A novel combined application of capacitive method and near-infrared spectroscopy for predicting the density and moisture content of solid wood
A novel combined application of capacitive method and near-infrared spectroscopy for predicting the density and moisture content of solid wood
复制标题
电容法和近红外光谱法预测实木密度和含水率的新颖组合应用
DOI：
10.1007/s00226-017-0974-x
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
S. Tsuchikawa
中科院分区：
文献类型：
--
作者：
V. T. H. Tham;T. Inagaki;S. Tsuchikawa
利用电容法和近红外光谱同时预测木材样品的密度和含水率（MC）。通过多变量分析分别考察两种方法的预测精度。结合两个波长在近红外范围内的容量和吸光度，通过特定的模型来预测两者的性质。对908 ~ 1676 nm范围内的所有波长组合进行了测试，选出了决定系数（R2）最高的最佳组合。这种新方法显示了预测和测量数据之间的强相关性，独立于样品厚度和木材种类。从青木到烘干状态的木材样品的预测精度在所有厚度下都显示出良好的结果，其thr2 = 0.79，交叉验证均方根误差(RMSECV) = 0.10 g/cm3，密度的残余预测偏差(RPD) = 2.22, r2 = 0.80, RMSECV = 25.70%， MC的残余预测偏差(RPD) = 2.22。在低于纤维饱和点到烘干状态时，预测MC的r2值略有下降，预测密度的RPD值略有上升。与MC从饱和点起的最大值39.56%相比，MC的RMSECV显著下降（最大5.46%）。这些结果明显优于单独模拟电容法或近红外法获得的结果，并且在估计密度方面的改进尤其明显。结果表明，一种结合电容法和近红外光谱的新装置可以更准确地预测密度和MC。
The use of a capacitive method and near-infrared (NIR) spectroscopy to simultaneously predict the density and moisture content (MC) of wood samples was investigated. Prediction accuracy of both methods was individually investigated by multivariate analyses. The capacity and absorbance at two wavelengths in the NIR range were combined to predict both the properties by the specific models. All wavelength combinations in the range of 908–1676 nm were tested, and the best combination yielding the highest coefficient of determination (R2) was chosen. This novel method showed a strong correlation between predicted and measured data, independent of sample thickness and wood species. The prediction accuracy of the wood samples, from green wood to oven-dried conditions, showed promising results for all thicknesses, withR2= 0.79, root-mean-square error of cross-validation (RMSECV) = 0.10 g/cm3, and residual predictive deviation (RPD) = 2.22 for density andR2= 0.80, RMSECV = 25.70%, and RPD = 2.22 for MC. In the case of below fiber saturation point to oven-dried state,R2value was slightly decreased in the prediction of MC and slightly increased in the prediction of density, but RMSECV of MC declined significantly (maximum 5.46%) compared to the range of MC from saturated point (maximum 39.56%). These results were considerably better than those obtained by modeling the capacitive or NIR method individually, and improvement was particularly apparent in estimating density. The results suggest the possibility of a new device combining the capacitive method and NIR spectroscopy to predict density and MC more accurately.