关键词: 动平衡, 主轴振动预报, DBSCAN聚类算法, RBF网络

Abstract: In order to solve the existing problem of low accuracy and the difficulty of modeling in dynamic balance process modeling method of mechanical spindle rotor, a method of establishing mechanical spindle vibration prediction model using RBF network is proposed to realize the prediction of the spindle vibration amplitude under the different moving position of balance blocks. In the prediction model based on RBF network, DBSCAN clustering algorithm is introduced to determine the radial basis function centers in hidden layer of the network, so as to identify the hidden layer nodes objectively and improve the precision of the prediction model. Finally with the help of the dynamic balance test platform, the accuracy of the prediction model is verified, and the modeling method is compared with RBF network based on maximal matrix element method, RBF network based on K-means algorithm, BP network based on genetic algorithms and artificial neural network. The results show that the modeling method of mechanical spindle vibration prediction model proposed in this paper achieves effective prediction of the vibration amplitude with higher precision.

Key words:

Dynamic balance, spindle vibration prediction, DBSCAN clustering algorithm, RBF network