Tobacco leaves contain a certain amount of stalk sticks after being shred, and most of these stalk sticks are removed in the silk line by local wind picker. However, in the actual production process, due to the influence of leaf silk moisture, flow rate and the cleaning degree of dust filter bag, the removal effect of stem air in leaf silk after selection is not stable. In order to solve this problem, the regression equations of dust removal speed of blade wire local air separator, frequency of high-speed belt conveyor and stalk removal rate of two key factors were established through experiments, so as to calculate the best dust removal speed of each brand during production and the dust removal speed of each brand under no-load condition. Then, the difference of these two dust removal speed was added to the best dust removal speed of each brand before production. The optimal no-load dust removal speed of each brand before production was obtained. At the same time, the linear regression equation of dust removal wind speed and dust removal fan frequency in no-load state of equipment was established through experiments. Before production, the optimal no-load dust removal wind speed was substituted into the equation, and finally the optimal theoretical frequency of dust removal fan was obtained. After the system error prevention detection, the theoretical frequency was automatically brought into the control system, so as to stabilize the stalk removal rate of local air separator of blade wire.