Circular chemistry through network science and optimisation on big data

One of the largest challenges in the 21st century is the transition towards sustainable practices. In chemical engineering, the choice of feedstock, i.e. fossil or renewable, greatly influences the sustainability of chemical processing routes. At present, 90% of feedstocks in the chemical industry are non-renewable, thus, large-scale supply chain changes are urgently required. To enable this transition, it is of utmost importance that novel, yet competitive, processes based on renewable feedstocks are identified. Systematic early-stage sustainability assessment can cover large regions of chemical space and provide well-reasoned rankings of most promising reaction pathways. In this thesis, the hypothesis that networks are essential to support sustainability assessment of reaction pathways from big data is posed and answered. This thesis identifies three main areas for development: data, metrics, and decision-making, and investigates the use of networks within the areas. Networks provide an interlinked framework for reaction information (data), are key to assess flows of mass and energy (metrics), and form the basis of optimisation algorithms (decision-making). This work represents the chemical space by reaction networks assembled on large-scale data from Reaxys database. A methodology to identify the key molecules within the chemical supply chain, e.g. strategic molecules, is presented. Molecules are described by features based on their position within the network and an isolation forest outlier detection algorithm is employed to identify the key molecules. To assess pathways within network structures, chemical heuristics with following network optimisation are presented. This work introduces Petri net optimisation for reaction networks and compares the event-discrete modelling approach with the steady-state formalism used in reaction network flux analysis. This work explores a case study of reaction pathway identification from β-pinene to citral within chemical big data. Pathways are modelled in circular interaction with the supply network based on material availabilities and demands and an exergetic description of each reaction pathway is presented. The methodological pipeline automates early-stage sustainability assessment for large data sets. Last but not least, this thesis introduces a teaching approach to familiarize non-experts with network science and the complexity of sustainability problems.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2022

Author dc:creator

• Weber, Jana

Advisor dc:contributor.advisor

• Lapkin, Alexei

• sustainability education
• chemical reaction networks
• optimisation
• early-stage sustainability assessment

Licence

• https://www.rioxx.net/licenses/all-rights-reserved/

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.86775

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/339365