Abstract: In order to reduce the accident rate of CNG pressure vessel, acoustic emission signal features of the vessel when under tensile stresses were investigated. According to the failure process of CNG cylinder, tensile metal specimens were used to simulate the process and to collect the signals. Through the analysis of signal, the limitations of characteristic analysis on time domain and frequency domain were pointed out. So the signal was analyzed by the method of wavelet. Results indicated that different types of signals had different energy coefficient ratio in the same frequency. According to the characteristics, the energy ratio characteristic was used to characterize the AE signal. Recognition results indicated that this recognition method of BP network could well identify the signal type.