The lack of virtual prototype impedes the accurate measurement of the maintenance time for mechanical products. Inspired by cumulative time estimation, this paper designs a framework for maintenance time prediction, and details its implementation steps. Then, several maintenance tasks were simulated in an environment constructed on the digital manufacturing software DELMIA. The simulation environment includes virtual prototype, virtual human and maintenance tools. Among them, the posture of the virtual human in the maintenance process was obtained by the VBScript language, which covers four types of motions (i.e. body motion, upper limb motion, grab & place motion and operation motion). Based on the posture data, the time consumption of each type of motion was predicted by the modular arrangement of predetermined time standard (MOD), and then corrected in light of the visibility and reachability of maintenance motions. Finally, the proposed measurement method was applied to the maintenance, disassembly and assembly of an actual engine. The results show that the measured results agree well with the actual results, indicating that our method is feasible and effective.