Abstract

Partial least squares (PLS) regressions were carried out to establish a mathematical correlation between the wet-laboratory reference values and the Fourier transformed near infrared spectra. The wavenumber range resulting from the investigations in Part 1 of this series was examined in detail, evaluating several data pre-processing methods. The data set was spilt into a cross-validation set and a test set to validate each model. Then the data set first used for cross-validation was used as a test set and vice versa. The coefficients of determination and the errors were very similar, but the number of PLS vectors varied widely, depending on the pre-processing method used. Applying the evaluation step used in Part 1, namely the prediction of lignin content of 732 additional spectra (366 wood samples with unknown lignin content) revealed which pre-processing method gave the most consistent results. The most appropriate model was found with the underlying assumption that it will cover a wide range of the natural variability of spruce wood and thus be applicable to as many samples as possible. The results of the models of cross-validation and test-set validation were compared to the results of a model calculated with all spectra (CV) and thus comprising a higher variability of the wood spectra. It is shown that the three models were equal and that the final model with the following parameters is highly qualified for prediction: wavenumber range 6102 cm⁻¹-5762 cm⁻¹, data pre-processing method 1st Der-MSC, number of PLS vectors (rank) = 2, r² = 0.926, root mean square error of estimation (RMSEE) = 0.24%, root mean square error of cross-validation (RMSECV) = 0.25%, an estimated root mean square error of prediction (RMSEP) = 0.25% with a corrected estimated RMSEP = 0.22%, bias = 0.00038 and a residual prediction deviation (RPD) = 3.7. These results confirm the importance of careful validation.

© 2011 IM Publications LLP

High accuracy determination of copper in copper concentrate with double genetic algorithm and partial least square in laser-induced breakdown spectroscopy

Haochen Li, Meizhen Huang, and Huidi Xu
Opt. Express 28(2) 2142-2155 (2020)

Nondestructive quantification of the soluble-solids content and the available acidity of apples by Fourier-transform near-infrared spectroscopy

Yibin Ying, Yande Liu, and Yang Tao
Appl. Opt. 44(25) 5224-5229 (2005)

Acidity measurement of iron ore powders using laser-induced breakdown spectroscopy with partial least squares regression

Z.Q. Hao, C.M. Li, M. Shen, X.Y. Yang, K.H. Li, L.B. Guo, X.Y. Li, Y.F. Lu, and X.Y. Zeng
Opt. Express 23(6) 7795-7801 (2015)

Rapid determination of the main components of corn based on near-infrared spectroscopy and a BiPLS-PCA-ELM model

Lili Xu, Jinming Liu, Chunqi Wang, Zhijiang Li, and Dongjie Zhang
Appl. Opt. 62(11) 2756-2765 (2023)

Quantitative analysis of bayberry juice acidity based on visible and near-infrared spectroscopy

Yongni Shao, Yong He, and Jingyuan Mao
Appl. Opt. 46(25) 6391-6396 (2007)