Equipment feature extraction and state monitoring method based on triple approximate entropy

The invention relates to the field of equipment state monitoring and fault diagnosis, and discloses an equipment feature extraction and state monitoring method based on triple approximate entropy, which comprises the following steps of: calculating a rearrangement signal and an alternative signal of an original vibration signal by using a data rearrangement algorithm and an alternative algorithm respectively; then respectively calculating the approximate entropy of the original signal, the rearranged signal and the substitute signal, constructing a triple approximate entropy feature vector to describe the operation state of the equipment, obtaining the vibration signals of the equipment in different operation states, calculating the confidence interval of each component of the feature vector of the equipment in different operation states on the 95% confidence level, and calculating the confidence interval of each component of the feature vector of the equipment in different operation states; and if the featur