摘要
结合数值模拟与人工神经网络技术研究了汽车内覆盖件承载地板在冲压成形中压边力的预测。将板料模型零件导入到Dynaform中进行网格划分并对其拉深过程进行模拟仿真,结合正交试验获取不同参数条件下最佳压边力的数据样本,然后运用Matlab软件中的GRNN神经网络工具箱对数据进行训练学习,采用训练好的神经网络对板料成形过程中的压边力进行预测,获得了板料拉深过程中的压边力变化曲线。通过预测结果和模拟结果对比,预测误差在10%以内。将预测的曲线对零件模拟仿真,结果显示零件最大减薄率在25%以内,并对板料进行实际冲压验证。结果显示成形效果良好,无起皱、破裂缺陷,符合实际生产的要求,说明GRNN神经网络可以用于零件冲压过程中压边力的预测。
The prediction on blank holder force of bearing floor for automobile covering parts in the stamping process was studied combining numerical simulation with artificial neural network technique.First,the model part was imported into Dynaform for mesh generation and the drawing process was studied,and data samples of the optimal blank holder force under different parameter conditions were obtained by the orthogonal test.Then,the data were trained and learned by GRNN neural network toolbox in the Matlab software,and the blank holder force was predicted by the trained neural network.Therefore,the changing curve of blank holder force in the sheet metal drawing process was obtained,and the prediction deviation is within 10% comparing with simulation results.Furthermore,the drawing process of parts was simulated based on the predicted curve,and the maximum reduction ratio is displayed within 25%.Finally,the actual stamping test for sheet metal was conducted,and the forming effect met the actual production requirements without wrinkles and rupture defects.It indicates that GRNN neural network can be applied to predict blank holder force in the stamping process.
出处
《锻压技术》
CAS
CSCD
北大核心
2017年第7期43-48,共6页
Forging & Stamping Technology
基金
华中科技大学材料成形与模具技术国家重点实验室开放基金课题(P2015-01)
关键词
压边力预测
冲压
拉深
GRNN神经网络
正交试验
汽车覆盖件
blank holder force prediction
stamping
drawing
GRNN neural network
orthogonal experiment
automobile covering parts
