航空学报 > 2024, Vol. 45 Issue (10): 429313-429313   doi: 10.7527/S1000-6893.2023.29313

基于数据驱动的纤维增强复合材料自动铺放速度预测与规划

杨倩, 王彦哲, 杨迪(), 李泽众, 曲巍崴   

  1. 浙江大学 机械工程学院,杭州 310027
  • 收稿日期:2023-07-12 修回日期:2023-07-27 接受日期:2023-09-20 出版日期:2024-05-25 发布日期:2023-10-08
  • 通讯作者: 杨迪 E-mail:yangdi0518@hotmail.com

Prediction and planning of automatic laying speed for fiber reinforced composite materials based on data⁃driven model

Qian YANG, Yanzhe WANG, Di YANG(), Zezhong LI, Weiwei QU   

  1. College of Mechanical Engineering,Zhejiang University,Hangzhou 310027,China
  • Received:2023-07-12 Revised:2023-07-27 Accepted:2023-09-20 Online:2024-05-25 Published:2023-10-08
  • Contact: Di YANG E-mail:yangdi0518@hotmail.com

摘要:

复合材料自动铺放(AFP)过程中,铺放速度的突变极易引起丝束翻折、褶皱、滑移等缺陷,从而降低铺放质量和铺放效率。基于铺放速度的预测结果进行优化调整,是提高铺放速度稳定性、保障铺放质量的重要途径。为此,提出一种基于数据驱动的铺丝机速度预测及规划方法。首先,基于随机森林方法,建立了以铺丝机运动轴为子树的铺放速度预测模型,提出以关节标称速度、加速度、关节轨迹夹角为输入特征,以关节实际速度为输出特征的随机森林模型特征参数定义方法;进一步,基于铺放速度预测结果分析,提出了指令速度的分段匀速规划方法;最后,给出了参考指令速度的制造周期预估方法。采用六自由度卧式机床的进气道铺放实验对上述方法进行验证。结果表明,该方法对同训练角度铺层铺放速度的预测准确度达到91%,随着学习数据增加,各角度铺层路径的速度预测精度均有提升。采用基于铺放速度预测结果的指令速度分段规划方法,可显著降低速度突变,有效提升铺放质量。在计算成本方面,通过与神经网络方法相比,证明了随机森林方法具备高效的铺放速度预测水平。

关键词: 复合材料, 自动铺丝, 铺放速度, 速度预测, 数据驱动模型

Abstract:

During the Automated Fiber Placement (AFP) process of composite materials, sudden changes in laying speed can easily cause defects such as tows folding, wrinkles, and slippage, thereby reducing laying quality and efficiency. Optimizing and adjusting laying speed based on predictive results is an important approach to improve laying speed stability and ensure laying quality. Therefore, this paper proposes a data-driven method for predicting and planning the laying speed of fiber placement machines. Firstly, based on the random forest method, a laying speed prediction model with the motion axis of the laying machine as the subtree is established. A method for defining feature parameters of the random forest model is proposed, with joint nominal speed, acceleration, and joint trajectory angle as input features, and actual joint speed as output features. Furthermore, based on the analysis of speed prediction results, a segmented uniform command speed planning method is proposed. Finally, a method for estimating the manufacturing cycle based on command speed is given. The above methods are verified through laying experiments on a six-degree-of-freedom horizontal machine tool. The results show that the accuracy of predicting laying speed for layers laid at the same training angle reaches 91%, and with the increase of learning data, the prediction accuracy of laying paths at various angles is improved. The segmented planning method for command speed based on speed prediction results can significantly reduce speed fluctuations and effectively improve laying quality. In terms of computational cost, compared with neural network methods, the random forest method demonstrates efficient laying speed prediction capabilities.

Key words: composite materials, automated fiber placement, layup speed, speed prediction, data driven model

中图分类号: