摘要
针对BP神经网络在数控机床热误差建模时存在因局部最优解缺陷而导致的预测精度低的现象,提出了采用蝗虫优化算法(Grasshopper Optimization Algorithm,GOA)解决BP神经网络局部最优解缺陷,从而提升BP神经网络在数控机床热误差模型预测精度的方法。为验证所述方法的有效性,在Vcenter-55型立式三轴数控机床上进行热误差实验,得到6批次实验数据,通过模糊聚类结合灰色关联度的方法选择温度敏感点,建立多元回归、GOA-BP神经网络和BP神经网络热误差模型,对其它批次数据进行预测。结果表明:GOA-BP模型比多元回归模型平均残余标准差提升了10.210μm;比BP模型平均残余标准差提升了2.916μm,提升幅度达到50.36%;GOA有助于BP神经网络找到全局最优解,能有效提高BP神经网络热误差模型的预测精度。研究结果对进一步提高数控机床热误差模型的预测精度提供了一种新的方法。
To address the issue of low prediction accuracy caused by the local optimal solution defect in BP neural networks when modeling thermal errors in CNC machine tools,a method utilizing the Grasshopper Optimization Algorithm(GOA)to resolve the local optimal solution defect of BP neural network is proposed,thereby improving the prediction accuracy of BP neural networks in thermal error modeling of CNC machine tools.To verify the effectiveness of the proposed method,thermal error experiments are conducted on a Vcenter-55 vertical three-axis CNC machine tool,yielding six batches of experimental data.Temperature-sensitive points are selected using a method combining fuzzy clustering and grey relational analysis.Multiple regression,GOA-BP neural network,and BP neural network thermal error models are established to predict data and used to predict other data batches.The results show that the GOA-BP model achieves a 10.210μm improvement in average residual standard deviation over the multiple regression model,and a 2.916μm improvement over the BP model,corresponding to a 50.36%enhancement in prediction accuracy.This indicates that GOA effectively aids BP neural networks in locating the global optimum solution,thereby significantly boosting the prediction accuracy of the thermal error model.The findings offer a novel method to further improve the prediction accuracy of thermal error models in CNC machine tools.
作者
廖磊
廖彬钧
谢桢
LIAO Lei;LIAO Binjun;XIE Zhen(School of Mechanical Engineering,Chongqing University of Technology,Chongqing 400054,China)
出处
《西安航空学院学报》
2025年第1期52-59,共8页
Journal of Xi’an Aeronautical Institute
基金
国家重点研发计划(2019YFB1703700)
重庆市技术创新与应用发展专项重点项目(cstc2019jscx-mbdxX0045)。
作者简介
廖磊(1999-),男,重庆武隆人,硕士研究生,主要从事机床热误差研究。
