To address the problem of large redundancy in the audio data of belt conveyors, a method is proposed for identifying key audio data of belt conveyors based on the Improved Honey Badger Algorithm (IHBA) optimized Support Vector Machines (SVM). The Mel Frequency Cepstral Coefficients of the audio data are extracted as features; The Tent chaos mapping is used to increase population diversity, and new density factor and golden sine mechanism are introduced to overcome the defects of the Honey Badger Algorithm (HBA), such as easy to fall into local optimum, slow convergence speed, and low accuracy in finding the best solution. The performance of IHBA is verified through simulation experiments using standard test functions. The parameters of SVM are optimized by IHBA, and the Mel Frequency Cepstral Coefficients are input into the IHBA-SVM model for identification. The results indicate that the IHBA-SVM model can effectively improve the identification rate of key audio data from belt conveyors.