Abstract: In recent years, the research on fault prediction of discrete event systems has received considerable attention. In this research, the prediction of patterns fault for decentralized discrete event systems is investigated. Firstly, the notion of predictability of patterns fault for decentralized systems is defined and the patterns fault predictability of distributed discrete event systems formalized. By constructing a patterns fault recognizer, patterns fault is identified from all behaviors of the system, and the unobservable closure is constructed for different observation points of decentralized system. On this basis, the patterns fault prediction verifier is constructed by combining the events observed at each station, which solves the problem of patterns fault prediction of decentralized discrete event system. A sufficient and necessary condition for determining the predictability of the decentralized systems is obtained. The corresponding patterns fault prediction algorithm is proposed and implemented. Finally, the complexity analyses of the construction of the decentralized patterns fault prediction verifier and the patterns fault prediction algorithm are carried out.