Hidden Markov model (HMM), which is initially employed in the field of speech signal processing, is introduced into DC/DC converter failure detection. HMM is depicted briefly, and a new method to detect convert failure based on HMM is presented. First, this paper analyzes the failure mechanism of DC/DC convert, selects the output ripple voltage and inductor current and output power as characteristic vectors. Then, it models for every state utilizing HMM. Extensive experiments have been conducted on a typical boost circuit model. The calculation result shows that this method is able to detect failure correctly