Provided is an abnormality diagnosis device including: a physical quantity measurement processor configured to acquire, as time-series data, a physical quantity of an electric motor or a load; a feature parameter calculation processor configured to calculate feature parameters; a correlation function creation processor configured to create a correlation function for two or more feature parameters, and calculate a correlation coefficient corresponding to a measurement result of the physical quantity that changes depending on an abnormality occurrence state; a database configured to store in advance reference data in which a remaining time until a failure is caused and the correlation coefficient are associated with each other; and a diagnosis processor configured to extract from the reference data the remaining time corresponding to the correlation coefficient, and identify the remaining time until the electric motor or the load causes a failure or a part that has caused a failure.