Abstract: In order to solve the problem that the current event observability of stochastic discrete-event systems (SDESs) is fixed in advance and cannot be changed dynamically with system evolution, a decentralized fault prediction method based on dynamic observations is proposed. First, the D-copredictability of decentralized SDESs under dynamic observations is formalized. Then, in order to conduct the fault prediction of decentralized SDESs with dynamic observations, a D-copredictor is established using local stochastic predictors. A necessary and sufficient condition of D-copredictability for decentralized SDESs based on the D-copredictor is presented, while an algorithm for verifying the D-copredictability of SDESs is proposed. The proposed verification algorithm can be used not only for online fault prediction, but also for offline predictability verification. Finally, the applicability of the verification algorithm for decentralized fault prediction under dynamic observation is analyzed.