基于数值模拟和神经网络的圆台件渐进成形回弹量预测研究
Springback Prediction of Truncated Cone Formed by Incremental Forming Based on Numerical Simulation and Neural Network
李军超, 张 旭, 胡建标
点击:3940次 下载:0次
作者单位:(重庆大学 材料科学与工程学院, 重庆 400044)
中文关键字:渐进成形; 有限元; 神经网络; 回弹预测
英文关键字:incremental forming; finite element; neural network; springback prediction
中文摘要:以圆台件为研究对象,运用DYNAFORM对板材渐进成形加工过程和回弹过程进行了数值模拟,并结合正交实验设计方法,对渐进成形回弹影响因素的敏感性进行了深入分析;同时,分别建立了回弹量的线性回归预测模型和BP神经网络预测模型。结果表明,各因素对回弹量的敏感程度由高到低按顺序依次为成形半顶角、加工高度、工具头直径和层进给量;BP神经网络模型可有效逼近工艺参数与回弹量间的非线性关系,预测精度相对误差2%以内。
英文摘要: Based on DYNAFORM, the forming and springback process for a truncated cone in incremental sheet forming process were numerically simulated. Then, combined with orthogonal test, the sensibilities of influencing factors in incremental sheet forming were researched. Meanwhile, a linear regression model and a BP neural network model for predicting the springback quantity were set up. The results show that the main factors correlating with springback are in the sequence of wall angle, processing height, tool diameter and depth increment; the nonlinearity relationship between the processing parameters and springback can be approached through the proposed neural network model and the relative prediction errors are within 2%.