|
Estimation of Pinus brutia Ten. wood density from
Fourier Transform Infrared (FTIR) spectroscopic bands by
Artificial Neural Network (ANN)
Bilgin Guller* and Samim
Yasar
Department of Forest Products Engineering, Faculty of
Forestry, Suleyman Demirel University, Isparta-32260,
Turkey.
*Corresponding author. E-mail:
bilginguller@orman.sdu.edu.tr. Tel: +90 246 2113970.
Fax: +90 246 2371810.
Accepted 3 June, 2010 |
|
In this study, the rapid Fourier transform infrared (FTIR)
spectroscopic method was used to indirectly measure the wood
density of Pinus brutia Ten. samples. A model was
constructed to relate FTIR data to wood density determined
by laboratory analysis, through the application of
artificial neural network (ANN) modelling approach to a set
of calibration observations. The proposed model with two
hidden neurons performed very good to estimate the wood
density with high correlation R2 values of 0.9833
for training and 0.9814 for testing, respectively, and with
a low prediction error of 0.71% in the validation. This
analysis showed that ANN coupled with FTIR spectroscopy
could be used to accurately predict the density of wood
samples.
Key words:
Wood density, Fourier transform infrared (FTIR)
spectroscopy, Artificial neural network (ANN), Pinus
brutia. |
