A Spatiotemporal Deep Learning-Powered Electrohydrodynamic Framework for Fluid Viscosity Estimation

Authors:

Abstract
Fluid characterization is of paramount importance for understanding the behavior of materials in several fields, from biomedical to industrial applications. Fluidics properties, such as viscosity, provide information on fluid stability, flow behavior, and potential biomarkers of physiological conditions. In this context, this study proposes an artificial intelligence-based system to automatically retrieve rheological properties, in particular fluid viscosity, by analyzing the fluid spatio-temporal dynamics that occur during the electrohydrodynamics (EHD) process. To this end, a spatio-temporal (ST) deep learning network, based on a custom three-dimensional Convolutional Neural Network (3D-CNN), named ST-EHDeepNET, is developed to analyze sequences of frames capturing fluid droplet deformation caused by the EHD phenomena. ST-EHDeepNET simultaneously extracts spatial and temporal features from video recordings, enabling automated fluid viscosity estimation. Experimental results demonstrate the high predictive accuracy of the model (RMSE 0.1±0.02), encouraging its application in challenging industrial and clinical scenarios.
Funding Information
  • Ministero dell'Universit? e della Ricerca (EXEGETE-C53D23003650001, TIPER-C53D23001660006)

This publication has 29 references indexed in Scilit: