Objective To study methods of drum leak detection based on acoustic emission signal. Methods Acoustic emission sensors, preamplifiers, capture cards and computer were used to set up collection systems; the characteristics of leakage frequency was analyzed under conditions of drain hole diameter of 0.2 mm and no leakage; wavelet packet was used to decompose the acoustic emission signals collected, and the energy characteristics of three features bands (16 ~ 30 kHz, 33 ~ 47 kHz, 95 ~ 102 kHz) of the acoustic emission signals were extracted; the energy characteristics was used as the input feature vector of the support vector machine, and the trained support vector machines were finally used to test whether there was a drum leakage. Results The accuracy rate of the test was 100%. Conclusion Combining the acoustic emission signals of wavelet packet energy and support vector machine method, drum leakage with a diameter of 0.2 mm and above was successfully tested.