A Dynamic Process Adjustment Method Based on Residual Prediction for Quality Improvement

Dynamic process adjustment is an important way for improving product quality in industry production process. Focusing on the process monitoring and feedback adjustment, a residual prediction method for quality improvement is proposed in this paper. This method deals with the problem of dynamic process adjustment in three steps: 1) definition of adjustment rules; 2) building of residual series model; and 3) prediction of adjustment amount, respectively. First, the cost function and quality loss are combined to define the adjustment rules, which is used for judging whether the process should be adjusted. Second, a multivariate residual series model is built to illustrate the time series between input variables (technological parameters) and output variables (quality indices). Third, the double-order weights are introduced to support vector machine to build a prediction model for predicting the adjustment amount of controllable variables. In this way, the adjustment decisions can be made and conducted to realize the dynamic process adjustment. At last, to demonstrate the practical usefulness of the proposed method, a case study about coating process of purifier carrier is provided to validate its effectiveness. The result shows that the proposed method has good performance for industry application.