This work develops a digital twin framework to optimize printed electronics by integrating virtual metrology and machine learning to predict printing outcomes. Three printing platforms (NanoDimensions DragonFly, Voltera V-One, and BotFactory SV2) were used to fabricate parametrized layouts under varied process conditions. A virtual metrology workflow was implemented to extract geometric features from scanned images, enabling non-contact dimensional measurements. Electrical resistance was measured via four-probe testing and served as ground truth. Two modeling approaches were implemented: a feature-based pipeline for predicting resistance and dimensions, and an image-to-image translation model using HyperPix2Pix to reconstruct printed structures from layout images. These models were integrated into a closed-loop digital twin that combines image prediction, metrology, and resistance modeling, supporting simulation-driven process monitoring and yield optimization. The results highlight the potential of data-driven tools to capture printing variability and guide design and process decisions in printed electronics manufacturing.