Khan, MR
1990
Modeling and Predicting the Composition of Fossil Fuel Derived Pyrolysis Liquids by Using Low-Voltage Mass Spectrometry and Canonical Correlation Analysis
Chakravarty, T.; Khan, M.R. and Meuzelaar, H.L.C.
Industrial and Engineering Research, 29 (11), 2173-2180, 1990. Funded by Consortium for Fossil Fuel Liquifaction Science.
Low-voltage electron ionization mass spectrometry (LV-EIMS) was performed on 25 fossil fuel samples (21 coals, 2 oil shades, 1 tar sand, and 1 coal resin concentrate) and their respective pyrolysis liquids prepared at Morgantown Energy Technology Center (METC) by means of a fixed-bed reactor. By using principal component analysis, the tar evaporation spectra and the solid fuel pyrolysis spectra were classified in terms of the underlying structural variables. In both data sets, all 4 non-coal samples, as well as 2 less typical coal samples, were found to be outliers. After removing the 6 outliers, canonical correlation analysis was performed on the remaining subsets of 19 coal samples in order to bring out the compositional similarities and differences between the fossil fuel samples and their pyrolysis liquids. By determining the common sources of variance between the two data sets by means of canonical correlation analyses, it was demonstrated that the canonical variate model enabled prediction of the mass spectrum of a given coal tar sample from the measured pyrolysis mass spectrum of the corresponding coal sample. Agreement with the experimental results was reasonably good.
1988
Prediction of the Composition of Coal Tars from the Pyrolysis Mass Spectra of the Parent Coals Using Canonical Correlation Techniques
Chakravarty, T.; Meuzelaar, H.L.C.; Jones, P.R. and Khan, M.R.
ACS Preprints, 33, (2), 235-241, 1988. Toronto, CA. Funded by ACERC (National Science Foundation and Associates and Affiliates).
Numerical comparison of compositional data on coals and their corresponding pyrolysis tars enables the construction of empirical mathematical models to predict liquid yield and composition from spectroscopic data of the parent coal. This approach was successful when using spectroscopic methods combined with vacuum micropyrolysis techniques, viz. Curie-point pyrolysis mass spectrometry. Nineteen US coals and the corresponding pyrolysis liquids prepared by the SHRODR method were analyzed by means of Curie-point pyrolysis low voltage MS. The pyrolysis mass spectra of the coals were composed of mainly primary pyrolysis products typical of vacuum micropyrolysis and were substantially different from the low voltage mass spectra of the corresponding SHRODR tars produced under batch autoclave conditions which promote the formation of secondary pyrolysis products. Nevertheless, it proved feasible to model and predict SHRODR tar spectra from the vacuum micropyrolysis spectra of the coals with a high degree of precision by means of factor analysis-based canonical correlation methods.
Khan, MR
1990
Modeling and Predicting the Composition of Fossil Fuel Derived Pyrolysis Liquids by Using Low-Voltage Mass Spectrometry and Canonical Correlation Analysis
Chakravarty, T.; Khan, M.R. and Meuzelaar, H.L.C.
Industrial and Engineering Research, 29 (11), 2173-2180, 1990. Funded by Consortium for Fossil Fuel Liquifaction Science.
Low-voltage electron ionization mass spectrometry (LV-EIMS) was performed on 25 fossil fuel samples (21 coals, 2 oil shades, 1 tar sand, and 1 coal resin concentrate) and their respective pyrolysis liquids prepared at Morgantown Energy Technology Center (METC) by means of a fixed-bed reactor. By using principal component analysis, the tar evaporation spectra and the solid fuel pyrolysis spectra were classified in terms of the underlying structural variables. In both data sets, all 4 non-coal samples, as well as 2 less typical coal samples, were found to be outliers. After removing the 6 outliers, canonical correlation analysis was performed on the remaining subsets of 19 coal samples in order to bring out the compositional similarities and differences between the fossil fuel samples and their pyrolysis liquids. By determining the common sources of variance between the two data sets by means of canonical correlation analyses, it was demonstrated that the canonical variate model enabled prediction of the mass spectrum of a given coal tar sample from the measured pyrolysis mass spectrum of the corresponding coal sample. Agreement with the experimental results was reasonably good.
1988
Prediction of the Composition of Coal Tars from the Pyrolysis Mass Spectra of the Parent Coals Using Canonical Correlation Techniques
Chakravarty, T.; Meuzelaar, H.L.C.; Jones, P.R. and Khan, M.R.
ACS Preprints, 33, (2), 235-241, 1988. Toronto, CA. Funded by ACERC (National Science Foundation and Associates and Affiliates).
Numerical comparison of compositional data on coals and their corresponding pyrolysis tars enables the construction of empirical mathematical models to predict liquid yield and composition from spectroscopic data of the parent coal. This approach was successful when using spectroscopic methods combined with vacuum micropyrolysis techniques, viz. Curie-point pyrolysis mass spectrometry. Nineteen US coals and the corresponding pyrolysis liquids prepared by the SHRODR method were analyzed by means of Curie-point pyrolysis low voltage MS. The pyrolysis mass spectra of the coals were composed of mainly primary pyrolysis products typical of vacuum micropyrolysis and were substantially different from the low voltage mass spectra of the corresponding SHRODR tars produced under batch autoclave conditions which promote the formation of secondary pyrolysis products. Nevertheless, it proved feasible to model and predict SHRODR tar spectra from the vacuum micropyrolysis spectra of the coals with a high degree of precision by means of factor analysis-based canonical correlation methods.