Engineering Effect Equivalence Enabled Transfer Learning

摘要

Engineering effect equivalence (EEE) reflects the fact that different engineering factors generate the same effect on the final product quality. For example, in CNC machining process, the deviations of the machine tool and the fixture may induce the same machining error on the final product. Building a quality prediction model for an engineering process (e.g., CNC machining and additive manufacturing) with multiple engineering factors usually requires a large amount of training data. Utilizing EEE, this paper explores the similarity among multiple manufacturing processes to establish engineering-driven transfer learning, which greatly improves the performance of modeling and prediction of process variations/quality for those engineering factors having only a limited amount of training data. The method has been demonstrated by a machining case study.