Agunloye, Emmanuel;
Yusuf, Muhammad;
Chamberlain, Thomas;
Muller, Frans;
Bourne, Richard;
Galvanin, Federico;
(2023)
Cloud-based MBDoE applications for optimal design of experiments to accelerate kinetic model identification.
Presented at: Sargeant Centre Industrial Consortium Meeting, London, UK.
Preview |
Slideshow
Sargeat Centre Industrial Consortium 0112 .pdf - Published Version Download (1MB) | Preview |
Abstract
Cloud technologies offer the enabling platform for remote computing and interconnection of machines globally, thus increasing and accelerating industrial production. This work introduces a novel cloud-based platform driven by online model-based design of experiments (MBDoE) techniques to remotely optimise experimentation in a smart flow reactor situated at the University of Leeds. Using the dynamics of a physical system, the MBDoE algorithm can optimally design experiments for model discrimination and improve parameter precision to identify a reliable mathematical model for kinetics describing reaction processes occurring in the physical system. In MBDoE for parameter precision, the purpose is to design experiments to reduce uncertainties in model parameters for reliable model predictions. Uncertain parameters are used in the algorithm as prior estimates to search for the posterior experiment to conduct for the highest reduction in model parameter uncertainties. In this work, we have demonstrated the applicability of the cloud-based platform in a pharmaceutically relevant case study, homogeneous amide formation, where the synthesis can be described using reversible chemical steps. Prior uncertain estimates for kinetic parameters in the reversible kinetics were calculated from preliminary experiments designed using Latin hypercube sampling. Using these prior estimates, MBDoE for parameter precision designed a single additional experiment, which updated preliminary experiments to yield a statistically precise parameter estimation of parameters after just one additional run. The kinetic model with the precisely estimated parameters produced reliable predictions when tested against an unseen validation data set from experiments designed using a full factorial approach.
| Type: | Conference item (Presentation) |
|---|---|
| Title: | Cloud-based MBDoE applications for optimal design of experiments to accelerate kinetic model identification |
| Event: | Sargeant Centre Industrial Consortium Meeting |
| Location: | London, UK |
| Dates: | 08 December 2023 |
| Open access status: | An open access version is available from UCL Discovery |
| Publisher version: | https://www.imperial.ac.uk/events/170171/sargent-c... |
| Language: | English |
| Keywords: | cloud technologies, online model-based design of experiments, MBDoE for parameter precision, reversible kinetics |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Chemical Engineering |
| URI: | https://discovery-pp.ucl.ac.uk/id/eprint/10194488 |
Loading...
Loading...Loading...Loading...Archive Staff Only
 |
View Item |
