The invention relates to a method of identifying bearing failures, in particular to an infrared waveform based intelligent identification method for a railway bearing. The method is based on a dynamic standard waveform of a specific detecting spot, a vertical scan in the original method is substituted with a once horizontal scan for processing the infrared waveform, and therefore, the nature of distinguishing a normal infrared waveform from an abnormal infrared waveform is reflected. With the adoption of the technical scheme, the intelligent identification method based on waveform characteristics is standard, concise and good in effect, and the workload can be effectively reduced on the premise that the effect of the method is better or not worse than that of artificial identification.