Application Of Multivariate Analysis Algorithms To The Identification Of Recycled Fiber In Paper Products

In recent years,the effective use of recycled fibers,e.g.,deinked pulps,have received great attention and been widely used in many paper related products.In order to reduce the impact to environment and a sustainable development,many countries stipulate that some paper products must contain a certain amount of recycled fiber.However,paper products like handkerchief paper,tissues and napkins,should not contain recycled fiber,since it contains the harmful chemical substances to human body,such as the fluorescent whitening agent and mineral oils.Therefore,it is very important to develop a method that can detect the recycled fiber in paper products,so that the reasonable use of recycled fibers for the sake of both environmental protection and paper use safety can be realized.This thesis conducted the exploration and research in the following aspects:The traditional method to identify the recycled fiber in paper products is mainly to detect residual ink,packing,fluorescence intensity,fiber morphology,etc.However,a small amount of pollutants containing in the raw material or machines in the production process could be misclassified as containing recycled fiber.It is very difficult to discriminate the paper products between the purposely added with small amounts of recycled fiber and those due to the raw material and process factors by the current National Standard Testing methods.In this work,the principal component analysis and BP neural network multivariate analysis algorithms were used to set up classification discriminant models.The model can analyze the data offered by traditional method and identify the paper products contains recycled fiber.The accurate identification rate was 100%.Because of the lack of fast and effective qualitative and quantitative detection means of recycled fiber,a qualitative and quantitative detection model on the basis of multivariate analysis algorithm was developed.In this work,the modeling process and the selecting of algorithms and verifies the stability and accuracy of the model was established and discussed.The accuracy of qualitative analysis was 100% and the fitting relative root mean square deviation of the quantitative analysis was controlled at 2.3% and 1.9%.In view of the limitation of traditional methods for fluorescent whitening agent detection,this paper introduces the multivariate analysis algorithm to solve this problem.The ultraviolet spectrum of the standard solution of fluorescent whitening agent is used to build analysis model by multivariate analysis algorithm.Then the analysis model can calculate the fluorescent whitening agent content of the samples using the ultraviolet spectrum of the standard addition of fluorescent whitening agent.This method could work with good stability and accuracy.The average relative deviation of the quantitative detection was 1.10%.Based on the analysis models,a NIR detection software which can identify the recycled fiber by using a portable near infrared spectrometer was developed.This detection platform can not only provide rapid test of recycled fiber but also load data of paper products or improve the built-in models.