Abstract: Aiming at the problem of difficult diagnosis and maintenance in CNC machine work, a method of fault diagnosis and SVM fault classification using sound fusion features and one class support vector machine (OCSVM) is proposed. Firstly, the audio data of CNC machine operation under different working conditions are collected. The Mel frequency Cepstral coefficient (MFCC) and linear prediction Cepstral coefficient (LPCC) methods are used to extract multi-dimensional features of the data. After dimension reduction and normalization by PCA, the features are fused. Finally, the processed features are detected by OCSVM to determine whether there is a fault and identify the fault category. The normal operation of the CNC machine and nine types of abnormal fault tones are collected as data sets for analysis. Experiments show that the sound feature fusion and OCSVM can realize the accurate diagnosis of CNC machine faults, and the diagnosis accuracy can reach 96.1%. The CNC machine faults can be accurately classified by SVM, and the classification accuracy can reach 93.3%.