Abstract: In feature value extraction processing step, feature values are extracted from time-series vibroacoustic data in equipment in operation. In machine learning diagnosis step, accumulated are feature values corresponding to normal equipment-operation-state and general measurement data on equipment. Probability distribution model related to correlation between general measurement data and feature values corresponding to normal equipment-operation-state and constructed through machine learning is fed with general-measurement data and feature values extracted in feature value extraction processing step to calculate degree of anomaly based on probability distribution model.

Abstract: A damage probability calculating apparatus for calculating a damage probability of a specific estimation target includes: a storage unit 1 that stores a damage rate function representing a relationship between a damage rate and time; a damage rate function correcting unit 3 that corrects the damage rate function on the basis of examination data acquired at the time of general examination of the estimation target; and a damage probability calculating unit 4 that calculates the damage probability on the basis of the damage rate function corrected by the damage rate function correcting unit 3.