The reciprocating compressor is an important equipment for petrochemical production, and its safe and stable operation is closely related to the operation states of the gas valve. For realizing the rapid diagnosis of the gas valve fault of the reciprocating compressor, the wavelet packet decomposition is used to extract the fault characteristics. Based on support vector machine (SVM) method, the gas valve fault is identified. The grid search method is used to optimize the parameters. The model for fault diagnosis of compressor gas valve is built, which is combined with wavelet packet decomposition and SVM. And the effectiveness of SVM in fault diagnosis of the compressor gas valve is verified, which simplifies the traditional process of judging the type of gas valve fault by experienced personnel, and provides a theoretical basis for practical problems such as compressor gas valve fault analysis, maintenance and replacement.