Abstract: In order to improve the accuracy of transformer core loosening identification, this paper proposes the chimpanzee algorithm(COA) to optimize the support vector machine(SVM) transformer core loosening identification model, extracts the acoustic patterns of transformer cores with different loosening conditions in no-load experiments, and extracts the mayer factor of cancellation coefficients(MFCC) through the acoustic signals and inputs it into the transformer core loosening identification model used in this paper, and the final results are compared with those of genetic algorithm(GA) and particle swarm algorithm(PSO). The result verifies that the model in this paper has high feasibility in transformer core loosening recognition.