Fig. 2: The GPJet Pipeline Framework.
From: Physics-Informed Bayesian learning of electrohydrodynamic polymer jet printing dynamics
The Physics-informed Bayesian Machine Learning framework (GPJet) comprised by three different modules: a the Machine Vision module, which takes as an input timeseries video focusing on the polymer jet in the free flow regime and performs the extraction of high-fidelity jet features in real-time based on an automated image processing algorithmic workflow (denoted as “Jet Metrology” in Section Machine Vision Module) – the extracted jet features are denoted on jet profile images in grayscale (0–255) with the 0 value and the 255 value in the color bar representing the black background, and white segmented jet profile respectively, b the Machine Learning module and c the Physics-based Modeling Module, d the Multi-fidelity Modeling Module which takes as input high fidelity experimental data from the Machine Vision module and low fidelity modeling data from the Physics-based Modeling Module and performs a series of data-driven tasks to learn the jet dynamics. Filled contours (shading) represent uncertainty bounds (95% confidence intervals (CIs)) of the predictions.
