Paper Number
AM11 

Session
AI and ML Based Rheological Characterization

Title
Machine-learning-based measurement of the relaxation time of viscoelastic fluids via particle ordering

Presentation Date and Time
October 10, 2022 (Monday) 2:50

Track / Room
Track 6 / Mayfair

Authors (Click on name to view author profile)

1. De Micco, Maurizio (University of Naples Federico II, DICMaPI)
2. D'Avino, Gaetano (University of Naples Italy)
3. Villone, Massimiliano M. (University of Naples Federico II)

Author and Affiliation Lines (in printed abstract book)
Maurizio De Micco, Gaetano D'Avino and Massimiliano M. Villone
DICMaPI, University of Naples Federico II, Naples, NA 80125, Italy

Speaker / Presenter 
De Micco, Maurizio

Keywords
computational methods; AI based; ML based; suspensions

Text of Abstract
The rheological characterization of complex fluids is of great importance in several applications. Among the rheological properties, relaxation time has a special relevance, as it provides a measure of fluid elasticity. Depending on the liquid, the relaxation time can vary over several orders of magnitude, from seconds to milliseconds. The latter is the case of aqueous solutions at low polymer concentration and biological fluids, where conventional rheometers fail to give accurate measurements and alternative techniques need to be developed. In this work, we present a non-intrusive method to measure the relaxation time of viscoelastic fluids in a microchannel by exploiting the phenomenon of “particle ordering” combined with a machine learning approach. Particles suspended in a viscoelastic medium flowing through a channel tend to migrate at the channel centreline and self-assemble in strings due to particle-particle hydrodynamic interactions. The ordering mechanism depends on the fluid relaxation time. Hence, we propose to measure this quantity from the distribution of the interparticle distances by using machine learning techniques. To prove the effectiveness of our approach, we generate “in silico” a dataset of interparticle distances by following the simulation procedure reported by D’Avino and Maffettone [1]. The dataset is used as input for various machine learning algorithms, as support vector machines and random forests. Both classification and regression are carried out with high performances. In principle, the proposed approach can be used to measure other properties of fluids. [1] D’Avino G., Maffettone P.L. (2019). Numerical simulations on the dynamics of trains of particles in a viscoelastic fluid flowing in a microchannel. Meccanica 55, 317–330.