摘要
利用GS理论和神经网络遗传算法函数寻优法,搭乘非线性有限元分析软件Dynaform,对轻型卡车左后侧围外板拉延成形过程工艺参数寻优,以解决该零件在成形过程中出现的破裂和过度减薄质量缺陷。将GS理论和正交试验设计相结合,获得各工艺参数组合下的最大减薄率,并对获取的数据进行灰色关联度分析,找出影响减薄率的两个主要因素,即冲压速度和压边力;基于神经网络遗传算法函数寻优模型,借助拉丁超立方抽样对选出的两个主要因素进行随机抽样,将冲压速度和压边力作为输入,最大减薄率作为输出,获得输入与输出之间的非线性映射关系,并获得BP神经网络预测结果。最后,将预测结果进行个体适应度值计算,得到全局最优解和对应输入值。对比优化前后的数值模拟结果以及实验结果可知,采用此方法所得的工艺参数组合可有效提高板料成形的性能和质量。
Based on GS theory, Artificial Neural Networks and Genetic Algorithm, the stamping process of a left rear side panel was simulated by using the analysis software named Dynaform. Firstly, by using the orthogonal design to get the combinations in different parameter. Then, based on the GS theory, the data obtained from last step were analyzed to get the main factors influencing the ratio of thinning. Moreover, using LHS to obtain new data of the main factors. Based on neural network genetic algorithm function optimization model, blank holder force and stamping speed were taken as the inputs while the maximum ratio of thinning as the output to train BP neural network data. Through putting the BP neural network prediction results, which are trained by genetic algorithm optimization, as individual fitness value to find the global optimal solution and the input value. The numerical simulation results and quality of production show that the optimized parameters can effectively improve the formability of sheet metal.
出处
《兵器材料科学与工程》
CAS
CSCD
北大核心
2017年第4期84-89,共6页
Ordnance Material Science and Engineering
关键词
汽车覆盖件拉延成形
灰色关联分析
拉丁超立方抽样
神经网络遗传算法
参数优化
metal forming
gray relational analysis
latin hypercube sampling
artificial neural networks and genetic algorithm
parametrical optimization
作者简介
熊文韬，男，硕士研究生；主要从事数字化设计与制造方向的研究，已发表学术论文5篇。E—mail：417180627@qq．com。
