The Neural Network Prediction Model of Surface Roughness Based on Additional Momentum Method

In order to obtain an effective prediction method about premixed water jet shot peening surface roughness, 45 steel which always exist in project is used as the experimental material and the pressure, nozzle scanning speed and the target distance are as the influencing parameters, thus to do jet peening strengthen experiment with the method of two jars continuous supplying pill, and then get experiment data about the surface roughness changing with the shot peening parameters. On the basis of experimental data, the prediction model of shot peening surface roughness is established with the traditional multiple regression analysis method and the BP neural network of additional momentum item respectively. The application of model predicted that, in the same situation the BP neural network prediction model with additional momentum item has higher precision and better predictive ability than that of the prediction model with multiple regression. It can meet the requirement of prediction accuracy with shot peening surface roughness in the industrial production, and it has greater practical value.