|
| 基于神经网络遗传算法的深腔型零件拉深工艺参数优化 |
| Parameter Optimization of Deep Drawing Process for Deep Cavity Parts Based on Neural Network Genetic Algorithm |
| Received:October 31, 2020 |
| DOI：10.3969/j.issn.1674-6457.2021.05.023 |
| 中文关键词: 冲压成形 参数优化 灰度关联分析 神经网络 遗传算法 |
| 英文关键词: stamping forming parameter optimization grey correlation analysis neural network genetic algorithm |
| 基金项目:国家自然科学基金(51675029);四川省院校合作项目(2019YFSY0034) |
| Author Name | Affiliation | | QIU Chao-bin | AVIC Xi'an Aircraft Industry Group Company Ltd., Xi'an 710089, China | | ZHANG Meng | School of Energy and Power Engineering, Beihang University, Beijing 100191, China | | LANG Li-hui | School of Mechanical Engineering and Automation, Beihang University, Beijing 100191, China | | GUO Qing-lei | School of Mechanical Engineering and Automation, Beihang University, Beijing 100191, China | | JIANG Pei-cheng | School of Mechanical Engineering and Automation, Beihang University, Beijing 100191, China | | LI Kui | AVIC Xi'an Aircraft Industry Group Company Ltd., Xi'an 710089, China | | CHEN Lin | Tianjin Tianduan Aviation Technology Co., Ltd., Tianjin 300232, China |
|
| Hits: 2258 |
| Download times: 955 |
| 中文摘要: |
| 目的 解决冲压成形中工艺参数优化难的问题。方法 以一种深腔型零件的冲压成形为例。首先，借助灰度关联分析法对有限元中的工艺参数进行分析，获取该零件冲压成形中影响成形质量的2个主要因素——冲压速度和压边力。其次，借助拉丁超立方抽样法对上述2个因素进行随机取样，并借助DYNAFORM软件对其进行逐一模拟。再次，将冲压速度和压边力作为输入，最大减薄作为输出，训练在MATLAB中建立的BP神经网络，并借助遗传算法对其进行寻优。结果 最优成形压边力为1.372 MN，最优加载速度为1.5366 m/s。结论 与神经网络遗传算法预测相比，有限元结果的相对误差小于2%，零件试制结果的相对误差小于6%，该方法有较高的预测精度。 |
| 英文摘要: |
| In order to solve the problem of difficult optimization of process parameters in stamping forming. this paper takes the stamping forming of a deep cavity part as an example. Firstly, the process parameters in the finite element were analyzed by gray correlation analysis, and the two main factors affecting the forming quality, namely stamping speed and blank holder force, are obtained. Secondly, above two main factors were sampled randomly by Latin hypercube sampling method and simulated one by one with the help of DYNAFORM software. After that, the stamping speed and the blank holder force were used as the input data and the maximum thinning rate as output data to train the BP neural network established in MATLAB and optimize it with the help of genetic algorithm. The optimal blank holder force was 1.372 MN and the optimal loading speed is 1.5366 m/s. Compared with the prediction of neural network genetic algorithm, the relative error of finite element results is less than 2%, and the relative error of parts trial production results is less than 6%, which means that this method has higher prediction accuracy. |
|
View Full Text
View/Add Comment Download reader |
| Close |
|
|
|
渝公网安备 50010702501719号