Addressing the challenges of highly non-stationary vibration signals and strong feature coupling in On-Load Tap-Changer (OLTC), this paper proposes a hierarchical multi-domain feature diagnosis method based on Markov Transition Field (MTF) and Spectral Markov Transition Field (SMF). The MTF converts temporal state transition characteristics of vibration signals into Markov transition matrices, while the SMF constructs spectral energy distribution matrices. Singular Value Decomposition (SVD) is then applied to these matrices to extract the first four stable singular values, forming primary feature vectors. Subsequently, Principal Component Analysis (PCA) combined with Support Vector Machine (SVM) is employed for dimensionality reduction and classification refinement. Experimental results demonstrate that the proposed method achieves an average accuracy of 98.67% in diagnosing typical OLTC mechanical faults, outperforming conventional approaches by 5 to 12%.