The cleaning performance of cleaning device greatly affects the working performance of the machine. In order to overcome the shortcomings of the uneven airflow field in the longitudinal axial flow fullfeeding airandscreen cleaning device, the mixed flow field produced by the fan and the threshing roller was numerically simulated with computational fluid dynamics (CFD) software. A structure of fan with double airducts and six airoutlets was proposed based on the longitudinal axial flow fullfeeding cleaning device. By comparing and analyzing the improved structures, the air speed of the sieve has the influence on the distribution of the full airflow field. At the same time, the orthogonal simulation was carried out on a new type of cleaning device with the three factors (the fan speed, the incident angle of the fan, the included angle of the sieve) in Z section, and the influence of the three factors was analyzed. The optimal combinations are as follows: the incident angle of the fan is 30°, the included angle of sieve is 40°, the speed of fan is 1900r/min. With the best combination of these structural parameters, the speed of overall air flow in the front and the rear of sieve is uniform and efficient, and it is more beneficial to highload cleaning. Finally, field tests verified the accuracy of the improved structure and optimization program. The loss rate of the rice is 0.91% and the impurity rate is 0.87%. The loss rate of the wheat is 0.82% and impurity rate is 0.76%, it greatly improved the cleaning performance of the longitudinal axial flow.