Predictive Modeling for Quality Control in Multi-Stage Manufacturing Using Artificial Intelligence

Multi-stage manufacturing systems involve processes such as machining, stamping, and assembly. Quality control challenges arise from the cumulative variations in the quality characteristics' (QC) produced dimensions compared with the design specifications across the stages. The absence of quality prediction leads to increased scrap, production downtime, rework, and the need for multiple inspections. In such scenarios, in-process inspections are used for successive product's quality evaluation. This paper presents the use of artificial intelligence (AI) models to predict product quality at successive multi-stage production processes. In previous research, most predictive models depend on processes' parameters for modeling the predicted quality. We introduce an approach using quality inspection measurements from stage q-1 as input to predict the QCs at stage q. Consequently, from stage q, we predict the QCs at stage q+j. The AI algorithms of Partial Least Square Regression and Support Vector Machines have demonstrated effectiveness when confronted with limited data. In this paper, these approaches are implemented using a real case at GE Aerospace Bromont Manufacturing, which produces airplane engine parts. The results show that predictive performance at stages q and q+j exhibit a coefficient of determination R2 greater than 0.60, and an average RMSE of 0.00029 and 0.00020, respectively. This emphasizes the ability of AI to predict the QCs, even without the data that indicates the values of the process parameters. These results demonstrate the possibility of using our approach within multi-stage manufacturing processes. Future endeavors aim to extend QCs' prediction to multi-stage prediction using quality prediction at any stage q-1 as input to predict the QCs at the final stage q+j.