In order to alleviate the pressure of wood pulp supply in China and meet the actual demand of pulping with mixed pulpwood

a study was conducted on near infrared rapid analysis of mixed pulpwood.145 mixed samples of Eucalyptus urophylla× grandis-Acacia mangium were prepared

in which the content of Eucalyptus urophylla×grandis was manually controlled.The near infrared spectra of these samples were collected

and the contents of holocellulose

pentosan and Klason lignin were analyzed by traditional methods.After the original spectra were pretreated by first derivative and standard normal variate

the analysis models for contents of Eucalyptus urophylla×grandis

holocellulose

pentosan and Klason lignin were established by partial least squares method

support vector machine method

artificial neural network method and LASSO algorithm

respectively.Among them

models for contents of Eucalyptus urophylla×grandis and holocellulose established by LASSO algorithm were the best

with their root mean square error of prediction(RMSEP) values of 1.80% and 0.60%

and their absolute deviation(AD) ranges of -3.03%-3.17% and -1.03%-0.98%

respectively

which could be used for accurate and rapid analysis.Besides

the model for content of pentosan established by partial least squares was the best

with its RMSEP value of 0.75% and an absolute deviation range of -1.26%-1.33%

while the model for content of Klason lignin established by the support vector machine method was the best

with its RMSEP value of 0.48% and an absolute deviation range of -0.82%-0.86%.The performance of the two models was suitable for inaccurate analysis.This study provides the possibility for rapid analysis of mixed pulpwood

and also confirms the applicability of LASSO algorithm.