I. Introduction
The generator is one of the core equipment of nuclear power plant, which has the characteristics of complex equipment structure, various components, special operating conditions (such as high speed, high voltage and high current, continuous operation), and generally serious failure consequences [1]–[2]. With the progress of relevant technology and the upgrading of maintenance theory, generator maintenance has begun to change from the past preventive maintenance mode based on the equipment operation time and carried out regularly to the predictive maintenance mode based on the equipment status and carried out on true demand [3]. Among them, condition monitoring and fault diagnosis are the prerequisite and critical to carry out generator predictive maintenance, prognostics and health management (PHM) [4], which can help to timely grasp and accurately locate fault symptoms at the early stage of generator fault, and make in-depth analysis and arrange scientific solutions according to the type and severity of the fault to ensure safe and reliable operation of generator.