Qualified Rate Prediction of Typical Workpieces in Discrete Manufacturing Process

In the field of high-end manufacturing, with the in-depth advancement of digital transformation, massive data becomes available for analysis and “learning”. This makes, the decision-making and control links during the manufacturing process smarter. In literature, few works considered the prediction of the qualified rate of typical workpieces in the discrete manufacturing process. This paper, fills this gap by proposing a prediction model based on machine learning methods. By analyzing the dataset from a competition, and performing feature engineering, we carefully select the key factors that correlate to the qualified rate. These features are fed into CatBoost for continuous training. Finally, the parameters of the model are finely tuned to assure the prediction accuracy. The experimental results show a significant improvement of system robustness. This also improves the accuracy of prediction of the qualified rate of the workpiece quality.