摘要
为精确建立五轴机床旋转轴热误差模型,提出了一种基于奇异谱分析(SSA)的改进粒子群优化(IPSO)算法-长短期记忆(LSTM)神经网络旋转轴热误差建模预测方法。通过奇异谱分析,将输入温度数据分解、重构为趋势序列与噪声序列。将改进惯性权重与学习因子的改进粒子群优化算法用于优化长短期记忆神经网络的神经元数量与学习率。对趋势序列与噪声序列由改进粒子群优化算法-长短期记忆神经网络分别训练并求和,得到输入为旋转轴温度、转向、角度,输出为旋转轴角度定位热误差的热误差预测模型。试验结果表明,所提出的预测方法精度高达94.5%以上,均方根误差在0.3″内,均优于传统长短期记忆神经网络和传统粒子群优化算法-长短期记忆神经网络。
In order to accurately establish the thermal error model of the rotation axis of five-axis machine tool,an improved particle swarm optimization algorithm-long short-term memory neural network rotation axis thermal error modeling and prediction method based on singular spectrum analysis was proposed.The input temperature data was decomposed and reconstructed into trend sequence and noise sequence by singular spectrum analysis.The improved particle swarm optimization algorithm with improved inertia weight and learning factor was used to optimize the number of neurons and learning rate of long short-term memory neural network.The trend sequence and the noise sequence were trained and summed by the improved particle swarm optimization algorithm-long short-term memory neural network,and the thermal error prediction model with the input as rotation axis temperature,steering and angle and the output as the positioning thermal error of rotation axis angle was obtained.The experimental results show that the accuracy of the proposed prediction method is as high as 94.5%,and the root mean square error is within 0.3",which is better than the traditional long short-term memory neural network and the traditional particle swarm optimization algorithmlong short-term memory neural network.
作者
程涛
项四通
Cheng Tao;Xiang Sitong
出处
《机械制造》
2023年第12期56-61,共6页
Machinery
基金
国家自然科学基金资助项目(编号:52175470)
宁波市自然科学基金重点项目(编号:2022J074)
宁波市重点研发计划项目(编号:2021Z077)。
