Under Industrial 4.0 concept and the analysis of big data issues, to meet the customers demand and improve producing quality quickly is the major topic now. This study applies the real company manufacturing data adopting intelligent manufacturing concept integrates neural network system、Taguchi quality engineering improving process and response surface methodology to screening the critical parameters and training by the data. Using the 10 real varied parameters and 230 data selected is trained in the network. The other data is used to verify the model’s utility. In order to help process engineer analyze and search the critical factors for reducing the reliability of manufacturing experience as a point of view about data analysis. According to the result of this study, the improved scope achieves 93.33%.