This technical note concerns the fault diagnosis of stochastic discrete event systems. Specifically, the goal is to characterize a detected fault by estimating its occurrence date. For that purpose, partially observed stochastic Petri nets are used to model the system, the failure processes, and the sensors. From the proposed modeling and collected dated measurements. the probabilities of consistent trajectories are computed and diagnosis in terms of faults probability is established as a consequence. For each detected fault, the probability density function of its occurrence date is approximated. This estimation improves the diagnosis by providing the most probable time interval of the fault occurrence. The interest of fault datation and the applicability of the proposed approach are showed through a case study that represents a distribution system.