Work

Unsupervised learning bioreactor regimes

Year: 2024

Type: article

Abstract: Efficient bioreactor operation is essential for biomanufacturing success. Traditional Computational Fluid Dynamics (CFD) simulations are detailed but slow and complex, limiting their use in real-time applications. This study introduces a novel unsupervised learning algorithm that clusters bioreactors into coherent regions using CFD-generated data, with the potential to incorporate real-world data. This clustering aids in identifying reactor regimes or forming the basis for compartment models. Our custom k-means algorithm ensures spatial continuity within clusters and optimizes the number of compartments based on clustering scores, maintaining clear definitions while preserving essential data. The approach’s effectiveness is demonstrated through Pareto front analysis. Validation comes from case studies on a 202 m³ Rushton impeller bioreactor and an 840 m³ airlift reactor, emphasizing the benefits of 3-D compartmentalization in capturing fluid dynamics and cellular activities. This method improves bioreactor design and scaling, boosting industrial efficiency.

Source: Computers & Chemical Engineering

DOI: 10.1016/j.compchemeng.2024.108891

Authors: Víctor Puig I Laborda, Lars Puiman, Teddy Groves, Cees Haringa, Lars Keld Nielsen

Institutions: Novo Nordisk Foundation, Technical University of Denmark

Language: English

Topic: Fault Detection and Control Systems

Subfield: Control and Systems Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: hybrid

Estimated APC paid: $4,230