Abstract. Along with the urban development, the application of anti-glare glass is more and more widely, high transmittance and low reflection of glass manufacturing technology research is of great significance. Because of the complexity of the anti-glare glass preparation technology, anti-glare glass transmittance is affected by multiple factors. Because of complexity nonlinear relation between the real production data, response surface method can't solve the problem of anti-glare glass preparation process parameters optimization. The BP neural network is proposed in this paper to structure the complex nonlinear model between the design vector with the response vector. BP neural network has high learning and representation ability and have ability by establishing a good mapping model. Using BP neural network model of high generalization ability on the optimal parameter combination optimization search about the corrosion condition of glass, with less test data to get the ideal parameter design.