摘要
为实现数控机床主轴热误差的精准预测,提出一种基于加权平均算法(WAA)优化双向长短期记忆神经网络(BiLSTM)的数控机床主轴热误差预测建模方法。进行机床主轴热误差实验,采集机床主轴的温度数据和热误差数据,结合主成分分析(PCA)与随机森林(RF)筛选温度关键点;以筛选的温升数据为输入、轴向热误差数据为输出,使用加权平均算法对双向长短期记忆神经网络的超参数空间进行优化,建立WAA-BiLSTM热误差预测模型,并与其他模型进行主轴热误差预测性能对比分析。结果表明:与GWOA-BiLSTM、PSO-CNN以及BiLSTM预测模型相比,WAA-BiLSTM预测模型的RMSE分别降低了30.5%、40.6%、43.8%,MAE分别降低了31.13%、37.06%、43.3%;WAA-BiLSTM模型的R 2为0.9808,明显优于其他3种模型,表明WAA-BiLSTM模型能够准确预测数控机床主轴热误差,为机床主轴热误差预测提供了一种新思路。
In order to predict the thermal error of CNC machine tool spindle more accurately,a prediction modelling method for CNC machine tool spindle thermal error prediction based on weighted average algorithm(WAA)optimized bi-directional long short-term memory neural network(BiLSTM)was proposed.The temperature data and thermal error data of the machine tool spindle were collected by conducting thermal error experiments of the machine tool spindle.The screening of the temperature key points was carried out by using principal component analysis(PCA)combined with the random forest(RF)method.The temperature rise data obtained after screening was used as the input,and the axial thermal error data was used as the output,and hyperparameter space of the BiLSTM was optimized using the WAA,and a WAA-BiLSTM thermal error prediction model was established.And a comparative analysis of the performance of spindle thermal error prediction with other models was conducted.The results show that compared with the GWOA-BiLSTM,PSO-CNN and BiLSTM prediction models,the RMSE of the WAA-BiLSTM prediction model has decreased by 30.5%,40.6%and 43.8%respectively,and the MAE has decreased by 31.13%,37.06%and 43.3%respectively.The R 2 of the WAA-BiLSTM model is 0.9808,which is significantly better than the other three models.This indicates that the WAA-BiLSTM model can accurately predict the thermal error of the spindle of CNC machine tools and provides a new idea for the prediction of the thermal error of the spindle of machine tools.
作者
刘思怡
郭忠峰
刘琪
朱康博
陈骥驰
LIU Siyi;GUO Zhongfeng;LIU Qi;ZHU Kangbo;CHEN Jichi(School of Mechanical Engineering,Shenyang University of Technology,Shenyang Liaoning 110870,China)
出处
《机床与液压》
北大核心
2025年第23期134-139,共6页
Machine Tool & Hydraulics
基金
国家自然科学基金项目(62471319)。
关键词
数控机床
主轴热误差预测
加权平均算法
双向长短期记忆神经网络
CNC machine tool
spindle thermal error prediction
weighted average algorithm
bi-directional long short-term memory neural networks
