Abstract: An impeller is one of the most important core components for the extractor in a sugarcane chopper harvester. Aerodynamic characteristics of impeller mainly dominates the cleaning performance of an extractor. Therefore, this study aims to optimize the structural parameters of impeller for a better performance using numerical simulation on CFD platform and response surface method (RSM). A realizable K - ε turbulence model was selected to calculate the internal flow field of an extractor in CFD simulation. A verification test showed that the maximum error of simulated value was 5.24%, and the average error was 4.29%, indicating a higher accuracy of model. A Box-Behnken design was utilized for the response surface test. Specifically, the installation angle (β), the number of blades (N), and gap ratio (G) were taken as the factors, whereas, the wind speed (v) of an extractor under no-load condition was taken as the evaluation index. Additionally, the impurity rate was used as the evaluation index in the pretest to determine the range of gap ratio. The pretest results showed that there was no significant difference in the impurity rate of impeller with four kinds of G values under the same wind speed. It infers that the G value posed nothing effect on the relationship between no-load wind speed and impurity rate. Subsequently, the level of response surface test was determined after the pretest. The analysis of variance show that N2, β, β2, G and N2×G presented extremely significant effects on the evaluation index, while N and N×G significant influence. An investigation was made to explore the variation in the attack angle in the main work area of a blade and the vortex distribution at the root, in order to reveal the influence mechanism of each factor on the evaluation index. It was found that the v increased with the increasing β, but when β = 30°, the performance of blade tip was reduced with the upward trend of v. Furthermore, the vortex caused the flow field near the impeller to deteriorate, when N and G were at the lowest level. An optimal combination of structural parameters in the impeller were achieved, where N= 4, β = 30° and G = 55%. Two kinds of sugarcane from different sites were used in field experiment, thereby to evaluate the cleaning effect of an extractor on the sugarcane with high impurity content. The sugarcane on the site I was growing well, and the sugarcane on the site II was short while containing more impurities. The field test showed that the impurity rate was reduced by 2.34, 2.2, and 2.4 percentage points, when the driving speed was 2.5, 3.0, and 3.5 km/h, respectively, indicating a better performance under the optimized impeller on the site II. The optimized extractor also increased the loss rate. The loss rate increased by 1.01 and 0.85 percentage points, but the decrease in the impurity rate was greater than the increase in the loss rate, when harvesting the sugarcane in the normal and under growth state in the field.