【作者】 姜伟；

【导师】 韩光亭；

【作者基本信息】 东华大学 ， 纺织工程， 2013， 博士

【摘要】 植物纤维原料广泛应用于纺织、造纸、饲料及生物质能源等多个领域。而化学成分含量作为其主要的参数指标对这些领域的原料评价、科学研究及工业生产都具有重要的指导意义。然而,大量对植物纤维原料化学成分的定量研究一直停留在湿化学分析方法阶段,且部分领域的分析标准存在一定的误差。随着科技的发展,新的分析方法逐渐显露出巨大的优势。本研究就湿化学分析方法存在的问题,结合现代分析技术如高效液相色谱法(HPLC)、近红外光谱法(NIR)等的帮助建立了相对准确的植物纤维原料化学成分湿化学分析体系及快速的近红外模型预测体系。论文的主要工作及结论如下：1.利用真空干燥、傅里叶转换红外光谱及高效液相色谱技术建立了相对准确的植物纤维原料化学成分湿化学分析方法。(1)利用真空干燥技术解决了分析过程中有机溶剂抽提物的增重问题,同时避免了高温加热过程中抽提物蒸发和氧化等问题。(2)通过多次循环试验结合origin软件和数学拟合得到多糖热降解动力学方程,并使用高效液相色谱技术验证了纤维原料在湿化学分析中多糖高温降解的问题。接着,根据所建立多糖热降解动力学方程和多元正交实验建立了适当的湿化学分析温度体系。最终确定烘干温度60℃和水浴温度60℃下所得到的水溶物含量较为准确。(3)研究木质素和单糖定量分析方法。发现并解决了热碱降解木质素而在分析过程中造成的误差,使分析结果的相对误差大幅降低。2.优化样品近红外采集最佳条件。研究了原料粒径大小和光谱分辨率对近红外光谱采集准确度的影响,并确立了合理的近红外光谱采集参数。以美国南部松树为对象,通过对木板、1/8英寸木片以及20目、40目和80目木粉为样品进行近红外光谱扫描,分别建立了相应的木质素预测模型。结果发现：随着样品粒径的减小,近红外光谱重现性提高,其木质素预测模型交叉验证决定系数R2提高,且交叉验证预测误差均方根RMSEP减小。而在样品由木板粉碎成1/8英寸木屑和由40目到80目这两个过程中近红外模型预测能力提高显著。最终确定80目为本研究采集近红外光谱的最佳样品粒径。使用80目样品,在光谱分辨率16cm-1、4cm-1和2cm-1下对样品进行近红外光谱扫描,并建立相应木质素近红外光谱。结果显示在光谱分辨率为4cm-1下所建立的木质素近红外模型最好,最终确定光谱分辨率4cm-’为本研究近红外光谱扫描最佳分辨率。3.确定了关键化合物的近红外光谱波段,研究特征波段和全光谱建模对相应化学成分近红外模型质量的影响,并建立软木主要化学成分近红外预测模型。通过对样品进行针对性预处理以除去样品中相应化学成分,然后采集样品原料和预处理后样品的近红外光谱。对两种光谱进行对比并结合其它分析,确定相应化学成分近红外特征谱段。最终得到植物纤维原料主要化学成分近红外吸收的特征谱段。其分别为：水分5050-5360cm-1,色素8500-10000cm-1,脂蜡质(有机溶剂抽提物)4000-8500cm-1,木质素5800-6900cm-1,多糖类(纤维素、半纤维素)4000-5100cm-1,6900-8500cm-1。以美国南部松树(软木)为对象,使用研究得到的湿化学分析改进方法结合美国国家可再生能源实验室(NREL)提供的HPLC方法对其进行有机溶剂抽提物、木质素和各种单糖的定量分析。根据美国南部松树的半纤维素结构特征使用单糖成分对其进行定量,进而得到纤维素含量。使用全光谱和研究得到的各种化学成分特征谱段对所有化学成分进行近红外建模工作。结果发现使用特征光谱建模后,样品有机溶剂抽提物、木质素,综纤维素和纤维素预测能力均比全光谱建模有大幅提高。4.进行了近红外模型预测能力扩展研究和近红外模型稳健性研究。选取木质素为建模指标,以美国南部松树为样品,通过氯化法去木质素作用对高木质素含量样品进行逐级去除,从而扩展样品木质素含量范围。研究建立预处理样品木质素近红外模型、样品原料和预处理样品总体木质素近红外模型。最终建模结果为预处理样品木质素模型交叉验证决定系数R2=0.99,交叉验证均方残差为RMSEP=0.72%:预测结果决定系数R2=0.99,预测误差均方根为RMSEP=0.68%,相对分析误差RPD值为12.7。可对经预处理后的木材样品进行准确的预测工作。样品原料和预处理样品总体木质素模型预测结果决定系数r2=0.99,预测误差均方根为RMSEP=0.6%,相对分析误差RPD值为14.34。从而可对木材所有样品(木材原料和预处理材料)进行预测工作。继续使用新鲜样品对模型预测新样品的稳健性进行研究。研究结果表明当同类新鲜样品指标含量高于模型预测能力时,模型对新鲜样品预测能力较差。研究发现将新鲜样品其中一个样品信息加入所建立的近红外模型中时,模型对新鲜同类样品的预测能力将会大大提高,此方法可用于提高近红外模型质量,增加模型的稳健性和对未知样品集的预测。5.研究对各种类型植物纤维原料,包括苎麻(韧皮类植物),美国南部松树(软木)及多品种硬木分别建立了相应的近红外预测模型。建模结果如下：使用改进苎麻化学成分分析方法对苎麻主要化学成分进行分析,然后对其进行近红外建模,最终得到苎麻主要化学成分近红外模型纤维素预测决定系数r2=0.95, RPD=4.54,胶质含量r2=0.91, RPD=3.37,可以对苎麻纤维素和胶质进行精确预测工作；半纤维素r2=0.75, RPD=2.61,模型质量较差。建立了硬木和软木主要化学成分近红外模型。最终得到硬木和软木主要化学成分如有机溶剂抽提物、木质素、综纤维素和纤维素近红外模型决定系数均高于0.90,具有较好的预测能力。硬木和软木半纤维素同苎麻样品一样决定系数低于0.90,不能胜任精确的预测工作。研究发现,同时对多品种样品进行建模时的模型质量低于单一品种样品建模质量。更多还原

【Abstract】 Fibrous materials are wildly used in textile, paper making, bioenergy and other related area. Quantitative analyze chemical composition of fibrous materials is essential on raw material evaluation, scientific research and producing process. However, most of the researches use wet chemistry as a standard method to quantitive analyze chemical composition in fibrous materials, wet chemistry method also vary in different area. Recent years, modern analysis methods show high advantages than wet chemistry. This research studied the