To study the anti-resonance of motorized spindle influenced by variations of structural parameters, the structure of one motorized spindle is parameterized firstly using ANSYS and then the modal analysis is carried out. With the ISIGHT platform integrated into ANSYS, several significant geometric and material parameters of the motorized spindle system are selected out as design variables to obtain sufficient samples by Optimal Latin Hypercube method. To fit the function between the low-order natural frequency and the random variables, the BP neural networks are constructed and the reliability limit state equation of the motorized spindle frequency is obtained. Subsequently, the frequency reliability and sensitivity of the motorized spindle, calculated with updating first order second moment method (AFOSM), are verified by Monte-Carlo method. The results show that the density, elastic modulus and the total length of the motorized spindle significantly affect the frequency reliability, in terms of the mean and standard deviation. Meanwhile, the reliability limit state equation constructed by the BP neural networks is relatively rational. The AFOSM to analyze the frequency reliability-based sensitivity is comparatively precise and with higher efficiency than Monte-Carlo method.