Surrogate and Multiscale Modelling for (Bio)reactor Scale-up and Visualisation

Abstract

Bioresource production in bioreactors presents a sustainable biotechnology for tackling the ever-increasing energy and mass demands of the world’s surging population. To attain commercial viability, reaction engineers must efficiently design and upscale these bioreactors for the industrial production of high value biochemicals, fuels, and materials. These engineers utilise computational fluid dynamics (CFD) to visualise bioreactor fluid flow and optimise dead zones with poor mixing, leading to promising bioreactor configurations. An advanced route, yet to be widely deployed, is the integration of bioreaction kinetics within the CFD framework for multiscale optimisation and upscaling. To demonstrate its potential, a two-step coupling strategy of CFD hydrodynamics to light transmission and bioreaction transport was comprehensively demonstrated herein for photobioreactors (PBRs) of different configurations and scales. The problem of prohibitively high computational cost of simulating long lasting fermentation experiments was addressed with a recently published accelerated growth kinetics strategy. To further cut the simulation cost stemming from the computationally expensive objective evaluation during multiscale CFD optimisation, a Gaussian process model was trained as a surrogate of the expensive multiscale CFD model and utilised within a Bayesian optimisation (BO) framework. BO suggested a near-optimal static mixer configuration for a flat plate PBR yielding over a 95.3% increase in biomass concentration compared to the baseline without static mixers. This robust and sample efficient optimisation strategy provides enormous cost savings and presents a step forward towards the efficient design, optimisation, and upscaling of bioreactors.