Carbolea Biomass Research Group, Department of Chemical and Environmental Sciences, University of Limerick, Ireland.
Bioresour Technol. 2012 Sep;119:393-405. doi: 10.1016/j.biortech.2012.05.137. Epub 2012 Jun 7.
Miscanthus samples were scanned over the visible and near infrared wavelengths at several stages of processing (wet-chopped, air-dried, dried and ground, and dried and sieved). Models were developed to predict lignocellulosic and elemental constituents based on these spectra. The dry and sieved scans gave the most accurate models; however the wet-chopped models for glucose, xylose, and Klason lignin provided excellent accuracies with root mean square error of predictions of 1.27%, 0.54%, and 0.93%, respectively. These models can be suitable for most applications. The wet models for arabinose, Klason lignin, acid soluble lignin, ash, extractives, rhamnose, acid insoluble residue, and nitrogen tended to have lower R(2) values (0.80+) for the validation sets and the wet models for galactose, mannose, and acid insoluble ash were less accurate, only having value for rough sample screening. This research shows the potential for online analysis at biorefineries for the major lignocellulosic constituents of interest.
对不同加工阶段（湿切、风干、干燥和粉碎以及干燥和筛分）的芒草样本进行了可见和近红外波长扫描。基于这些光谱建立了预测木质纤维素和元素成分的模型。干燥和筛分扫描给出了最准确的模型；然而，对于葡萄糖、木糖和 Klason 木质素，湿切模型的预测精度也非常高，预测均方根误差分别为 1.27%、0.54%和 0.93%。这些模型适用于大多数应用。对于阿拉伯糖、Klason 木质素、酸溶性木质素、灰分、提取物、鼠李糖、酸不溶残渣和氮，湿模型的验证集 R(2) 值较低（0.80+），而对于半乳糖、甘露糖和酸不溶灰分的湿模型则准确性较低，仅对粗样品筛选有价值。这项研究表明，在生物精炼厂对主要木质纤维素成分进行在线分析具有潜力。
