A data-driven process-quality-property analytical framework for conductive composites in additive manufacturing

Abstract

Developing conductive composites via additive manufacturing (AM) enables unparalleled flexibility in material composition and product properties. However, it is challenging due to strong interactions between product properties and process parameters. This work aims to investigate and develop a data-driven process-quality-property (PQP) framework to explore the PQP relationships for conductive composites by statistical analysis with the fused filament fabrication (FFF) AM process. Three machine learning (ML) methods are used in the framework to model the PQP relationships. Two conductive composites, copper-based material (Electrifi) and carbon-based material (Protopasta), are investigated. The results show strong correlations among the process parameters, surface quality, and resistance. Among the three ML methods, the artificial neural network achieves the lowest mean absolute percentage error (MAPE) in predictions of resistance for Electrifi (MAPE = 0.1487) and Protopasta (MAPE = 0.1119), respectively. Compared with the trial-and-error method, the PQP framework demonstrates the potential to reduce the time and cost of property prediction and help practitioners to understand AM composites. This framework could contribute to applications and innovations of AM conductive composites.