Machine Learning-Assisted Discovery of Novel Reactor Designs via CFD-Coupled Multi-fidelity Bayesian Optimisation - Abstract

Aug 5, 2026

Upload by Anda

Additive manufacturing enables producing advanced reactor geometries and greatly expands reactor design spaces, making it difficult for human-centric approaches to identify and optimise promising configurations. The work introduces two novel coiled-tube parameterisations that create high-dimensional optimisation problems without requiring gradients. Multi-fidelity Bayesian optimisation, coupled with parameterised meshing and simulation, exploits faster lower-fidelity runs during search. Maximising plug-flow performance enables extrapolation to two 3D-printed designs validated experimentally.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Abstract

Introduction

Coiled tube reactors and flow physics

Additive manufacturing and data-driven optimisation

Aim and proposed approach

arXiv :2308 .08841v1 [ cs .CE] 17 Aug 2023 Machine Learning-Assisted Discovery of Novel Reactor Designs via CFD-Coupled Multi-fidelity Bayesian Optimisation Tom Savage 1,2 , Nausheen Basha2 , Jonathan McDonough3 , Omar Matar2 , and Ehecatl Antonio del Rio Chanona 1,2 1Department of Chemical Engineering, Imperial College London 2 Sargent Centre for Process Systems Engineering, Imperial College London 3 School of Engineering, Newcastle University Abstract Additive manufacturing has enabled the production of more advanced reactor geometries, resulting in the potential for significantly larger and more complex design spaces. Identifying and optimising promising configurations within broader design spaces presentsa significant challenge for existing human-centric design approaches. As such, existing parameterisations of coiled-tube reactor geometries are low-dimensional with expensive optimisation limiting more complex solutions. Given algorithmic improvements and the onset of additive manufacturing, we propose two novel coiled-tube parameterisations enabling the variation of cross-section and coil path, resulting in a series of high dimensional, complex optimisation problems. To ensure tractable, non-local optimisation where gradients are not available, we apply multi-fidelity Bayesian optimisation. Our approach characterises multiple continuous fidelities and is coupled with parameterised meshing and simulation, enabling lower quality, but faster simulations to be exploited throughout optimisation. Through maximising the plug-flow performance, we identify key characteristics of optimal reactor designs, and extrapolate these to produce two novel geometries that we 3D print and experimentally validate. By demonstrating the design, optimisation, and manufacture of highly parameterised reactors, we seek to establish a framework for the next-generation of reactors, demonstrating that intelligent design coupled with new manufacturing processes can significantly improve the performance and sustainability of future chemical processes. Introduction Coiled tube reactors have received attention across chemical engineering due to their desirable mixing and heat transfer characteristics [1–6] . Their applications span from flow chemistry [7], and bioprocesses [8], to chemical kinetic experiments [4] . At mesoscale, coiled tube reactors have been shown to combine the heat and mass transfer properties of microreactors with the economic benefits of higher throughput larger scale reactors [9] . When fluid flows through curved tubes at Reynolds numbers (Re) exceeding 300, it experiences a centrifugal force, leading to the development of strong secondary flow structures in the form of counter-rotating Dean vortices [10, 11] . These vortices significantly improve radial mixing, leading to enhanced plug flow performance. Previous work has revealed improvements in the performance of coiled tube reactors at low Reynolds numbers (Re ≤ 50) by super-imposing pulsed-flow operating conditions, which also induce Dean vortices [2, 12, 13] . However, it remains to be seen whether these flows can be induced ata low Reynolds numbers without the operational overhead of pulsed-flow operating conditions. In order to investigate the potential impact of novel coiled-tube reactors under steadyflow conditions, it becomes pertinent to investigate new reactor parameterisations and associated optimal designs. Furthermore, advancements in additive manufacturing have enabled the creation of complex and unintuitive reactor designs. Previously infeasible designs can now be manufactured and investigated, resulting in substantially larger design spaces. Coupled with data-driven design tools such as multi-fidelity Bayesian optimisation [14–20], which has enabled the optimisation of large-scale simulation based problems, a pathway has emerged towards the identification of optimal reactors from larger design spaces with the potential for improved performanc

FAQs

Why are traditional coiled-tube reactor parameterisations difficult to use for complex design spaces?

Existing coiled-tube parameterisations are low-dimensional, and optimising more complex solutions is limited by the expense of optimisation. Larger design spaces become challenging for human-centric approaches to search effectively.

What does the paper propose to enable optimisation of novel coiled-tube designs?

It proposes two new coiled-tube parameterisations: varying cross-section along the reactor length and varying coil path, yielding high-dimensional optimisation problems solved under steady-flow conditions.

How does multi-fidelity Bayesian optimisation help in the optimisation workflow?

The method characterises multiple continuous fidelities and couples with parameterised meshing and simulation, allowing faster lower-quality simulations to be used throughout optimisation when gradients are unavailable.

Aug 5, 2026

Upload by Anda