Abstract: Abstract: To determine the optimal parameter combination of the machine mounted seed cotton pretreatment apparatus of stripper cotton harvester, in 2014 and 2015, we conducted optimization experiments by using an orthogonal and a response surface experiments designed by Box-Behenken in Anyang, Henan. The current stripper cotton harvester takes in lots of large impurities like branches, which are unable to be cleaned out during harvesting. This becomes the most limitation for the machine, and is caused by its property of being easy to twist with seed cotton. In the past experiments, researchers found that most branches mixed with seed cotton reaching a length at 20 cm, which makes the problem harder to be resolved. Small impurities such as bolls and leaves, though take a smaller percentage, are not only difficult to be cleaned out once they entered pretreatment apparatus due to its brittleness, but also easy to be crushed down and twisted deeply with seed cotton which would directly influence the price of seed cotton. The stripper harvester we used in this experiment was a brush rolling cotton harvester invented by Nanjing institute of Agricultural Mechanization，Ministry of Agriculture, on which the seed cotton pretreatment apparatus was mounted. The apparatus is consisted of a mesh box, a material striking and opening apparatus and 'one-cleaning and two-recycling' type cleaner. Seed cotton cleaned collected from the cotton room of the harvester and impurities recycled from the pretreatment apparatus would be cleaned again by hand to determine the impurity rate and loss rate of seed cotton. In the experiments, we took the moving speed of the cotton harvester, linear speed of the cleaning rollers and the distance between the cleaning bar and the cleaning roller as experiment factors, and the impurity rate of seed cotton and loss rate of cleaning as response functions to optimize the construction configurations and working parameters of the apparatus. The orthogonal experiment conducted in 2014 showed that the apparatus worked best when the moving speed of the cotton harvester, the linear speed of the cleaning rollers, and the distance between the cleaning bars and the cleaning roller were 1.3 km/h, 10.8 m/s and 10 mm, respectively. However, the orthogonal experiment only showed the better parameter combination of the level selected which was not accurate enough. Thus, to testify the result of the experiment conducted in 2014 and obtain more precise results, a Box-Behnken response surface experiment was conducted in 2015. The experiment showed that, with a result being close to the former one, that the apparatus worked best when the moving speed of the cotton harvester, the linear speed of the cleaning rollers, and the distance between the cleaning bars and the cleaning roller were 1.3 km/h, 9.83m/s and 13.23mm, respectively, with the impurity rate and loss rate being 5.56% and 0.84%, respectively. At the same time, we established the mathematical regression model of factors and performance indexes, and determined the effects of the factors on the impurity rate and loss rate. The data showed that the technical indexes could meet the demands of design. The cotton fiber detection conducted by authorized agencies showed that indexes of machine picked seed cotton reached standard or above the standard, and could meet the demand of cotton production.