Present work describes the application of NIR spectroscopy coupled with Partial least square (PLS) regression to estimate the residual chemical composition (in terms of holocellulose and lignin) of wood decayed by attack of biotic agents (such as bacteria) when preserved for a long-time in waterlogged conditions. The evaluations allowed assessing how various parameters may influence the predictive ability of calibration models. These pa- rameters include: pre-processing manipulations (first and second derivatives curves, both normalized and not, were compared with baseline corrected spectra), the ranges used for model calibration, the preliminary condi- tioning of the samples (meals conditioned at both 22 °C/50% r.h. and 60 °C), and the way to take account of the presence of inorganic fillers permeating wood. To such aim, 59 samples from different excavations were con- sidered, with samples belonging to several hardwoods and softwoods (Alnus sp.p., Cupressus sempervirens, Larix decidua, Picea abies, Pinus sp.p., Quercus sp.p., Ulmus sp.p.), and to different states of preservation and burial en- vironments. Values used for calibration were obtained by conventional wet analyses. Results showed that NIR spectroscopy coupled with PLS multivariate analysis could be used to reliably assess the residual chemical com- position of waterlogged decayed wood provided that the calibration range is carefully selected and that original data are suitably pre-processed. However, the best models were different for the two considered components (lignin and holocellulose) and depending on which data set (softwoods or hardwoods) the samples belonged. Fi- nally, it was also shown that the models predictive ability is affected by high ash content (too contaminated sam- ples had to be excluded in order to attain good results).
Measuring the chemical composition of waterlogged decayed wood by near infrared spectroscopy
Pecoraro E;Pizzo B;Macchioni N;
2015
Abstract
Present work describes the application of NIR spectroscopy coupled with Partial least square (PLS) regression to estimate the residual chemical composition (in terms of holocellulose and lignin) of wood decayed by attack of biotic agents (such as bacteria) when preserved for a long-time in waterlogged conditions. The evaluations allowed assessing how various parameters may influence the predictive ability of calibration models. These pa- rameters include: pre-processing manipulations (first and second derivatives curves, both normalized and not, were compared with baseline corrected spectra), the ranges used for model calibration, the preliminary condi- tioning of the samples (meals conditioned at both 22 °C/50% r.h. and 60 °C), and the way to take account of the presence of inorganic fillers permeating wood. To such aim, 59 samples from different excavations were con- sidered, with samples belonging to several hardwoods and softwoods (Alnus sp.p., Cupressus sempervirens, Larix decidua, Picea abies, Pinus sp.p., Quercus sp.p., Ulmus sp.p.), and to different states of preservation and burial en- vironments. Values used for calibration were obtained by conventional wet analyses. Results showed that NIR spectroscopy coupled with PLS multivariate analysis could be used to reliably assess the residual chemical com- position of waterlogged decayed wood provided that the calibration range is carefully selected and that original data are suitably pre-processed. However, the best models were different for the two considered components (lignin and holocellulose) and depending on which data set (softwoods or hardwoods) the samples belonged. Fi- nally, it was also shown that the models predictive ability is affected by high ash content (too contaminated sam- ples had to be excluded in order to attain good results).I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
