Thermal equilibrium test of CNC machine tool spindle is a necessary step in thermal error modeling and compensating, and also an experimental method to obtain the thermal characteristics of spindle system, such as the thermal sensitive points, the data of temperature field and thermaldisplacement field and so on. A novel method was presented for fast identification of a machine tool spindle temperature rise, based on a modified adaptive fading unscented Kalman filter (AFUKF). Firstly, a fading factor was introduced into the normal UKF. This factor can be automatically updated by using the residual normalization, and it was also introduced into the gain matrix to reduce the influence of system model deviation on estimation accuracy and enhance the stability of the filter. Secondly, by using adaptive law, the process noise and measurement noise covariance matrix were dynamically adjusted to reduce the influence of external disturbance on temperature rise prediction, so that the better filtering performance can be obtained. A vertical machine tool was used to validate the effectiveness of the presented method. Taking any selected point, we could identify the temperature rise at the point in 28min. The root mean square error (RMSE) between the estimated and measured temperatures in the period of 400min was 0.1291℃, and the error between the estimated and measured steady-state temperature was 0.097℃. The simulation experiments showed that the method of fast identification of machine tool spindle temperature rise can predict the temperature rise of the selected point in a short time, and the prediction results were in good agreement with the results of thermal equilibrium test. The feasibility and validity of the method were verified, and it can greatly improve the efficiency of thermal equilibrium test.