Many manufacturing environments are beginning to leverage Industry 4.0 technologies and methodologies built upon a networked OT environment. 3D printer farms, as an example, contain a collection of AM devices with varying networking capabilities. While AM is praised for its affordability and flexibility, it suffers from quality and reliability issues. In this work, the networked environment is leveraged to collect a shared dataset across three 3D printers, and federated learning is employed to predict and compensate for geometric deviations in printed parts. Leveraging a 3D CNN developed in previous literature and deployed in a federated environment, 3D scans of printed geometry and process parameters are combined between three printers of similar models to determine whether (1) the individual geometric error distributions are distinct enough to cause negative transfer and (2) total bandwidth can be reduced by transferring weights rather than the full dataset. This work determined that negative transfer did not occur. Furthermore, 90% of the network bandwidth can be saved by leveraging federated transfer learning with minimal or no loss to model performance. Additionally, the lack of negative effects creates an opportunity for parallel data collection for model training.