Calculation method of convective heat transfer coefficients for thermal simulation of a spindle system based on RBF neural network

Results from the temperature field and thermal deformation simulation of a spindle system are greatly affected by the accuracy of convective heat transfer coefficients (CHTCs). This paper presents a new method based on radial basis function (RBF) neural network to calculate CHTCs. First, the temperature field and thermal deformations of a spindle system were obtained by experimental and finite-element (FE) methods. However, the simulation results are significantly different from the experimental results because boundary conditions used for the FE model were derived empirically. Second, the relationship between the simulated temperature values and CHTCs were established by a RBF neural network. Using the experimental temperature values as an input vector of the RBF neural network, CHTCs of the spindle system can be predicted through an iterative calculation taking 14 cycles. Finally, the effectiveness of the proposed method was proved using steady-state and transient-state analyses of the spindle system. Results from the steady-state simulation show that temperature errors were less than 4 % at the seven thermal-critical points and deformation errors in the three directions were less than 6 %. Results from the transient-state simulation of the spindle system show that the variations for each of the thermal characteristics are in good agreement with the experimental results. The method provides guidance for modifying boundary conditions of a FE model.