Development and analysis of machine-learning guided flash nanoprecipitation (FNP) for continuous chitosan nanoparticles production

Chitosan-based nanoparticles (CNPs) are widely used in drug delivery, cosmetics formulation and food applications. To accelerate the manufacturing of CNPs, the present study develops a workflow to prepare CNPs in a continuous model. Based on machine learning, the workflow precisely predicts size and polymer dispersity index (PDI) value of CNPs, which impacts on the colloidal stability and applications. Multi-inlet vortex mixer (MIVM) device was fabricated by 3D printing as the reactor. Peristaltic pump was applied to deliver the reaction streams into the MIVM device and produce CNPs by flash nanoprecipitation (FNP) in a continuous way. The developed MIVM device produces CNPs in a controlled manner at a higher output which is promising for upscale applications. Twelve machine learning algorithms were employed to investigate the potential relationship between the reaction independent variables and hydrodynamic characteristics of CNPs. Random Forest, Decision Tree, Extra Tree and Bagging algorithms performed better than other algorithms with the average prediction accuracy around 90 %. The current study demonstrated that supervised machine learning guided FNP using the developed MIVM device is an effective strategy for accurate and intelligent production of CNPs and other similar nanoparticles.