Numerous deadlock control policies have been developed for automated manufacturing systems. However, a majority of the studies assume an absence of resource failures within the system and overlook the consideration of operation time. This paper presents a robust deadlock prevention policy that is based on the state space of time Petri nets incorporating unreliable resources. The proposed approach involves the classification of state classes into two distinct sets: the robust legal state class set and the illegal state class set. Subsequently, controllers are designed to prevent the illegal state classes, thereby guaranteeing the liveness of the controlled system regardless of any resource failures. Finally, an example is presented to show the efficiency of the proposed methodology.