锻压技术

机器学习耦合有限元分析预测板料气弯回弹行为

英文标题：Springback behavior on sheet metal in gas bending predicted by machine learning coupled with finite element analysis

作者：徐承亮张祥林王大军

单位：1. 广州科技贸易职业学院产业学院 2. 华中科技大学材料成形与模具技术国家重点实验室 3. 重庆邮电大学自动化学院

关键词：空气辅助弯曲弯曲回弹机器学习神经网络有限元分析

分类号：TG302

出版年,卷(期):页码：2022,47(6):107-112

摘要：

采用机器学习神经网络（NN）耦合有限元分析（FEA）的方法来构建弯曲成形过程的非线性回弹模型,并且考虑了不同材料、工艺参数和模具几何形状，可以有效和准确地预测工件的弯曲回弹行为。当模具开口量V=11 mm、板料厚度t= 3 mm时，对于结构钢HC220材料，机器学习NN模型的预测值（YNN）与回弹后分析解（yJBP）的均方根误差RMSE分别为0.28和1.70；对于双相钢DP590材料，YNN与yJBP的RMSE分别为0.45和0.22。采用NN模型、回弹后分析解（yJBP）和FEA方法的CPU计算时间分别为3.1、6.3和278 s，NN模型的CPU计算时间最少，实验结果表明，NN模型可以在良好的预测精度和高效的求解速度之间达到一个最佳平衡。

A nonlinear springback model of bending process was constructed by machine learning neural network (NN) coupled finite element analysis (FEA), and considering different materials, process parameters and die geometry shapes, the bending springback behavior of workpiece could be predicted effectively and accurately. When the die opening amount V=11 mm and the sheet thickness t=3 mm, for structural steel HC220 material, the root mean square errors RMSE of prediction value (YNN) by machine learning NN model and analytical solution after springback (yJBP) were 0.28 and 1.70 respectively. For dual-phase steel DP590 material, the RMSE values of YNN and yJBP were 0.45 and 0.22 respectively. The CPU calculation time of NN model, analytical solution after springback (yJBP) and FEA methods were 3.1, 6.3 and 278 s respectively, and the NN model was of less CPU calculation time. The experimental results show that the NN model can achieve an optimal balance between good prediction accuracy and efficient solution speed.

基金项目：

广东省普通高校特色创新项目（自然科学类）（2018GKTSCX053）；2021年度广州市基础研究计划基础与应用基础研究项目(2021-02-08-13-0018)；材料成形与模具技术国家重点实验室基金资助项目(P2021-016)

徐承亮(1970-)，男，硕士，高级工程师，副教授 E-mail：281552074@qq.com

参考文献：

[1]宋云鹤, 秦雪娇, 鲍益东. 基于DKT12壳单元的板料冲压成形中重力加载变形模拟仿真[J]. 塑性工程学报, 2020, 27(11):65-69.

Song Y H, Qin X J, Bao Y D. Simulation of gravity loading deformation in sheet metal stamping based on dkt12 shell element[J]. Journal of Plasticity Engineering, 2020, 27 (11): 65-69.

[2]Pacheco J B, Santos A D. A study on the nose radius influence in press brake bending operations by finite element analysis[J]. Key Engineering Materials, 2013, 554-557(2):1432-1442.

[3]丁立波, 李健东, 李明杰,等. 高强钢辊弯成形过程中成形力的影响因素研究[J]. 精密成形工程, 2021, 13(4):6-12.

Ding L B, Li J D, Li M J,et al. Study on influencing factors of forming force in roll bending process of high strength steel[J]. Precision Forming Engineering, 2021, 13 (4): 6-12.

[4]洪凯骏, 杜晟强, 昝祥. 不同背压方式对静液挤压过程影响的模拟研究[J]. 精密成形工程, 2019, 11(2):17-21.

Hong K J, Du S Q, Min X. Simulation study on the influence of different back pressure on hydrostatic extrusion process[J]. Precision Forming Engineering, 2019, 11 (2): 17-21.

