Work

Data-driven inverse design for magnesium hydroxide precipitation: Coupling CFD–PBM with deep learning to control hexagonal platelets morphology formation and particle size distribution

Year: 2026

Type: article

Abstract: Magnesium hydroxide precipitation is a key industrial process whose outcome strongly depends on operating conditions, hydrodynamics, and supersaturation dynamics, ultimately affecting particle morphology and particle size distribution (PSD). In this work, a mechanistic and data-driven inverse design framework is proposed by coupling Computational Fluid Dynamics and Population Balance Modelling (CFD-PBM) with deep learning techniques to identify precipitation kinetics and predict particle characteristics under different semi-batch feeding strategies. A neural-network-based surrogate model is employed to infer kinetic parameters governing nucleation, growth, and agglomeration directly from experimentally measured characteristic particle sizes, drastically reducing the computational burden associated with traditional optimization loops. The identified kinetic parameters are shown to be transferable across operating conditions and reactor configurations, enabling accurate predictions without re-calibration. The proposed CFD-PBM model successfully captures the effect of supersaturation generation rate on particle morphology, distinguishing between globular primary particles formed under rapid supersaturation buildup and hexagonal platelet-like structures obtained under slower supersaturation conditions. Shape effects are accounted for through morphology-dependent shape factors, demonstrating that particle shape primarily influences the quantitative magnitude of characteristic sizes while preserving qualitative trends. Model predictions are in good agreement with experimental data within the observed variability. Overall, this study provides a robust and computationally efficient framework for the mechanistic interpretation of magnesium hydroxide precipitation, offering new opportunities for morphology and PSD control through the integration of CFD-PBM modelling and advanced machine learning techniques.

Source: Chemical Engineering Journal

DOI: 10.1016/j.cej.2026.176546

Authors: Antonello Raponi, Daniele Marchisio

Institutions: Centre National de la Recherche Scientifique, Mines Saint-Étienne, Georges Friedel Laboratory, Politecnico di Torino

Language: English

Topic: Crystallization and Solubility Studies

Subfield: Materials Chemistry

Field: Materials Science

Domain: Physical Sciences

Open Access status: hybrid

Estimated APC paid: $4,830

Funder: Centre National de la Recherche Scientifique