Fault prognosis is the fundamental of realizing condition based maintenance. Uncertain problems exist in the general process of fault prognosis. Therefore, the key issues of fault prognosis based on uncertainty are summarized. Then existing fault prognostic methods are categorized into four types, i.e. methods based on randomness, fuzziness, grey property and integrated property. A review is made on the current research status and shortcomings of the main methods. The development trends of fault prognostic methods based on uncertainty are indicated, and the feasibility of possible fault prognostic methods based on interval uncertainty is discussed.