Back to search

University of Leeds

Model predictive control of industrial crystallisation processes

Abstract

dc:description.abstract

For fine chemicals produced by batch cooling crystallisation, control of the product crystal size distribution (CSD) is necessary for good performance properties and downstream processing. To improve upon the resource-heavy process of developing an empirical model-based control system, this thesis presents a model predictive control (MPC) methodology for the control of supersaturation and product CSD using commercially available tools. A one-dimensional (1D) population balance model (PBM) was parameterised in gPROMS FormulatedProducts software (Siemens-PSE Ltd.) for the hexamine-ethanol crystallisation system, using kinetics reported by Myerson et al. (1986). Seeded batch crystallisation experiments were run to validate the model, and scope out suitable process conditions for control performance tests. The gPROMS model was coupled with the control software PharmaMV (Perceptive Engineering Ltd.) to create a digital twin of a 500 ml crystalliser. Simulated crystallisation data was used to train statistical MPC structures for supersaturation control, in addition to a batch end-point control (EPC) strategy for the D50 of the product CSD. Within a relative supersaturation set-point range of 0.012-0.036, the digital twin performed with excellent stability. When the MPC strategy was implemented on a real 500 ml crystallisation setup, supersaturation stability was slightly poorer (standard deviations of order 10^-3 instead of 10^-4). A similar level of stability was maintained when perturbing the growth rate kinetics by manipulating the seed CSD and adding a growth rate inhibitor. Performance was mostly linked to whether cooling equipment could meet the demand to accurately maintain the required cooling rate. Simulations of the crystal size EPC strategy achieved different target D50 values within a range of 20 μm. However, this was not reflected in physical experiments where secondary nucleation was negligible, implying that particle generation is mandatory for CSD control. Adaptations to improve the industrial applicability of this control methodology have been proposed in the conclusions.

Degree

thesis:*
Name dc:type.qualificationname
Integrated Ph.D and Master
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Leeds
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Leeming, Ryan Paul
Advisors dc:contributor.advisor
  • Mahmud, Tariq
  • Roberts, Kevin
  • Simone, Elena
  • George, Neil
  • Webb, Jennifer
  • Brown, Cameron
ORCID person:identifier.orcid

Rights

Language dc:language.dcterms_iso639_2
eng

Identifiers

dc:identifier.*
Dc Identifier Grantnumber
EP/L015285/1
HH13054

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Leeming, Ryan Paul. Model predictive control of industrial crystallisation processes. doctoral thesis, University of Leeds, 2024.