Project 9

A hybrid Physics-BASED and Machine Learning framework for electrospinning process understanding

(dr. luis segura, iE)

Objective: Develop a framework to integrateelectro-hydrodynamics models and in-situ data to have accurate predictions (e.g., Taylor cone, jet, whipping, etc. behaviors) in electrospinning (see Fig. 1) and relate them to the quality of the products (e.g., fiber alignment, diameter distribution, etc.).  

Project description: Obtaining reliable/consistent electrospun fibers is crucial for the electrospinning process since this will improve the product performance (e.g., scaffolds, high-performance filters, etc.). The process variations are influenced by process parameters (e.g., voltage, solution feeding rate, needle diameter, etc.), material parameters (e.g., density, viscosity, etc.), and ambient conditions (e.g., humidity, temperature, etc.). Having mechanisms to understand and detect these variations are important for the process stability, reliability, and repeatability. Data-driven approaches have demonstrated important progress at improving manufacturing processes but they lack the physics principles that govern the manufacturing phenomenon (e.g., electro-hydrodynamics in electrospinning). On the other hand, physics-based models can be utilized to encode the physics, but they are computationally inefficient for real time applications. Hence, integrating these two becomes paramount to improve the model performance for anomalies detection in electrospinning.

Data to be collected: CFD models of the electrospinning, videos of the different electrospinning stages (e.g., taylor cone, jet, whipping, etc.), SEM images of the deposited nanofibers. These can be tested for different collectors to have a complete dataset for various analysis.