Aimed at the problems that the research on mechanical seal life prediction is lack,and the model of degradation of mechanical seal can not be clearly defined,a new prediction method of mechanical seal was proposed based on extracting features from AE signal of mechanical seal and support vector regression.The full life data of mechanical seal were acquired based on the AE technology by experiments,and the data was processed by wavelet denoising.The features in time domain and frequency domain which can represent mechanical seal operation state were extracted by wavelet packet decomposition,and the highdimensional features were treated by KPCA and Mahalanobis distance,the mechanical seal degradation index was acquired after these process methods.The index was inputted into SVR to get a prediction model.The experimental result shows that predication method of mechanical seal life based on AE signal features has strong generalization and high prediction precision.