Abstract: The traditional reliability evaluation of the bearings of machine tool is not related to the condition characteristics of the using process. A condition maintenance decision utilization of condition-based maintenance was proposed. First, operation condition information of the bearings was obtained by the senor, such as the acceleration senor. Then, the condition characteristics of the root mean square and frequency center were extracted, Weibull proportional hazard model (WPHM) was also establish, which regarded Weibull distribution as failure rate function. Then, combined with the lifetime data of similar parts, three parameters were estimated through Newton-Raphson method. Finally, a maintenance decision model based on WPHM was established. Using the similar type of parts lifetime data to determine the maintenance failure threshold, according to failure threshold, maintenance decision can be conducted. Results show that this maintenance decision model can improve the availability of the bearings greatly, and make full use of the effective life of the machine tools.