Aiming at the problem that the failure state of concrete is complex and changeable, and the acoustic emission(AE) signal is difficult to be separated from the background noise, the function of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) method was coupled with the support vector machine (SVM) method to identify and predict the concrete destructive AE signal. Firstly, the CEEMDAN method is used to decompose the acquired AE signal, obtaining a certain number of adaptive characteristic modal components (IMF). The correlation coefficient between each component and the original AE signal is calculated, and the IMF component containing more information about the original AE signal is preferred. Secondly, the energy coefficient and waveform coefficient of each component eigenvalue are calculated and inputted into the SVM respectively to classify and identify different failure states of concrete. The results show that the prediction rate of energy coefficient as eigenvalue is 92.39%, and the prediction rate of waveform coefficient as eigenvalue is 91.30%.