| 本文已被：浏览 246次 下载 287次 |
码上扫一扫！ |
|
|
| 脉冲PAW快速成形焊缝尺寸的预测模型 |
|
徐富家1,2, 吕耀辉2, 刘玉欣2, 徐滨士1,2, 何鹏1
|
|
1.哈尔滨工业大学先进焊接与连接国家重点实验室, 哈尔滨 150001;2.装甲兵工程学院装备再制造技术国防科技重点实验室, 北京 100072
|
|
| 摘要: |
| 介绍了一种基于脉冲等离子焊接快速成形的方法,并采用Taguchi法对单道成形试验进行了合理设计,从而获得了多组焊接工艺参数下的熔宽和余高数据.通过遗传算法(GA)结合BP神经网络的方法建立了等离子焊接快速成形的预测模型,该模型预测了不同焊接成形工艺参数下单道成形的熔宽和余高.结果表明,通过误差分析和线性回归的方法验证了模型具有较高的预测精度和泛化能力,能够合理的预测焊道尺寸,并能推广到多层多道堆积成形尺寸的预测. |
| 关键词: 脉冲等离子焊接 快速成形 田口法 遗传算法 反向传播神经网络 |
| DOI： |
| 分类号: |
| 基金项目:国家973计划资助项目(2011CB013403) |
|
| Prediction model of bead geometry shaped by rapid prototyping based on pulsed PAW |
|
XU Fujia1,2, LÜ Yaohui2, LIU Yuxin2, XU Binshi1,2, HE Peng1
|
|
1.Harbin Institute of Technology, State Key Laboratory of Advanced Welding and Joining, Harbin 150001, China;2.Academy of Armored Forces Engineering, National Key Laboratory for Remanufacturing, Beijing 100072, China
|
| Abstract: |
| In this paper,rapid prototyping based on pulsed PAW technology is introduced.The Taguchi method is applied to design single bead forming experiments properly and then multi-group experimental data of weld width and height are obtained.The prediction model of bead geometry is developed by applying BP neural network based on genetic algorithm.It turns out that the model holds high prediction accuracy and generalization ability verified by error analysis and linear regression,which can predict single and multiple beads geometry accurately. |
| Key words: pulsed PAW rapid prototyping Taguchi method genetic algorithm BP neural network |
