China Mechanical Engineering ›› 2014, Vol. 25 ›› Issue (7): 862-866.

Previous Articles     Next Articles

Identification of Temperature Measuring Points for Machine Tools with Kohonen Self-organizing Competitive Network

Gao Feng;Liu Jiang;Li Yan;Yang Xingang   

  1. Xi'an University of Technology,Xi'an,710048
  • Online:2014-04-10 Published:2014-04-11
  • Supported by:
    National Natural Science Foundation of China(No. 51375382);Shaanxi Provincial Natural Science Foundation of China(No. S2009JC1400)

基于Kohonen自组织竞争网络的机床温度测点辨识研究

高峰;刘江;李艳;杨新刚   

  1. 西安理工大学,西安,710048
  • 基金资助:
    国家自然科学基金资助项目(51375382);陕西省自然科学基金资助重点项目(S2009JC1400) 

Abstract:

A novel optimization algorithm based on Kohonen neural network for identifying temperature measuring points was presented, where temperature variables and positioning errors at the feeding system in the machine tools served as input sample to train the network. The winners resulted from self-organizing competition were transferred into the corresponding classified modes, so the thermal critical points were determined according to the correlation coefficients among temperature variables and thermal errors in each class. Finally, a thermal error model was established with multi variable linear regression analysis method using optimized thermal critical points, whose results are more feasible and practicable in comparison with the thermal error model based on grouping optimization.

Key words: Kohonen neural network, critical thermal point, identification of temperature measuring point, correlation coefficient

摘要:

提出一种基于Kohonen神经网络的温度测点辨识优化算法,用机床进给系统上不同位置处的温度测点变化值及定位误差作为输入样本来训练神经网络。利用该网络的自组织竞争将胜出的结果输出到相应的分类模式中,根据各类分类模式中温度变量与热误差之间的相关系数,确定出机床热关键点。通过多元线性回归理论建立了热误差模型,与基于变量分组优化方法的热误差模型比较发现,该方法具有更好的可行性和有效性。

关键词: Kohonen神经网络, 热关键点, 测点辨识, 相关系数

CLC Number: