The invention discloses a blade-root bolt fracture fault detection method which comprises the steps of S01, selecting multiple characteristics for preprocessing, standardizing three variables describing the running status of a blade by adopting mean value-variance, then summing, and outputting; S02, using data without bolt fracture fault for training an LSTM normal model; using data before bolt fracture fault occurrence for training an LSTM fault model; putting the data of S01 into the LSTM normal model and the LSTM fault model, and extracting an error vector characteristic; S03, substitutingthe error vector characteristic into a random forest algorithm for training a random forest model; and substituting fan operation data into the random forest algorithm model for carrying out fault diagnosis. The invention further discloses a computer readable storage medium stored with a computer program, and the program implements the method when being executed by a processor. The detection method and the medium provided by the invention have the advantages of high automation degree, high detection accuracy, cost reduction and the like.