[5]Nath U, Yadav V, Purohit R. Finite element analysis of AM30 magnesium alloy sheet in the laser bending process[J]. Advances in Materials and Processing Technologies, 2021，(3): 1-13.

[6]方刚, 陈祝, 雷丽萍. 非关联本构模型在铝合金板料成形有限元模拟中的应用[J]. 塑性工程学报, 2021, 28(6):8-18.

Fang G, Chen Z, Lei L P. Application of uncorrelated constitutive model in finite element simulation of aluminum alloy sheet forming[J]. Journal of Plasticity Engineering, 2021, 28 (6):8-18.

[7]熊文韬, 谢三山, 黄兆飞,等. 基于神经网络遗传算法函数寻优与回弹补偿技术的某型汽车前幅拉延成形优化[J]. 塑性工程学报, 2020, 27(6):8-12.

Xiong W T, Xie S S,Huang Z F，et al. Optimization of drawing forming of front panel of an automobile based on neural network genetic algorithm function optimization and springback compensation technology[J]. Journal of Plasticity Engineering, 2020, 27 (6): 8-12.

[8]Liu S, Xia Y, Shi Z. Deep learning in sheet metal bending with a novel theoryguided deep neural network[J]. Journal of Automatica Sinica, 2021, 8(3): 565-581.

[9]张新艳, 郭鹏, 余建波. 应用深度强化学习的压边力优化控制[J]. 哈尔滨工业大学学报, 2020, 52(7):9-13.

Zhang X Y, Guo P, Yu J B. Optimal control of blank holder force using deep reinforcement learning[J]. Journal of Harbin Institute of Technology, 2020, 52 (7): 9-13.

[10]Ingarao G, Di Lorenzo R. Design of complex sheet metal forming processes: a new computer aided progressive approach[J]. International Joural of Material Forming, 2010,3(1): 21-24.

[11]Le T T, Asteris P G, Lemonis M E. Prediction of axial load capacity of rectangular concretefilled steel tube columns using machine learning techniques[J]. Engineering with Computers, 2021: 1-34.

[12]陈光耀，李恒，贺子芮，等. 基于机器学习的管材弯曲回弹有效预测与补偿[J]. 中国机械工程, 2020, 31(22):8-13.

Chen G Y, Li H, He Z R，et al. Prediction and compensation of pipe bending springback based on machine learning[J]. China Mechanical Engineering, 2020, 31 (22): 8-13.

[13]Ly H B, Le T T, Vu H L T, et al. Computational hybrid machine learning based prediction of shear capacity for steel fiber reinforced concrete beams[J]. Sustainability, 2020, 12(7): 2709-2732.

[14]Miranda S S, Barbosa M R, Santos A D, et al. Forming and springback prediction in press brake air bending combining finite element analysis and neural networks[J]. The Journal of Strain Analysis for Engineering Design, 2018, 53(8): 584-601.

[15]Fu Z, Mo J. Springback prediction of highstrength sheet metal under air bending forming and tool design based on GA-BPNN[J]. The International Journal of Advanced Manufacturing Technology, 2011, 53(5): 473-483.

[16]Narayanasamy R, Padmanabhan P. Comparison of regression and artificial neural network model for the prediction of springback during air bending process of interstitial free steel sheet[J]. Journal of Intelligent Manufacturing, 2012, 23(3): 357-364.

[17]Sharad G，Nandedkar V M. Springback in sheet metal U bendingFEA and neural network approach[J]. Pracedia Materials Science, 2014,6:835-839.

[18]ElZeghayar M, Topper T, Bonnen J, et al. Effective strainfatigue life of dual phase 590 steel[A]. Proceedings of the 12th International Conference on Fracture[C]. Ottawa， 2009.

[19]Pham Q T, Kim Y S, Nguyen D T. Application/comparison study of a graphical method of forming limit curve estimation for DP590 steel sheets[J]. Journal of the Korean Society for Precision Engineering, 2019, 36(9): 883-890.

服务与反馈：

【文章下载】【加入收藏】