The invention discloses a track circuit fault precursor discovery method based on small fluctuation detection, and the method comprises the following steps: S1, selecting m monitoring data points as data windows, and recording the variance value of each data window; S2, obtaining a variance sequence in a period of a normal working state, and extracting a variance threshold from the variance sequence to detect abnormal small-amplitude data fluctuation; and S3, summarizing the abnormal events under the condition of small-amplitude fluctuation on site into n types, wherein n is an integer greaterthan 1; and S4, extracting t fluctuation characteristic values capable of covering all types of abnormal events to serve as characteristic attributes for judging the fault mode. Wherein t is an integer greater than 1; and S5, on the basis of the decision tree classification model, carrying out real-time classification on the detected monitoring data under the abnormal small-amplitude fluctuation,and identifying the category of the abnormal event. The method is high in recognition accuracy, is quick, simple and convenient in detection of abnormal small-amplitude fluctuation, and is suitable for a maintenance site.