Using Near Infrared Spectral To Identify Papermaking Raw Material Combining Wavelet Transform With Artificial Neural Network

The papermaking raw materials have different pulping and paper-making properties because of their difference of inner compositions and structures. The traditional chemical analysis method is time-consuming and high cost, cannot satisfy the online demand of pulp and breeding process, lots of rapid analysis and evaluation in practice. The advantages (rapid, efficient and low cost etc) of near infrared spectral analysis (NIRSA) technology solved this problem rightly and lead it becomes a promising application in pulping and paper-making industry.Wavelet transform have the property of local analysis in detail and can reveal the information ignored by the other signal analysis method. Compared with the traditional signal analysis technology, wavelet transform can de-noising and compress the signal in almost no information lost. As a new pretreatment method, wavelet transform is not only better than conventional method, but also have the advantages of simpler and flexible operation procedure. As an intelligent bionic model, artificial neural network (ANN) provides a powerful tool for the complicated problem of pattern recognition due to its nonlinear treatment, strong fault-tolerance and self-learning ability as well as injecting a new vitality into chemistry especially into the field of pattern recognition.Based on the requirements of paper pulping and paper-marking industry and the nature of near infrared spectral technology, this research used wavelet transform as a new method to pre-treat near infrared spectrum of the papermaking raw materials, used the principal component analysis to reduce dimensions of near infrared spectral and extract characteristics for clustering. Then the artificial neural network based on the principle of error back propagation (BPANN) was employed to build the model which can identify the different kinds of wood materials with near infrared spectrum effectively.