The work aims to propose a fault identification method based on parameter optimization with respect to the problem of low recognition rate of fault in power machine bearings of packaging machinery. Firstly, principal component analysis algorithm was used to extract the principal component of vibration data of power machine in packaging machinery and reduce the correlation between the data. Then, LSSVM was used to identify the fault in various kinds of data samples. In order to overcome the local extremum and poor convergence precision of LSSVM penalty factors and kernel function parameters, an ICS algorithm was proposed for the optimization of LSSVM state parameter to improve the recognition rate of power machine bearings in packaging machinery. Taking the measured vibration data of packaging machinery in candy factory as an example, the validity of the proposed method was verified. Experimental results showed that the algorithm could identify the fault in all kinds of power machines with high precision when the type of fault in power ma-chine bearing of packaging machinery. The proposed algorithm realizes the adaptive selection of the classifier parameters, and provides a reliable method for improving the recognition rate of fault diagnosis of power machine bearings in packaging machinery.