In view of the fact that the vibration signal of rolling bearing under non-stationary conditions is vulnerable to the interference of velocity fluctuation, amplitude or frequency modulation, noise and other irrelevant components, which leads to the complexity of the generated time-frequency plane, and makes it difficult to identify the fault characteristic frequency of rolling bearing. A novel method based on adaptive chirp mode decomposition and ridge detection is proposed. The proposed method constructs a high-resolution timefrequency representation, improves the accuracy of diagnosis, and has very strong adaptability. Through the analysis of the vibration signals of rolling bearing with different health conditions, it is found that the proposed method is very suitable for fault diagnosis of rolling bearing under variable speed conditions, and the diagnosis effect is better than the newly developed time-frequency analysis method.