Abstract: Mechanical transmission devices are the core components of fully automated production lines. Traditional fault diagnosis methods mainly rely on manual inspections or sensor monitoring. The former is time-consuming and prone to missed inspections, while the latter is limited due to the difficulty of sensor installation. Therefore, this paper proposes a fault diagnosis method for mechanical transmission devices based on the operating parameters of servo motors, which can quickly diagnose faults without the need for additional external sensors. First, the torque current and speed of the motor operation are used to calculate features, including output power, loss power, copper loss, iron loss, and other parameters. Next, the PCA algorithm is utilized to process the aforementioned high-dimensional features to identify low-dimensional features that can characterize the faults of mechanical transmission devices. Finally, these low-dimensional features are input into an improved KNN algorithm that incorporates the idea of K-means, achieving fault diagnosis of mechanical transmission devices. Experimental results show that the system can effectively diagnose connecting rod faults and bearing faults, with a fault diagnosis accuracy of over 95%. It can also be applied under various working conditions, solving the problems of traditional fault diagnosis schemes that require the installation of external sensors and are inefficient, thus meeting the demands for efficiency and reliability in industrial applications.