DWT-PLS Regression on Near-Infrared Spectra for Moisture and Volatile Determination of Coal

Abstract

We studied moisture and volatile determination in bituminous coal samples using near-infrared (NIR) spectra. This research was developted by applying partial least squares regression (PLS) and discrete wavelet transform (DWT). Firstly, NIR spectra were pre-processed by DWT for fitting and compression. Then, DWT coefficients were used to build regression model with PLS. We used NIR spectra to determination moisture and volatile determination in coal samples seperately and simultaneously. Through parameters optimization, the results show that DWT-PLS can obtain satisfactory performance for separate and simultanous determination.