Power transformer is an important equipment of power system. For transformer mechanical fault

vibration analysis method has scientific and strong feasibility.In this paper

a transformer vibration measurement system based on vibration sensor array is built.The vibration distribution of single-phase transformer tank surface is studied.The vibration cloud pictures of the oil tank under normal and fault conditions are measured respectively.Histograms of oriented gradient（HOG） of the vibration cloud pictures are extracted and the support vector machine（SVM）is selected as the learning method. Two types of classifiers are formed to distinguish normal and fault states of transformer winding. Finally

a fault diagnosis method of transformer winding loosening based on vibration image HOG and SVM is proposed.The experimental results show that the method has a high recognition rate.