Abstract: The mechanical defects of gas insulated metal enclosed switchgear (GIS) equipment are the important factor leading to its failures. To solve the problems of missing information and insufficient accuracy in single measurement point and single evidence mechanical defect diagnosis model, this paper proposes a multi-layer fusion data analysis method for GIS mechanical vibration defects diagnosis. First, based on the real GIS vibration simulation platform, the influence of measurement point positions and defect types on vibration behavior are studied through experiments. Then, combined with statistic analysis, mode decomposition, and scale transformation methods, a composite parameter analysis method is proposed for the overall and local information of mechanical vibration signals, and principal component analysis (PCA) is introduced to perform feature layer fusion and dimensionality reduction of multiple measurement points vibration information. Finally, a strong and weak basis learner decision layer fusion mechanism is proposed based on the scaling weight Dempster-Shafer (DS) evidence theory and Bagging voting mechanism, and the GIS vibration defects diagnosis method based on multi-layer fusion data analysis is jointly constructed. Results show that the position of the measuring points and the types of defects have influence on the amplitude, frequency, and distribution range of vibration response. The distribution of composite vibration characteristic parameters shows significant differences. The diagnosis model based on multi-layer data fusion analysis has achieved an effective accuracy of 98.66% for GIS mechanical defects, which improves the diagnostic performance by 5.83% to single classifier. This paper can provide a valuable reference for the diagnosis method of GIS mechanical defects.