Abstract: Sufficient clearing time, lower impurity content, and loss rate are highly required in the cleaning device of the harvester. The higher generality, larger size, and sieve surface of the device are also demanded during the clearing operation of the two crops of soybean and corn. In this study, an improved cleaning bench was optimized for the 4LZ-3.0Z small self-walking combined grain harvester. A testbed was also built for the cleaning device of soybean and corn. Firstly, the discrete element model was established for the main explants of corn cleaning using EDEM software. Secondly, the EDEM-Fluent coupling simulation was conducted to determine the trajectory and velocity changes of materials in the sieving box during cleaning. A comparison was made on the original sieve box A (Straight upper sieve and lower sieve), the improved sieve box B (Upper sieve section and lower sieve concave surface), and the improved sieve box C (Concave surface is larger). Thirdly, the force analysis was then verified to be the optimal design. The experimental factors were selected as the vibration frequency of the vibrating sieve, the inclination angle of the upper sieve, and the inclination angle of the lower sieve, particularly for the generality and performance of the cleaning device. The single factor test and Response Surface Method (RSM) were carried out for the soybean and corn, with the impurity rate and the loss rate of the cleaning as the experimental indexes. Finally, the best parameter combination was obtained to clarify the influence of experimental factors on the indicators in the cleaning device for two crops. The simulation results showed that the grain movement in the process of cleaning was consistent with the force analysis. Specifically, there was miscellaneous accumulation in the sieve box C. Much more contribution of cleaning was achieved in the grain penetration area and the material movement trend in the sieve box B, compared with the box A and C. The bench test results of the two crops showed that the three selected test factors presented a significant influence on the parameters (P<0.05). Once the vibration frequency increased, the loss rate and impurity rate of the grain cleaning showed a trend of first decreasing and then increasing during the cleaning of the two crops. By contrast, the impurity content decreased first and then increased, while the loss rate continued to decrease, as the inclination angle of the upper and lower sieve increased significantly. The RSM showed that the optimal working parameters of the equipment for the soybean cleaning were the vibration frequency of 5.9 Hz, the inclination angle of the upper screen at 10.5°, and the inclination angle of the lower screen at 6.5°. The optimal cleaning was achieved in this case, where the average impurity rate and loss rate were 0.622% and 0.439%, respectively. In corn cleaning, the optimal working parameters of the cleaning device were the vibration frequency of 4.7 Hz, the inclination angle of the upper screen at 10.3°, and the inclination angle of the lower screen at 8.6°. Correspondingly, the optimal cleaning was achieved, where the average impurities rate and loss rate were 0.956%, and 0.771%, respectively. Therefore, the impurity content and loss rate of the improved soybean cleaning were reduced by 38.8% and 45.9%, respectively, compared with the original. In corn cleaning, the impurity content and loss rate were reduced by 29.9% and 30.1%, respectively. This finding can provide a theoretical basis for the design soybean and corn combined harvester in the soybean and corn intercropping.