Abstract: In recent years, frequent equipment failures have been observed in GIS three-position switches due to inadequate mechanical movement. In order to enhance the detection and maintenance capabilities for mechanical defects causing the incomplete movement of GIS three-position switches, in this paper, a 126 kV fully-functional three-position switch test platform was firstly utilized to simulate three typical mechanical defects, namely, mechanism sticking, transmission decoupling, and motor stalling. The motor currents, angular displacement trajectories, and vibration signals were experimentally collected during the opening and closing actions of the three-position switch under normal conditions and with the three typical mechanical defects. Subsequently, the "current-mechanical" characteristic parameters under the three types of defects were extracted and analyzed. Finally, the random forest algorithm was employed to construct a diagnostic model for typical mechanical defects in GIS three-position switches and validates its effectiveness. The results indicate that individual characteristics such as motor current, angular displacement, or vibration signal patterns alone are insufficient for effectively distinguishing different types of mechanical defects. However, the "current-mechanical" fusion feature parameter demonstrates significant effectiveness in distinguishing the three typical mechanical defects. The random forest diagnostic model, based on the "current-mechanical" fusion feature parameter, can reliably diagnose mechanism sticking, transmission decoupling, and motor stalling.The result can provide a technical reference for improving the safety operation of GIS three-position switches in practical engineering applications.