In order to analyze the status of the transformer core more effectively, a method for analyzing the characteristics of loosening defects in the transformer core based on chaos theory and sparrow optimization K-means algorithm was proposed. Firstly, the C-C method was used to solve the embedding dimension and delay time of the reconstructed phase space, and reconstruct the phase space of the transformer vibration signal. Secondly, the maximum Lyapunov index of the transformer vibration signal was calculated to determine whether the system has chaotic characteristics, and the correlation dimension and Kolmogorov entropy were selected as a set of chaotic characteristics to identify the degree of looseness of the core. Thirdly, the sparrow search algorithm was introduced into the K-means clustering algorithm to optimize the selection of the initial center cluster and use the average displacement of the cluster center and cluster points as a quantitative feature to describe the loose state of the transformer core. Finally, the two sets of characteristics were combined to form a diagnostic index for the loosening fault of the transformer core, which provides a theoretical basis for the diagnosis of the loosening fault of the transformer core and put the classifier into fault diagnosis to verify the superiority of the combination of the two sets of features.