Process-monitoring laboratories in the pulp and paper industry generally use a combination of wet chemical analyses and physical measurements to certify the fitness of a production pulp for a specific end-use. These laboratory tests require time and the effort of trained personnel, limiting their utility for real-time process control. Here we show that Raman probes of unrefined cellulosic pulps, well-suited to the online measurement of in-process materials, can predict the quality attributes of manufactured papers. The accuracy of prediction improves when the covariance is modelled in a reduced measurement space selected by a data-driven, feature-selection technique referred to as a Template Oriented Genetic Algorithm (TOGA). TOGA, combined with discrete wavelet transform (DWT), isolates functional-group features that correlate best with mechanical properties paper derived from refined pulp. Paper makers refine market pulps to build sheet strength using a beating process that decreases freeness as it increases fibre-fibre bonding. Methods demonstrated here predict manufactured sheet properties obtainable after any specified degree of refining from the Raman spectrum of an unrefined pulp. This analysis capacity will enable both vendors of market pulp and makers of sheet paper to specify in advance the amount of beating required to produce a desired product, thereby saving cost and conserving resources.

纸浆和造纸工业中的过程监控实验室通常结合使用湿化学分析和物理测量来证明生产纸浆是否适合特定的最终用途。这些实验室测试需要时间和受过训练的人员的努力，从而限制了其在实时过程控制中的效用。在这里，我们表明，未精制纤维素纸浆的拉曼探针非常适合在线测量过程中材料，可以预测人造纸的质量属性。当在由称为模板定向遗传算法（TOGA）的数据驱动的特征选择技术选择的缩小的测量空间中对协方差建模时，预测的准确性会提高。TOGA，结合离散小波变换（DWT），分离出与精制纸浆的机械性能最相关的功能组特征。造纸商使用打浆工艺来精炼市场纸浆以增强纸张强度，因为打浆工艺会增加纤维与纤维的粘合力，从而降低游离度。此处演示的方法可根据未精制纸浆的拉曼光谱预测在任何指定的精制度后可获得的成品板的性能。这种分析能力将使市场纸浆供应商和单张纸生产商都能预先指定生产所需产品所需的打浆量，从而节省成本并节省资源。此处演示的方法可根据未精制纸浆的拉曼光谱预测在任何指定的精制度后可获得的成品板的性能。这种分析能力将使市场纸浆供应商和单张纸生产商都能预先指定生产所需产品所需的打浆量，从而节省成本并节省资源。此处演示的方法可根据未精制纸浆的拉曼光谱预测在任何指定的精制度后可获得的成品板的性能。这种分析能力将使市场纸浆供应商和单张纸生产商都能预先指定生产所需产品所需的打浆量，从而节省成本并节省资源。