The invention belongs to the field of intelligent fault diagnosis of large rotating machines, and discloses an operation state fault diagnosis system and method for large rotating machines. The systemcomprises a data collection module used for acquiring the vibration data of a large rotating machine at the current operation time, a vibration feature extraction module used for extracting various vibration features according to the vibration data, a vibration feature classification module used for classifying each vibration feature through a classification rule to obtain a corresponding vibration sub-feature, a fault probability calculation module used for presetting multiple fault types and calculating fault probabilities under different fault types according to all the vibration sub-features, and a fault diagnosis module used for making judgment according to the calculation result of the fault probability calculation module and judging a fault type corresponding to the fault probability greater than 40% as the current operation fault of the large rotating machine. According to the invention, intelligent analysis can be carried out according to the vibration data features of the large rotating machine, and the probability of a specific fault can be automatically calculated.