The complexity of chemical reaction processes and the short market window of some chemical products mean that detailed model building can often not be justified. With little knowledge of chemistry, this work aims to provide a new methodology for model building of chemical reaction systems with minimum experimental measurements, for the purpose of reactor design and optimisation. Most often reactor designs are scaled from experimental measurements, especially for the manufacture of fine and speciality chemicals. Yet, without a model of the reaction system, major opportunities can be missed in the design and optimisation of the reactor. When models are developed for a reaction system in the laboratory, they are often inappropriate for reactor design and optimisation. In the first part of this thesis, the reaction scheme that best describes the production of a given chemical and suitable kinetic equations are obtained simultaneously using optimisation. A hybrid optimisation method is used to deal with this large problem where more than one model fits the same experimental data within a certain confidence level. Stochastic optimisation methods provide multiple solutions that are rival models for model discrimination. An NLP method improves model precision from the stochastic optimisation in the narrowed search space. A strategy for reaction scheme construction is used to generate all reactions from the reacting species and to provide plausible reaction schemes during optimisation. These reaction schemes are screened simultaneously with kinetic models to fit the most appropriate reaction scheme and kinetic model from the rival models. Optimal experiments then need to be designed to discriminate among rival models. The experimental design exploits the potential for mixing, as well as temperature and concentration effects to discriminate between models through the reactor superstructure. The oleic acid epoxidation reaction is used to demonstrate the methodology. For refinery heterogeneous catalytic reactions, due to the complex nature of catalysis, a large number of rival models pose difficulties for model building and discrimination. In the rest of the thesis, three-level kinetic study method is developed for model building to reduce the model complexity by separating diffusion effects from kinetic equations. In addition, catalyst characterisation is used to assist model discrimination. There are a large number of techniques available to connect catalyst properties, catalyst activities with model performance with different capabilities and limitations. However, not all of these will be useful in a given application. A classification of those techniques specified for hydrodesulphurisation (HDS) processes provides guidance for selecting suitable techniques to yield the most information with accuracy, speed, and economy. Furthermore, plausible ways for model discrimination and model improvement for thiophene and diesel HDS are explored, including operating condition, feedstock and catalyst effects.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:503069 |
| Date | January 2004 |
| Creators | Zhang, Wenling |
| Publisher | University of Manchester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0029 seconds