Rapid compositional analysis of sawdust using sparse method and near infrared spectroscopy

Abstract

This paper proposes to measure the components of sawdust by combining a new sparse method with near infrared (NIR) spectroscopy technology. The spectroscopic data of sawdust samples are acquired by the means of Fourier transform near-infrared (FT-NIR) spectrometer. Wavelet filter is used to remove undesired noises from the spectroscopic data, and multivariate statistical methods, such as principal component regression (PCR), partial least squares regression (PLS) and least absolute shrinkage and selection operator (LASSO) are used to model the relationship between the spectroscopic data and sawdust composition. The constructed model is then tested on a set of new samples. Compared with PCR and PLS, it is shown that LASSO, a sparse method, is capable of constructing a sparse model with stronger ability in interpretation while retaining good modeling accuracy.