Abstract

Mass personalization in metal manufacturing requires processes that accommodate unrestricted geometric variability while preserving cost scalability and surface quality. Additive manufacturing enables geometric freedom but is limited in material and surface reliability, whereas electroforming provides high precision and parallel metal deposition but is sensitive to electrolyte condition and process variability. This thesis addresses this challenge by integrating a hybrid additive manufacturing–electroforming process architecture with a data-driven framework for adaptive electroforming control.
In this work, a hybrid workflow was developed in which additively manufactured polymer molds enable low-cost personalization, while electroforming provides scalable, high-precision metal fabrication. Experimental validation confirmed the feasibility of this approach for producing personalized metal components in a cost-efficient manner. However, sustaining consistent surface quality within this hybrid process requires systematic control of electroforming behavior as the electrolyte evolves during repeated use.
To support this requirement, a comprehensive experimental and data-driven framework was established. An open dataset for additive-free copper electroplating was constructed, comprising 114 electrolyte baths across multiple CuSO₄ concentrations, aging stages, and pulse–reverse deposition conditions. Analysis showed that surface roughness and gloss cannot be reliably predicted using static process parameters or conventional bath measurements alone.
An operational bath-state representation was therefore introduced, inferred from measurable physicochemical bath properties and dynamic electroplating pulse-response features. Machine-learning models were trained to predict surface properties, with Random Forest providing the most robust performance. Scenario-based modeling demonstrated that data-driven adjustment of electroforming parameters can maintain target surface quality within defined chemical and operational bounds, enabling adaptive control within the hybrid manufacturing workflow.