Integrated thermal error modeling of machine tool spindle using a chicken swarm optimization algorithm-based radial basic function neural network

• Published in: International journal of advanced manufacturing technology Vol. 105; no. 5-6; pp. 2039 - 2055
• Main Authors: Fu, Guoqiang, Gong, Hongwei, Gao, Hongli, Gu, Tengda, Cao, Zhongqing
• Format: Journal Article
• Language: English
• Published: London Springer London 01.12.2019; Springer Nature B.V
• Subjects: Accuracy; Algorithms; CAE) and Design; Cluster analysis; Clustering; Computer-Aided Engineering (CAD; Correlation analysis; Engineering; Error analysis; Error compensation; Error reduction; Fitness; Industrial and Production Engineering; Machine tool industry; Machine tools; Machining; Machining centres; Mathematical models; Mechanical Engineering; Media Management; Model accuracy; Modelling; Neural networks; Optimization; Optimization algorithms; Original Article; Parameters; Radial basis function; Training; Vector quantization; Thermal error modeling; Machine tools; Chicken swarm optimization (CSO) algorithm; Radial basis function (RBF) neural network; KC-RBF approach
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-019-04388-5

Thermal errors are one of the main error sources affecting the machining accuracy of the machine tool. Thermal error modeling is a prerequisite for thermal error compensation to reduce the thermal error and improve machining accuracy. In this paper, a chicken swarm optimization algorithm-based radial basic function (CSO-RBF) neural network is applied to integrated thermal error modeling. At first, correlation analysis-based K -Means clustering and radial basis function neural network (KC-RBF) approach is proposed to screen optimal temperature-sensitive point combination. The correlation analysis-based K -Means clustering is used to obtain temperature-sensitive point combinations corresponding to different K values. The mean value of residual and root mean square error are established to evaluate the results of RBF model to filter the optimal temperature-sensitive point combination. Secondly, one CSO-RBF neural network is proposed to handle the nonlinear relationship between temperature variables and thermal errors. RBF model-based fitness function is proposed for CSO to obtain the optimal initial structure parameters of RBF. The optimal thermal error model is established by training RBF with the optimal initial structure parameters and the measured data. At last, different experiments are carried out on VMC850 machining center: training and testing of thermal error models at a fixed speed for Y-direction thermal drift error; verification of thermal error models for different speeds of different error parameters. It is worth mentioning that the model trained with one thermal error parameter measured at a certain speed is also applied for different thermal error parameters at different speeds. Results show that the proposed CSO-RBF model has high accuracy and strong robustness.