摘要
拉伸是板料冲压成型的关键工艺,回弹是拉伸过程中的控制难点。笔者以板料拉伸为对象,选择材料各项异性系数、硬化指数、压边力、摩擦系数、及凹模圆角为试验参数,借助正交试验工具,进行了冲压成型数值模拟试验,并在此试验基础上建立了基于BP神经网络的回弹预测模型。针对BP算法的不稳定性,易陷入局部最优值的缺陷,设计了一种基于蚁群算法的BP网络优化算法,该优化算法融合了蚁群算法的全局性能、启发式优点和神经网络的泛化性能。利用该算法建模并将其应用于冲压成型回弹预测,结果表明:此模型可将预测误差控制在5%以内,满足工业应用标准。
We first construct a BP neural network model for springback prediction on the basis of numerical simulation.Then,we design an algorithm based on ant colony optimization algorithm and BP neural network（ ACO-BP） to integrate the overall performance and heuristic advantages of ant colony algorithm and the generalization performance of neural networks.Simulation results show that the prediction error of this model can be controlled within 5% to satisfy the requirements of industrial applications.
出处
《机械科学与技术》
CSCD
北大核心
2010年第6期714-718,共5页
Mechanical Science and Technology for Aerospace Engineering
关键词
板料
冲压成型
回弹
神经网络
预测
蚁群算法
sheet metal
stretch forming
springback
neural networks
prediction
ant colony optimization algorithm
