Cleaning is one of the most important working processes of corn kernel harvester. However, high loss and high impurity seriously restrict the effect of cleaning process. The impurity rate and loss rate were directly affected by working parameters of cleaning devices. So working parameters optimization experiment of cleaning devices was carried out by using corn kernel harvester. Fan speed, vibration frequency and opening degree of upper screen were selected as experiment factors. The better levels of three experiment factors were obtained by single factor experiment. The better levels of fan speed were ranged from 800r/min to 1000r/min. The better levels of vibration frequency were ranged from 6Hz to 8Hz. The better levels of opening degree of upper screen was ranged from 15mm to 25mm. Based on single factor experiment results, the optimal combination of three test factors and regression model were obtained by orthogonal experiment. The optimal factor combination of cleaning impurity rate was fan speed of 1000r/min, vibration frequency of 7Hz, and opening degree of upper screen of 20mm. The optimal factors combination of cleaning loss rate was fan speed of 900r/min, vibration frequency of 6Hz, and opening degree of upper screen of 20mm. The optimal factor combination of cleaning comprehensive weighted index was fan speed of 900r/min, vibration frequency of 7Hz, and opening degree of upper screen of 20mm. Three regression models were verified by field test. The relative error of regression model of cleaning impurity rate was 5.56%. The relative error of regression model of cleaning loss rate was 5.10%. The relative error of regression model of cleaning comprehensive weighted index was 4.60%. The results showed that three regression models were reliable.