Data-driven optimisation of coiled tube reactors via computational fluid dynamics

Performance optimisation via data-driven methods is a promising approach to minimise computational effort. In this research we demonstrate the application of a data-driven optimisation technique coupled with CFD to enhance the plug flow performance in a coiled tube reactor subjected to oscillatory flows at highly laminar flow conditions (at least as low as Re=10). We apply Bayesian optimisation techniques based on Gaussian processes to build a surrogate model which is iteratively updated using OpenFOAM simulations evaluated through the PyFoam library. We run transient CFD simulations using the ScalarTransportFoam solver, where a tracer is injected into the water working fluid for a fixed coil geometry to produce a residence time distribution (RTD). We explore the parameter space of Reynolds number (10–50), oscillation amplitude (1–8 mm), and oscillation frequency (2–8 Hz) to find the narrowest RTD, corresponding to optimal plug flow under the laminar flow conditions. We also study the resulting vortex structures to further understand the optimal plug flow performance conditions to guide potential future reactor designs. We expect this low-cost integrated modelling approach to be easily applicable to a wide range of industrial mixers to identify opportunities for performance improvement.