drawbacks of normal wet chemistry analysis method, and tried to establish a more fast and accurate modified method combining with High Performance Liquid Chromatography (HPLC) and near infrared modeling method (NIR). The main results and conclusions of this research are listed blew:1. Established a relatively accurate wet chemistry analysis method using vacuum dry, Fourier Transform Infrared spectroscopy (FTIR) and High Performance Liquid Chromatography (HPLC).(1) Improve the accuracy in testing extractives using vacuum-dry technique, this mehod can avoid the extractive to be evaporated or oxidezed during oven dry.(2) Studied on thermal degradation kinetics of polysaccharides in fibrous materials by using cyclic test combined with origin software and mathematic curve fitting technique. And validate the equation with HPLC test. Using multivariate orthogonal experiments coupled with thermal degradation kinetics equation to get the best temperature system of wet chemistry on test fibrous materials chemical composition. Finally we found that to oven dry sample at60℃and then extractive samples using water in60℃water bath is the best temperature system to test water solube matter in fibrous materious.(3) Resarch found lignin could be degraded in hot alkali liquor, and the optimized the procedure to reduce the error in testing hemicellulose from10%to1%. Study aslo determined the best testing method for lignin and monosaccharides.2. Optimize the best parameters on collecting near infrared spectrum. Studing the relationship between particle size of sample and spectrum resolution and the quality of spectrum, finally determined the best particle size and spectrum resolution.Using southern pine as the raw material collected the near infrared spectra and then constructed near infrared models on particle size of wood lumber,1/8inch,20mesh,40mesh and80mesh samples, respectively. Results showed that the precision of near infrared model could be improved with the particle size decreasing. There were significantly impovements during particle size changed from wood lumber to1/8inch sample and from40mesh to80mesh sample. Study finally determined that80mesh is the best particle size to collect the near infrared spectrum for near infrared model.Colleted the near infrared spectra and construed the NIR model of80mesh samples under spectrum resolution16cm-1,4cm-1and2cm-1, respectively. Results showed that spectrum resolution of4cm-1was the best parameter to collect NIR spectrum.3. Identified the wavenumber ranges in near infrared spectrum of key chemical composition. Comparing the selected wavenumber range and the whole spectrum range on calibrating the near infrared models quality.Using targeted pretreatment method to remove one chemical composition, then collect near infrared spectrum on raw material and pretreated material. Then identify the wavenumber range of this chemical composition by comparing the NIR spectrum of raw material and pretreated material. By