摘要
为精确预测管材弯曲回弹并设计合理的补偿方案,选择经过优化处理的BP机器学习算法建立预测模型,之后对其开展了控制性能评价。大幅提升了泛化性能并获得更高的预测精度,促进算法更快完成收敛过程。并对模型开展了验证分析。研究结果表明:当以PSO算法优化BP建立预测模型进行预测时跟目标结果间形成了15.7%的平均误差,相对于BP预测模型,大幅提升了预测精度,但会导致计算效率明显下降,所需计算时间接近1.5h。以改进粒子群算法对BP进行优化后,可以有效提升神经网络泛化性能,跟目标值相比平均误差只有6.2%。先对基本PSO算法实施优化处理,再利用优化后的PSO算法调整BP,由此建立得到机器学习预测模型。此模型可以达到高预测精度以及高效率的要求,可以有效满足管材数控弯曲回弹以及补偿的计算需求。
In order to accurately predict the springback of pipe bending and design a reasonable compensation scheme,the opti-mized BP machine learning algorithm was selected to establish a prediction model,and then the control performance of the model was evaluated.The generalization performance is greatly improved and the prediction accuracy is higher,which promotes the al-gorithm to complete the convergence process faster.The model is validated and analyzed.The results show that when the PSO algo-rithm is used to optimize BP to build a prediction model for prediction,the average error between the PSO model and the target re-sult is 15.7%.Compared with the BP prediction model,the prediction accuracy is greatly improved,but the calculation efficiency is decreased significantly,and the calculation time is close to 1.5h.After optimizing BP by improved particle swarm optimization algorithm,the generalization performance of neural network can be effectively improved,and the average error is only 6.2%com-pared with the target value.Firstly,the basic PSO algorithm is optimized,then the OPTIMIZED PSO algorithm is used to adjust BP,and the machine learning prediction model is established.This model can meet the requirements of high prediction accuracy and high efficiency,and can effectively meet the calculation requirements of springback and compensation of pipe NC bending.
作者
刘贵彬
解小琴
王平
张利虎
LIU Gui-bin;XIE Xiao-qin;WANG Ping;ZHANG Li-hu(Department of Mechanical and Electrical Engineering,Sichuan Institute of Electronic and Mechanical Technology,Sich-uan Mianyang 621023,China;Bazhong Vocational and Technical College,College of Science and Technology,Sichuan Ba-zhong 636000,China;School of Mechanical and Vehicle Engineering,Chongqing University,Chongqing 400030,China)
出处
《机械设计与制造》
北大核心
2023年第10期130-133,共4页
Machinery Design & Manufacture
基金
2019年度绵阳职教中心科研课题(MZJYB33)。
关键词
回弹
管材弯曲
机器学习
成形精度
Rebound
Pipe Bending
Machine Learning
Forming Precision
