Prediction of strength properties of poplar alkaline peroxide mechanical pulp using near infrared spectroscopy and multivariate calibration

Abstract Application of near infrared (NIR) spectroscopy to pulp and paper research is almost based on chemical pulping, but little research has been directed toward the use of NIR to mechanical pulping. Compared with chemical pulp, mechanical pulp exhibits certain attractive qualities such as high pulping yield, sufficient opacity and bulk. However, because non-cellulose components (hemicelluloses and lignin) are preserved largely in the produced fibre, mechanical pulping process emerges easily production problem, such as inefficient chemical impregnation and unstable pulp properties. Therefore, it is necessary to explore a rapid measurement technique suitable for pulping process control. This study investigated the feasibility of NIR to predict rapidly the strength properties of mechanical pulp. Alkaline peroxide mechanical pulp (APMP), using poplar as raw material, was produced at pilot plant scale, different type of mechanical compressive pre-treatment and different pulping degree were implemented to increase variability in the obtained pulp samples. NIR spectra were collected from the rough and smooth surface of pulp handsheets to develop calibration models for tensile, burst, and tear index. Calibration statistics results indicated that rough surface spectra could provide more accurate and robust calibrations compared with smooth surface spectra. Additionally, spectral signal correction was proved to be a critical step for NIR calibration optimization. Based on spectra collected with multiplicative scatter correction and derivatives pre-processing, excellent NIR prediction performance was obtained with the root mean square error of 1.776 Nm g−1 for tensile index, 0.108 kPam2 g−1 for burst index and 0.116 mNm2 g−1 for tear index, the RPD values greater than 2.0 satisfying quantitative analysis.