Fault detection method and device for monitoring state of tower drum of wind generating set

The invention discloses a fault detection method for monitoring the state of a tower drum of a wind generating set. The fault detection method comprises the following steps: acquiring state data of the tower drum; carrying out marking processing on the obtained tower drum state data; the tower drum state data obtained after marking processing are preprocessed; carrying out segmentation processing on the preprocessed tower drum state data, so that the preprocessed tower drum state data are segmented into a plurality of fixed-length time sequence fragment data; performing feature extraction processing on each piece of time sequence fragment data; constructing a fault detection model, and sending all the time sequence fragment data and the feature information thereof to the fault detection model for training; and deploying the trained fault detection model to an actual system, and carrying out prediction and fault detection processing on the obtained tower drum state data. The invention further discloses a fault