Fourier Transform Infrared Spectroscopic Estimation of Residual Lignin Content in Wood Holocellulose (Pinus brutia Ten.) Using Artificial Neural Network Modelling

Abstract

Wood samples from brutian pine (Pinus brutia Ten.) were delignified at different temperatures and different times to generate holocellulose samples of varying residual lignin content. Fourier transform infrared (FTIR) spectra of holocellulose samples were studied in the range of 4000- 400 cm^-1 at a resolution of 4 cm^-1. The spectral bands at 1508, 1421, 1372, 1265, 1158, 1054 and 1030 cm^-1 were used for estimating residual lignin content in the holocellulose samples. Artificial neural network (ANN) modelling was used to predict the amount of residual lignin from FTIR spectral data of holocelluloses. Artificial neural networks trained by Levenberg-Marquardt algorithm were applied for constructing and optimizing calibration models using MATLAB software. For four hidden neurons, three layered ANN model was scored an average relative error of prediction skill of 0.58 %.