近红外光谱法测定闽楠纤维长度及木质素质量分数

Determination of Fiber Length and Lignin Content of Phobe bournei Using Near-Infrared Spectroscopy

以20年生闽楠(Phobe bournei)人工林中的闽楠为研究对象,钻取直径为12 mm的木芯带回实验室并制成粉末状。采用台式近红外光谱仪采集木芯粉末样品的光谱信息,利用硝酸-氯酸钾法、紫外分光光度计法测定闽楠纤维长度和木质素质量分数,使用偏最小二乘法建立相关模型。结果表明:建立纤维长度和木质素最佳预测模型的波段分别为1000.00~2000.00、1250.00~2000.00 nm。使用不同方法对近红外光谱数据进行预处理,发现建立纤维长度和木质素预测模型的最佳方法均为“趋势归一化+一阶导数处理”。最佳纤维长度预测模型的校正集相关系数(R_(C))为0.930、交互验证集相关系数(R_(V))为0.981、校正集均方根误差(R MSEC)为0.861、交互验证均方根误差(R_(MSEV))为0.763;木质素质量分数预测模型的校正集相关系数、交互验证集相关系数分别为0.94、0.95,校正集均方根误差、交互验证均方根误差分别为0.013、0.013。外部样品检验模型预测性能表明,预测值与实测值间差距较小,说明近红外光谱分析技术能够用于闽楠纤维长度和木质素质量分数的快速预测。

Using 20-year-old Phobe bournei from a P.bournei plantation as the research subject,wood cores with a diameter of 12 mm were drilled and brought back to the laboratory to be made into powder.The spectral information of the wood core powder samples was collected using a desktop near-infrared spectroscopy instrument.P.bournei fiber length and lignin content were measured using nitric acid-potassium chlorate method and UV spectrophotometry,and correlation models were established using partial least squares method.The results showed that the optimal prediction models for fiber length and lignin content were found in the wavelength ranges of 1000.00-2000.00 and 1250.00-2000.00 nm,respectively.Different preprocessing methods were applied to the near-infrared spectroscopy data,and it was determined that the best methods for establishing the prediction models for fiber length and lignin content were“trend normalization+first-order derivative treatment”.The best prediction model for fiber length had a calibration set correlation coefficient(R_(C))of 0.930,validation set correlation coefficient(R_(V))of 0.981,calibration set root mean square error(R MSEC)of 0.861,and validation set root mean square error(R_(MSEV))of 0.763.The prediction model for lignin content had calibration set correlation coefficient and validation set correlation coefficient of 0.94 and 0.95,respectively,with calibration set root mean square error and validation set root mean square error of 0.013 and 0.013,respectively.External sample verification of the model’s predictive performance indicates that there is a small difference between predicted values and actual measured values,demonstrating that near-infrared spectroscopy analysis technology can be used for rapid prediction of P.bournei fiber length and lignin content.