摘要
针对高强度板成形后回弹大的问题,以工件回弹前后对应节点的位移偏差和等效塑性应变裕度最小化为目标,以板料最大增厚率和最小减薄率为约束条件,建立基于车身侧外板回弹控制的工艺面多目标优化模型。采用实验设计(DOE)和径向基函数(RBF)神经网络方法建立优化代理模型,对均匀实验设计方案进行改进以提高优化精度,并与未改进的RBF神经网络和响应面(RSM)代理模型的优化结果进行对比分析。结果表明,建立的多目标优化模型是合理的,改进RBF神经网络代理模型的优化精度较高,在所抽取的满意解中,回弹和等效塑性应变裕度目标函数的相对误差分别为15.9%和2.2%。与实验设计中回弹量最大的样本方案相比,优化后车身侧外板回弹量减少5.149 2 mm。
To solve the springback problem of advanced high strength steel,a multi-objective optimization model was established by taking an automobile sidewall panel as the example,with a shape error function and an equivalent plastic strain function as the multi-objective functions,the maximum thickening and thinning of the panel as constraint conditions.A optimization surrogate model was constructed on the basis of the even experimental design by using design of experiment（DOE） technique and radial basis function（RBF） neural networks method,to improve the accuracy of the optimization result.The optimization result of the surrogate model was compared with those of the unimproved model and response surface model（RSM）.The results have proved the feasibility of the optimization model,with the springback amount decreasing by as much as 5.1492 mm after optimization compared with sample experiments,and the accuracy of the improved surrogate model is also higher.For the selected satisfactory solution,the relative error of shape error function and equivalent plastic strain function are just about 15.9% and 2.2%,respectively.
出处
《武汉科技大学学报》
CAS
2011年第3期223-227,共5页
Journal of Wuhan University of Science and Technology
基金
国家自然科学基金资助项目(50805050)
广东省教育厅广东高校优秀青年创新人才培育资助项目(wym09110)
关键词
回弹
多目标优化
工艺面
代理模型
板料成形
springback
multi-objective optimization
addendum
surrogate model
sheet metal forming
