[Paper Review] Physics-Informed Bayesian Learning of Electrohydrodynamic Polymer Jet Printing Dynamics

Athanasios Oikonomou, Θεόδωρος Λούτας+5 more|arXiv (Cornell University)|Apr 16, 2022

Electrohydrodynamics and Fluid Dynamics4 citations

TL;DR

This paper introduces GPJet, a physics-informed Bayesian learning framework that enables real-time, self-calibrating electrohydrodynamic polymer jet printing by integrating machine vision, physics-based modeling, and machine learning. It achieves accurate jet dynamics prediction with minimal experimental data by fusing video-derived features with physical laws, demonstrating a closed-loop, data-efficient approach to optimizing complex AM processes.

ABSTRACT

Calibration of highly dynamic multi-physics manufacturing processes such as electro-hydrodynamics-based additive manufacturing (AM) technologies (E-jet printing) is still performed by labor-intensive trial-and-error practices. These practices have hindered the broad adoption of these technologies, demanding a new paradigm of self-calibrating E-jet printing machines. To address this need, we developed GPJet, an end-to-end physics-informed Bayesian learning framework, and tested it on a virtual E-jet printing machine with in-process jet monitoring capabilities. GPJet consists of three modules: a) the Machine Vision module, b) the Physics-Based Modeling Module, and c) the Machine Learning (ML) module. We demonstrate that the Machine Vision module can extract high-fidelity jet features in real-time from video data using an automated parallelized computer vision workflow. In addition, we show that the Machine Vision module, combined with the Physics-based modeling module, can act as closed-loop sensory feedback to the Machine Learning module of high- and low-fidelity data. Powered by our data-centric approach, we demonstrate that the online ML planner can actively learn the jet process dynamics using video and physics with minimum experimental cost. GPJet brings us one step closer to realizing the vision of intelligent AM machines that can efficiently search complex process-structure-property landscapes and create optimized material solutions for a wide range of applications at a fraction of the cost and speed.

Motivation & Objective

• To address the lack of automated, efficient calibration in electrohydrodynamic (E-jet) additive manufacturing, which currently relies on time-consuming trial-and-error methods.
• To develop an end-to-end framework that enables real-time, self-calibrating control of E-jet printing by integrating sensory feedback with physical and data-driven models.
• To reduce experimental cost and accelerate process optimization in multi-physics additive manufacturing by leveraging high-fidelity video data and physics-informed machine learning.
• To enable intelligent AM machines capable of autonomously navigating complex process-structure-property landscapes for optimized material fabrication.

Proposed method

• The Machine Vision module extracts high-fidelity jet features (e.g., jet shape, velocity) from real-time video using an automated, parallelized computer vision pipeline.
• The Physics-Based Modeling Module encodes the underlying electrohydrodynamic equations governing jet formation and dynamics, providing physical constraints for learning.
• The Machine Learning module employs Bayesian inference to fuse low- and high-fidelity data from vision and physics, enabling uncertainty-aware online learning.
• The framework operates in a closed-loop configuration, where real-time visual feedback and physical models jointly guide the ML planner to adapt process parameters dynamically.
• A data-centric approach is adopted, minimizing the need for extensive experimental trials by leveraging synthetic and real-time data with physical priors.
• GPJet is validated on a virtual E-jet printing setup with in-process jet monitoring, demonstrating robustness and efficiency in dynamic process learning.

Experimental results

Research questions

• RQ1 Can a physics-informed Bayesian learning framework effectively reduce the experimental cost of calibrating E-jet printing processes?
• RQ2 How well can real-time machine vision extract accurate jet dynamics features from video data in a high-speed, multi-physics manufacturing context?
• RQ3 To what extent can the integration of physical laws with visual and experimental data improve the accuracy and robustness of process modeling in E-jet printing?
• RQ4 Can the framework enable online, adaptive learning of jet dynamics with minimal human intervention?
• RQ5 How does the fusion of high- and low-fidelity data through Bayesian inference enhance predictive performance in complex, dynamic manufacturing systems?

Key findings

• The Machine Vision module successfully extracted high-fidelity jet features in real time from video data using a parallelized, automated workflow.
• The integration of vision data with physics-based models enabled accurate, uncertainty-aware prediction of jet dynamics with minimal experimental data.
• The GPJet framework demonstrated effective closed-loop sensory feedback, allowing the ML planner to actively learn and adapt to changing process conditions.
• The data-centric approach significantly reduced the number of required experiments by leveraging physical priors and real-time visual feedback.
• The framework achieved robust and efficient learning of complex E-jet printing dynamics, positioning it as a scalable solution for intelligent, self-calibrating AM systems.
• The results indicate that GPJet can navigate complex process-structure-property landscapes with high efficiency, enabling optimized material fabrication at reduced cost and time.