In this paper, a formal verification approach for diagnosability analysis of intermittent faults is proposed. In this approach, the industrial systems are abstracted as discrete-event systems (DES) and modeled by finite state automata (FSA), then a model-checking framework is set to deal with diagnosability issues. Intermittent faults are defined as faults that can automatically recover once they occur. We first revisit two existing definitions of diagnosability of intermittent faults, regarding the occurrence of faults and their normalization (i.e., disappearance of faults). Then, necessary and sufficient conditions are developed based on the twin plant construction, and reformulated as linear temporal logic (LTL) formulas in order to use model-checking for actual verification. A benchmark is used to illustrate the contributions discussed and to assess the efficiency and the scalability of the proposed approach.