Fault signal extraction method and device of mechanical device, equipment and medium

The invention discloses a fault signal extraction method and device of a mechanical device, equipment and a storage medium. The method comprises the following steps: constructing a noise distribution signal according to spectrum correlation characteristics of a vibration signal; taking the vibration signal as a to-be-filtered signal, filtering the to-be-filtered signal based on a filter, outputting a filtered signal of the iteration, and determining a signal-to-noise kurtosis ratio matched with the iteration; and according to the signal-to-noise kurtosis ratio, on the basis of a gradient descent model, updating the filter, updating the to-be-filtered signal, returning to execute filtering of the to-be-filtered signal until an iteration termination condition is satisfied, outputting a filtering signal of the last iteration, and determining a fault signal. According to the technical scheme, the problem that the fault features of the fault signals extracted from the vibration signals of the mechanical device are