Masters Theses

Date of Award

12-2007

Degree Type

Thesis

Degree Name

Master of Science

Major

Biosystems Engineering

Major Professor

X. Philip Ye

Committee Members

Douglas G. Hayes, Alvin R. Womac, Arnold M. Saxton

Abstract

A new method for rapid chemical analysis of lignocellulosic biomass was developed using Fourier transform near-infrared (FT-NIR) spectroscopic techniques. The new method is less time-consuming and expensive than traditional wet chemistry. A mathematical model correlated FT-NIR spectra with concentrations determined by wet chemistry. Chemical compositions of corn stover and switchgrass were evaluated in terms of glucose, xylose, galactose, arabinose, mannose, lignin, and ash. Model development evaluated multivariate regressions, spectral transform algorithms, and spectral pretreatments and selected partial least squares regression, log(1/R), and extended multiplicative signal correction, respectively. Chemical composition results indicated greater variability in corn stover than switchgrass, especially among botanic parts. Also, glucose percentage was higher in internodes (>40%) than nodes or leaves (~30- 40%). Leaves had the highest percentage of lignin (~23-25%) and ash (~4-9%). Husk had the highest total sugar percentage (~77%). Individual FT-NIR predictive models were developed with good accuracy for corn stover and switchgrass. Root mean square errors for prediction (RMSEPs) from crossvalidation for glucose, xylose, galactose, arabinose, mannose, lignin and ash were 0.633, 0.620, 0.235, 0.374, 0.203, 0.458 and 0.266 (%w/w), respectively for switchgrass, and 1.407, 1.346, 0.201, 0.341, 0.321, 1.087 and 0.700 (%w/w), respectively for corn stover. A unique general model for corn stover and switchgrass was developed and validated for general biomass using a combination of independent samples of corn stover, switchgrass and wheat straw. RMSEPs of this general model using cross-validation were 1.153, 1.208, 0.425, 0.578, 0.282, 1.347 and 0.530 %w/w for glucose, xylose, galactose, arabinose, mannose, lignin and ash, respectively. RMSEPs for independent validation were less than those obtained by cross-validation. Prediction of major constituents satisfied standardized quality control criteria established by the American Association of Cereal Chemists. Also, FT-NIR analysis predicted higher heating value (HHV) with a RMSEP of 53.231 J/g and correlation of 0.971. An application of the developed method is the rapid analysis of the chemical composition of biomass feedstocks to enable improved targeting of plant botanic components to conversion processes including, but not limited to, fermentation and gasification.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS