节点文献

基于人工神经网络的U型材单轴柔性滚弯成形的预测研究

Prediction Research on Single-axle Rotary Shaping of U-Profile Base on Artifial Neural Network

【作者】 刘海燕

【导师】 金霞;

【作者基本信息】 南京航空航天大学 , 航空宇航制造工程, 2009, 硕士

【摘要】 单轴柔性滚弯成形是一种先进的、柔性的钣金制造工艺,适于加工比强度高的材料,成形的工件具有表面质量好、成本低等特点,适用于航空、航天及汽车工业等小批量生产中。但与其他板料成形方法一样,回弹是单轴柔性滚弯成形中不可避免,对零件的尺寸精度和生产效率会造成很大的影响,因此,对回弹进行有效预测与控制研究具有重要意义。论文的主要研究工作如下:1.详细系统地介绍了单轴柔性滚弯技术的原理、特点及其优越性;对滚弯成形回弹过程数值模拟中的关键技术进行了研究。2.利用自行研制的单轴柔性滚弯装置进行U型材滚弯成形实验,同时用有限元分析软件Msc.Marc对成形及回弹过程进行了有限元仿真,将实际工艺实验结果与有限元结果进行对比,结果证明,采用有限元数值模拟结果在一定程度上可以代替实验结果,为回弹预测研究建立基础。3.列出了影响工件成形回弹半径的主要因素及各因素对曲率半径的影响趋势;采用正交设计方法制定正交试验方案,用级差的方法和图示法对各主要因素对滚弯成形回弹半径的影响进行了定性分析。4.基于正交数值模拟试验提供的训练样本,建立了基于BP神经网络的单轴柔性滚弯成形预测模型,输入测试样本,将预测结果与实验值进行对比,验证了该预测模型的准确性。以MTALAB图形用户界面为开发环境设计了预测系统的人机交互界面,操作简便,有效提高了工作效率。本文工作的特色在于对U型材滚弯成形作了多工艺参数综合影响分析,将数值模拟技术、正交试验方法与人工神经网络三者相结合,应用于对U型材单轴柔性滚弯成形回弹半径的预测中,实现了在保证分析精度的前提下,提高了工艺参数设计效率,对发展精确成形技术具有现实的指导意义。

【Abstract】 One-Axle rotary shaping with elastic medium (RSEM) is an advanced and highly flexible manufacturing technology of sheet metal forming. The bending products have the advantages of high strength, good surface quality, low cost and so on. It is applied in aeronautics & astronautics, automobile industries for mini-batch productions. But the springback is inevitable in One-Axle RSEM like other sheet metal forming methods, which would influence the precision and the production efficiency of the products. So it is very important to predict and control the springback. The main research contents of the paper are as follows:1.The working principle, characteristics and superiority of One-Axle RSEM technology have been introduced in detail. The key technologies of roll bending and its springback processing numerical simulation have also been researched.2. The U-shaped roll bending experiments have been finished by using a self-developed One-Axle RSEM device. At the same time, we have simulated the roll bending and its springback processing with finite element analysis (FEA) software Msc.Marc. Compared with actual experiment results and simulation results, it proved that the finite element numerical simulation results could replace the experiment results in a certain extent. It established a foundation for the springback prediction research.3. The main influence factors and their impact trend to the curvature radius are listed. The orthogonal design method has been used to set up the experiment project, and then we have made qualitative analysis of the main influence factors to the bending and springback curvature radius with the differential and icon methods.4. Based on training samples provided by numerical simulation experiment, the One-Axle RSEM prediction model is built, which is based on BP neural network. The precision of this prediction model has been verified comparing with the prediction results and experimental results when entering in test samples. a man-machine interface of the prediction system which is easy to operate, is designed based on graphical user interface development environment of MATLAB. The working efficiency would be improved effectively.In this paper, the characteristic is that the U-shaped roll bending is made more comprehensive impact analysis, the combination of orthogonal test method, numerical simulation technology and artificial neural network, has effectively been using in predicting the U-shaped One-Axle RSEM curvature radius. On the case of ensure the analysis precision, the design efficiency of process arameters have been improved, It would guide the development of precision forming technology.

节点文献中: