Measurement and Process Relation for Cast Surface Defects
Abstract
This thesis presents an in-depth investigation into the classification and prediction of surface defects in aluminum sand castings using a combination of experimental trials and machine learning techniques. Key process variables such head height, pouring temperature, resin %, sand type and process were systematically varied to study their influence on defects like veining, and burn-on. Comparative analysis indicated superior performance of the deep learning model in handling nonlinear interactions between process parameters. The study emphasizes the importance of integrating data-driven models with foundry practices to enhance surface quality, reduce rework, and support intelligent decision-making in casting operations.
This paper has been withdrawn.