Estimating dense-packed zone height in liquid–liquid separation: A physics-informed neural network approach
Abstract
• Phase heights in liquid–liquid separation are estimated from flow-rate data only. • A PINN is pretrained using synthetic data from a low-fidelity mechanistic model. • The pretrained PINN is then fine-tuned with scarce experimental settler data. • The trained PINN is embedded in a state estimation framework. • The two-stage trained PINN outperforms data-driven models in phase-height estimation.
Fields of Study & Subjects
Fluid Dynamics and Mixing
Phase Equilibria and Thermodynamics
Block Copolymer Self-Assembly
Publication details
- Journal / source
- Chemical Engineering Journal
- Publisher
- Elsevier BV
- Volume
- 541
- Pages
- 177517–177517
- Publication date
- 22 May 2026
- Language
- EN