using this method, we finally identify the wavenumber range of chemical composition, which are:water5050-5360cm-1, color element8500-10000cm-1, extractives4000-8500cm-1, lignin5800-6900cm-1, sugars (holocellulose, cellulose and hemicellulose)4000-5100cm-1,6900-8500cm-1.Using southern pine (softwood) as the raw material, quantitatively determine its extractives, lignin and sugars using modified wet chemistry and HPLC method which is proposed by national renewable laboratory (NREL). Then determine the cellulose and hemicellulose content based on the structure of softwood. Then further construted NIR models of the chemical compositon using full wavenumber and selected wavenumber range. Result showed that using selected wavenumber range can significantly improve the pricison of NIR model than using full wavenumber range.4. Research on expand the predictiablity and robust the NIR model.Using southern pine as the raw material, and lignin was determined for NIR model construction. This research tried to expand the lignin content of raw materials by using delignification to decrease the lignin content. The NIR models were established on raw wood samples, pretreated wood samples and all wood samples (combine raw wood sample and pretreated wood sample together). The calibration data and prediction data of pretreated wood lignin model were listed as follows:r square of cross validation0.99, prediction error of cross validation0.72%; r square of prediction0.99, prediction error of prediction0.68%, residual predictive deviation (RPD)12.7. The prediction data of all wood sample lignin model were listed as follows:r square of prediction0.99, prediction error0.6%, RPD value14.34. The results show high predictiablity of lignin model both in prediction precision and prediction range on raw and pretreated wood samples.We also studied the method of robust NIR model. Research found NIR model has low predictiability on predicting sample which chemical composition content beyong the calibration data of NIR model. However, the prediction precision could be improved by adding one of the new bath sample’s information in old NIR model.5. Established NIR models on different species of fibrous materials chemical composition content, including ramie, softwood (southern pine), mixed hardwood (aspen, cotton wood and other species).Quantitative analyzed chemical composition of ramin using modified method in this research. Then NIR models of these chemical compositions were calibrated and validated. Results showed high predictability of NIR models for ramie cellulose, gum. The results of NIR models of cellulose, gum and hemicellulose were r2=0.95,0.91,0.75; RPD=4.54,3.37and2.61, respectively.We also established NIR models for chemical composition on softwood and hardwood, all the r2were over0.9which impled they had high quality to be used for prediction. Study also found that it is much easier to establish a NIR model on pure species than on multispecies samples.更多还原