A method (200) for identifying abnormalities in a mechanical apparatus or a mechanical component, comprising at least the following steps of: i) acquiring at least two types of undersampled measurement data collected in or on a mechanical apparatus or mechanical component, wherein the at least two types of undersampled measurement data are different from each other in any one or both of the following aspects: a delay △t relative to the occurrence time t0 of a trigger event, and a used frequency fs; and ii) identifying abnormalities in the mechanical apparatus or mechanical component on the basis of the acquired at least two types of undersampled measurement data by using a machine learning-based abnormality identification model used to identify abnormalities in the mechanical apparatus or mechanical component. Further, provided are a method for training the machine learning-based abnormality identification model, a computer apparatus, a computer program product, and a detection apparatus. The computer apparatus comprises a processor (10) and a computer scale storage medium (20) communicatively connected to the processor (10), so as to form a cost-effective and reliable fault diagnosis or predictive maintenance solution.