Chandwani, Roopkumar Mohan; (1995) Matching of bioprocess data records for the rapid diagnosis of chromatographic processes. Doctoral thesis (Ph.D), UCL (University College London).

Text out.pdf Download (12MB)

Abstract

This thesis considers the application of pattern recognition to the at-line diagnosis of chromatographic processes. The application of control is reviewed for high purity bioprocesses. It is concluded that advanced diagnostic techniques (e.g. artificial intelligence and multivariate data analysis) have successfully been applied to the fermentation of biomolecules. Extending these to downstream processing (e.g. chromatography) leads to a more efficient process. A three-stage data driven diagnostic tool, based upon principal component analysis (a technique of multivariate data analysis) has been developed and tested. The tool compares current process performance by matching with historical data. The first stage of the tool provides a qualitative match of the current process chromatogram with the entire historical data. The second stage provides a quantitative description indicating whether the process has deviated from the bounds of a template. The third stage provides information on batch-to-batch performance. The tool is used to diagnose process performance of a test system based upon size exclusion chromatography. In order to test the viability of the tool, three datasets and seven test chromatograms were generated using the test system. Two historical datasets, generated using an experimental design approach, were used to test the first two stages of the tool by classifying the test charges. A third dataset tested the third stage of the tool. Several weaknesses with the second stage of the tool were identified which concern the F-test to assess membership of test chromatograms to the template. Alternatives to the F-test are discussed. A description of the test system, a mimic of an industrial-scale chromatographic separation, is presented. The test system is characterised by applying techniques of experimental design to the discrimination factor, a chromatographic separation criteria. The analysis indicated multivariate relationships between the process variables and showed that parameters determined from chromatographic profiles alone could not adequately describe the performance of chromatography systems. Consequently, the whole UV profile of the chromatography eluent was used for analysis by the tool developed in this thesis. The variation of chromatographic peak parameters is quantified and their utility in monitoring by statistical process control is discussed.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item