Near Infrared Calibration Models for Pretreated Corn Stover Slurry Solids, Isolated and in situ. Issue 4 (August 2013)

• Record Type: Journal Article
• Authors: Sluiter, Amie; Wolfrum, Ed
• Is Part Of: Journal of near infrared spectroscopy. Volume 21:Issue 4(2013)
• Page Start: 249
• Page End: 257
• Publication Date: 2013-08
• Subjects: near infrared -- NIR (near infrared) -- biomass -- biofuels -- slurry -- corn stover -- pretreatment -- rapid analysis -- calibration model; Near infrared spectroscopy -- Periodicals; 543.5
• DOI: 10.1255/jnirs.1065
• Languages: English
• ISSNs: 0967-0335

Abstract

Biomass pretreatment processes often yield slurry, a two-phase material consisting of an aqueous phase with solubilised components and a solid phase with insoluble constituents. Chemical characterisation of this material using conventional wet chemical analysis requires that the two phases be analysed separately. We have previously demonstrated near infrared (NIR) models that successfully predict the chemical composition of the solid phase after separation, washing and drying. In this work, we present the current version of this calibration model, as well as a model that uses spectra of the whole slurry samples (without separation) to predict the solids composition in situ. Removing the slurry solid/liquid separation step saves large amounts of time and effort during analysis. The model using washed and dried solids provided predicted vs measured correlation coefficient ( R 2 ) values of 0.97, 0.99 and 0.98 and root mean square error of calibration ( RMSEC ) values of 1.5, 0.8 and 0.8 dry weight percent for glucan, xylan and lignin, respectively. These RMSEC values are similar to established wet chemical analysis uncertainties. Validation samples also showed similar uncertainties and an average r 2 value of 0.98 for the major constituents. The whole slurry model provided R 2 values of 0.93, 0.93 and 0.95 and RMSEC values of 2.3, 1.7 and 1.0 dry weight percent for glucan, xylan and lignin, respectively. The RMSEC values are larger than established wet chemical analysis uncertainties. Validation samples showed uncertainties and r 2 values that were not statistically significantly different ( p = 0.05) from calibration model values for glucan, xylan, and lignin. The slurry model was not equivalent to the washed and dried pretreated solids model, with relative increases in RMSEC values of 20%–50% for major constituents. However, the model was highly successful for the intended purpose, which was to predict the composition of samples without the significant added effort of separating, washing and drying solids prior to scanning.