(t)为12 mm,上模宽度(d)为90 mm,上模圆角半径(r)为9 mm,载荷速度(v)为10 mm/s。BB递归核函数SVR算法、有限元模拟和实际测量的α分别为16.3°、17.5°和18.2°,尽管有限元结果更接近于实际值,但是BB递归核函数SVR算法可以为有限元模拟提供筛选出的参数(t,d,r,v)的数据,以快速进行模拟并预测张开角α,并实现回弹补偿装置的高效设计。

There are many factors affecting the springback of U-bend bending parts,such as workpiece size,mechanical properties,load conditions and material anisotropy,which are highly nonlinear and lead to the uncertainty of springback prediction.In this study,a support vector regression(SVR)model with recursive kernel function was constructed,and deployed to the branch and bound(BB)method to select the optimal feature variable parameter subset with the dimension of 4.The coefficient of determination(R^(2))was 0.982147 and the mean square error(MSE)was 0.00433,indicating that the prediction accuracy of the model was relatively high.The bending part thickness(t)was 12 mm,the upper die width(d)was 90 mm,the upper die fillet radius(r)was 9 mm,and the load speed(v)was 10 mm/s.The BB recursive kernel SVR algorithm,the finite element simulation,and the actual measuredαare 16.3°,17.5°,and 18.2°,respectively.Although the finite element results are closer to the actual values,the BB recursive kernel SVR algorithm can provide the data of the filtered parameters(t,d,r,v)for the finite element method(FEM)to quickly perform the simulation and predict the angleα,and the efficient design of the springback compensation device is realized.