The acid catalysed esterification of ethanol by acetic acid, a batch process, has been investigated on a laboratory scale at the high temperature range of 78 - 80°C. The data has been collected by Raman Spectroscopy and successfully de-noised using Principal Components Analysis. The first principal component (PCI) was found to describe the fluorescence and other sources of noise in the data and the reconstituted data due to the variation captured in the second principal component (PC2) contained the actual Raman spectra. Thus the reaction profile as well as the profiles of individual reaction components have been clearly mapped out. Validation of this denoising technique has been done by calculating the kinetics of the reaction with the reconstituted data, which has been found to follow the theoretical first order reaction kinetics. The effect of variable selection procedures on model building has been investigated using data from a continuous industrial process, for which reaction profiling as was done for the batch system is not applicable. Two variable selection techniques, General Randomised PRESS-based Elimination (GRAPE) and the genetic algorithm (GA), improve the prediction ability of MLR models by a great deal, indicated by Root Mean Square Error of Cross-Validation (RMSECV) values of 1.0649 - 1.1277 and 1.0977 - 2.0064 respectively. Predicted concentrations are a good estimate for the actual concentrations.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:445281 |
Date | January 2006 |
Creators | Ampiah-Bonney, Richmond Jerry |
Contributors | Walmsley, Anthony |
Publisher | University of Hull |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://hydra.hull.ac.uk/resources/hull:5861 |
Page generated in 0.002 seconds