Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.489843
Title: Oil fingerprinting : the development of OPDS, a novel chemometric analysis and diagnostic pattern recognition tool
Author: Buxton, Robert
Awarding Body: University of Staffordshire
Current Institution: Staffordshire University
Date of Award: 2006
Availability of Full Text:
Access through EThOS:
Abstract:
The Oil Pollution Diagnostic System (OPDS) is a proof of concept prototype pattern recognition system designed to assist chemists in assessing the type and source of refined oil spills that occur in freshwater. The pollution of inland waters by refined oils can cause damage to the local environment that takes many months to repair since oils are toxic to many of the flora and fauna often found in these habitats. The issue of assessing the exact type and source of the spill based on chemical analysis are separate but closely linked problems. The lesser of the two problems relates to the reliable identification of the oil type based upon distinctive patterns contained in the analysis data. The problem of identifying the likely source is more challenging however and the work describes an approach where the patterns contained in small sections of the analysis data are used as the basis of a match. Manual methods of analysing chemical data to determine the type and source of a spill have existed for some time. However in a courtroom situation where the analysis is presented as evidence in the attempt to secure a prosecution, manual methods of matching could be seen as subjective and unreliaole and possibly be exploited by the defence. The OPDS was developed as an alternative to the manual methods already in use. The system illustrates the advantages of transformable oil type templates, the advantages of automatic calibration to reduce the impact of retention time drift and the use of the lesser peaks between the major peaks for oil source fingerprinting. The major contribution in this section of the development project was to incorporate all the features mentioned in a system dedicated to the matching of gas cl1romatograms (GC), and furthermore to incorporate the facility to allow the system to automatically match GCs with minimal input from the user. The template based approach for type matching required the construction of a dynamic type template for each oil type to be identified. The template modelled a collection of distinctive features identified from the weathered information provided by the Environment Agency (of England and Wales). The templates were dynamic since they model the way the distinctive features are altered by the effects of weathering. This ability to negate the effects of weathering from the pattern together with some calibration allowing for instrument variation allowed the comparison of a sample to a set of templates via a metric based on the Euclidean distance between matched peaks. A similar approach was adopted with the problem of source identification where the emphasis was more on matching small collections of GC peaks that are present between the major features. This approach was based on a suggestion contained in the feasibility study. The EA provided a number of GCs relating to actual pollution incidents and these formed the basis of a set of case studies which were then used to develop and test the source matching system. Tests carried out on the type matching component of the system using independent test data produced correct classification figures of greater than 90% and kappa values of greater than 0.9. The source matching testing was less extensive than that done on the type matching component however a correct classification figure of 88% was observed. This thesis contains an explanation of the problems caused by spills of refined m inland waters, an explanation of the teclmiques used to analyse oil spills, a review of pattern recognition and chemometric tecrwiques, at1 overview of the software design and development the results achieved and the conclusions that can be drawn from these.
Supervisor: Not available Sponsor: Not available
Qualification Name: University of Staffordshire, 2006 Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.489843  DOI: Not available
Share: