The sensor installation position is limited and the collected signal is easily disturbed by external noise in the existing RV reducer condition recognition. The comprehensive utilization of servo characteristic information to monitor the load condition of the RV reducer can improve the service performance of industrial robots in the field of manufacturing. Firstly, according to the structural parameters and working mechanism of the RV reducer, analyzed the correlation between the input speed and the important frequency, the servo characteristic information and the load of the RV reducer. Then, constructed the correlation identification model between servo feature information and RV reducer load based on the K-means clustering algorithm. Finally, the experimental platform of the RV reducer is built to collect the feedback information of the servo system under different load conditions. After corresponding processing, the correlation identification model is used to realize the accurate identification of load state, and the recognition rate is as high as 97.45%. This paper can provide technical support for monitoring the running state of RV reducer based on servo characteristic information.