Abstract: Bearings generate a large amount of frictional heat when a high speed motorized-spindle runs, and the temperature of the bearings is a main factor affecting stiffness and accuracy of the spindle.When the high speed motorized-spindle without load runs in different speeds, 151 sets of temperature are tested.Based on the principle of BP neural network, as far as each tested temperature is concerned, the former 100 temperature data are used to establish and train the network.Then the network is used to solve absolute value of accumulated error of the latter 51 data.The network training results show that the BP neural network has higher generalization ability.Accordingly, the network is used to predict the temperature in 5 different operating conditions.The results show that there is a small absolute error between the predicted values and the experimental values.In addition, bearing preload, spindle speed and the viscosity of the lubricant are investigated so as to analyze the effect on the increase of bearing temperature.The results show that the spindle speed is the main factor affecting the bearing temperature increase.