Compositional analysis and classification of Miscanthus using Fourier transform near infrared spectroscopy

Williams, Daniel

Permalink

https://hdl.handle.net/2142/46756 Copy

Description

Title

Compositional analysis and classification of Miscanthus using Fourier transform near infrared spectroscopy

Author(s)

Williams, Daniel

Issue Date

2014-01-16T18:01:23Z

Director of Research (if dissertation) or Advisor (if thesis)

Danao, Mary-Grace C.

Department of Study

Engineering Administration

Discipline

Agricultural & Biological Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Miscanthus
• Biomass
• Fourier transform near infrared (FT-NIR) spectroscopy
• partial least squares regression
• linear discriminant analysis

Abstract

Miscanthus × giganteus is a woody rhizomatous C4 grass species that is a high yielding lignocellulosic material for energy and fiber production. The cellulose and hemicellulose fractions of Miscanthus can be converted into energy and chemicals through biological conversion. Since only a fraction of the biomass can be converted into chemical energy, bioethanol yields per unit mass of biomass are directly proportional to the composition of the biomass, which can vary due to age, stage of growth, growth conditions, and other factors. It is advantageous to know these variations prior to conversion so that enzyme mixtures, yeast strains, and process control parameters can be adjusted accordingly to maximize yields. Knowing the composition at earlier stages of the supply chain can also help in the development of quality-based valuations which incentivize farmers and suppliers to implement best management practices to ensure a uniform and consistent supply system. Therefore, in this study, the variability of composition of Miscanthus bales stored under a variety of conditions for a period of 3 to 24 months was described, along with the compositional variability of its botanical fractions. High throughput assays based on Fourier transform near infrared (FT-NIR) spectroscopy, partial least squares regression (PLSR), and linear discriminant analyses (LDA) to provide quantitative and qualitative measures of Miscanthus composition were developed. Results showed large variations (mean ± S.D.) in glucan (40.4 ± 2.70%), xylan (20.7 ± 1.50%), arabinan (1.90 ± 0.40%), acetyl (2.84 ± 0.28%), lignin (20.5 ± 1.40%), ash (2.60 ± 1.80%), and extractives (5.60 ± 0.86%) - contents were observed for samples that were collected from Miscanthus bales stored indoors, under roof, outdoors with tarp cover, and outdoors without tarp cover for i 3 to 24 months after harvest and baling. There was also a wide variability for all components: glucan, 32.2 to 46.1%; xylan, 20.9 to 25.3%; arabinan, 0.0 to 6.1%; lignin, 18.7 to 25.5%; and ash, 0.4 to 8.9%, observed in botanical fractions of Miscanthus. The ranges in composition were comparable to corn stover botanical fractions. While the sum of glucan, xylan, and arabinan contents for the rind, pith, and sheath fractions were not different from each other, the variations across some botanical fractions were significant with the blade having lowest glucan, lowest lignin, and highest ash contents. PLSR models were developed to predict glucan, xylan, lignin, and ash contents in Miscanthus bale samples with RPD values of 4.86, 4.08, 3.74, and 1.71, respectively. The geometric mean particle size ranged from 0.36 to 0.49 mm, with the smallest size observed with samples from bales stored outdoors for 17 months and the largest size observed with samples from bales stored outdoors with a tarp cover for 5 months. On average, PLSR predictions of glucan, arabinan, and lignin content were not sensitive to the particle size of ground Miscanthus, but predictions of xylan and ash content were. The predicted xylan content using the non-sieved samples was lower than those for sieved samples and ash levels increased with decreasing particle size. When the PLSR models were coupled with LDA to classify the Miscanthus samples based on their glucan, lignin, and ash contents, the best classification results were found with the PLS-DA lignin model. While the PLSR and PLS-DA models developed in this study were based on a small sample size, the approaches presented in this study demonstrated FT-NIR spectroscopy is a practical tool for screening biomass at different stages of the supply chain, making the delivery of consistent feedstock to conversion facilities year round a realistic possibility.

Graduation Semester

2013-12

Permalink

http://hdl.handle.net/2142/46756

Copyright and License Information

Copyright 2013 Daniel Williams

Owning Collections