锻压技术

基于BP神经网络的管材数控弯曲多参数优化方法研究

英文标题：Research on optimization method of multi-parameter in NC tube bending based on BP neural network

作者：吴超严勇胡志力

分类号：TG386.43

出版年,卷(期):页码：2015,40(6):131-137

摘要：

针对管材数控弯曲成形过程中多工艺参数耦合的特点，基于BP神经网络，结合多目标优化算法研究了弯曲成形工艺参数的优化方法。采用ABAQUS对管材数控弯曲过程进行有限元仿真，并实验验证了结果的准确性。基于MATLAB平台，以芯棒直径、芯棒伸出量、防皱块与管材间摩擦系数等主要工艺参数为优化对象，以外壁减薄率、内壁增厚率（起皱）为优化目标，通过验证的数值模型获得样本数据，利用BP神经网络建立了优化对象和优化目标之间的映射关系，并采用多目标优化算法进行寻优求解，最后通过数值仿真实验验证了优化方法的准确性。结果表明：薄壁管数控弯曲有限元数值模拟结果与实验数据吻合较好，可以为神经网络提供准确可靠的训练样本； BP神经网络结合多目标优化算法可以有效地对弯曲工艺参数进行优化；优化的工艺参数有效地改善了弯曲管材内侧起皱和外侧减薄。

For the characteristics of multi parameters coupling during NC tube bending, the optimization method of process parameters in tube bending was studied based on the BP neural network combined with the multi-objective algorithm. The process of NC tube bending was simulated numerically by ABAQUS, and the simulation result was verified by the experiment. Then based on MATLAB platform, the sample data were obtained by the numerical model, the mapping relationship between the optimization object(the mandrel diameter, the extension of mandrel and the friction coefficient between the wiper die and tube) and the optimization goal(the outside wall thinning and inside wall thickening(wrinkle)) was established by the BP artificial neural network, and the optimal process parameters were obtained by optimization algorithm for multiple targets. The validity of the optimization method was verified by simulation. The results show that the numerical simulation results show good agreement with the experimental data in tube NC bending, thus the reliable sample data for the neural network training can be provided by the numerical model. The process parameters can be effectively optimized by the BP neural network combined with multi-objective algorithm. The outside wall thinning and inside wall wrinkling of the bending tube are effectively improved by the optimized process parameters.

基金项目：

中央高校基本科研业务费专项资金资助项目(2014-IV-042）

吴超（1989-），男，硕士研究生

参考文献：

［1］杨合,孙志超,林艳，等.面向21世纪的先进塑性加工技术与管成形研究发展[A].中国科学协会第二届学术年会文集[C].北京:科学技术出版社,2000.Yang H,Sun Z C,Lin Y,et al. Advanced processing technology and research progress on tube forming[A]. China Association for Science Proceedings of the Second Annual[C]. Beijing: Science and Technology Press,2000.
［2］闫晶,杨合,詹梅,等. 基于多成形指标的大直径铝合金薄壁管数控弯曲极限[J].中国科学：技术科学，2010,(6):601-618.Yan J,Yang H,Zhan M,et al. Forming limits under multi-index constraints in NC bending of aluminum alloy thin-walled tubes with large diameters[J]. Science China: Technical Science,2010,(6):601-618.
［3］张尧武,曾卫东,戴毅,等. 基于虚拟正交试验的热推弯管工艺参数优化设计[J].塑性工程学报，2009,16(6):91-95.Zhang Y W，Zeng W D,Dai Y,et al. Optimal design of technological parameters for hot-pushing pipe-bending based on virtual orthogonal experiment[J]. Journal of Plasticity Engineering,2009,16(6):91-95.
［4］贾美慧,王成林,孙卫华. 基于神经网络和粒子群算法的管材弯曲工艺参数优化[J].制造业自动化，2014,36(12):109-113.Jia M H,Wang C L,Sun W H. Optimization of process parameters in tube bending based on neural network and particle swarm algorithm[J].Manufacturing Automation,2014,36(12):109-113.
［5］林高用,陈兴科,蒋杰,等. BP人工神经网络与遗传算法在型材挤压模具参数优化中的应用[J].湘潭大学自然科学学报，2006,28(2): 89-94.Lin G Y,Chen X K,Jiang J,et al. Application of BP artificial neural network and genetic algorithm to the parameters optimization of profile extrusion die[J]. Natural Science Journal of Xiangtan University,2006,28(2): 89-94.
［6］胡贵强. 多目标优化的遗传算法及其实现[J].重庆文理学院学报:自然科学版，2008,27(5): 12-15.Hu G Q. The research and implementation of genetic algorithm for multi-objective optimization[J].Journal of Chongqing University of Arts and Sciences: Natural Science Edition，2008,27(5): 12-15.
［7］李恒，杨合，詹梅,等. 大口径薄壁小弯曲半径数控弯管有限元建模和实验[J].锻压技术，2006,31(5):136-139.Li H,Yang H,Zhan M,et al. FEM modeling and experimental of NC bending process of thin-wall tube with large diameter and small bending radius[J]. Forging & Stamping Technology,2006,31(5):136-139.
［8］张德丰.MATLAB神经网络仿真与应用[M].北京：电子工业出版社，2009.Zhang D F. MATLAB Neural Network Emulation and Application[M].Beijing: Electronic Industry Press,2009.
［9］朱凯,王正林. 精通MATLAB神经网络[M].北京：电子工业出版社，2010.Zhu K,Wang Z L. Proficient in MATLAB Neural Network[M]. Beijing: Electronic Industry Press,2010.
［10］沈花玉，王兆霞，高成耀.BP神经网络隐含层单元数的确定[J].天津理工大学学报，2008,24(5): 13-15.Shen H Y,Wang Z X,Gao C Y. Determining the number of BP neural network hidden layer units[J]. Journal of Tianjin University of Technology,2008,24(5): 13-15.
［11］Pawar P J,Rao R V,Davim J P. Multi-objective optimization of grinding process parameters using particle swarm optimization algorithm[J]. Materials and Manufacturing Processes,2010,25(6): 424-431.
［12］林焰，郝聚民，纪卓尚.基于模糊优选的多目标优化遗传算法[J].系统工程理论与实践，1999,(12):31-37.Lin Y,Hao J M,Ji Z S. Genetic algorithms based on fuzzy evaluation for multi-criterion function optimization[J]. Systems Engineering Theory and Practice,1999,(12):31-37.
［13］张彦. 基于多目标优化随机权系数加权和的机组负荷分配[J].电网技术，2008,32(2): 64-67.Zhang Y. Multi-objective optimization of economic dispatch problem based on stochastic weighted sum method and multi-attribute decision making[J].Power System Technology,2008,32(2): 64-67.

服务与反馈：

【文章下载】【加入收藏】