Quality prediction in pulp bleaching: application of a neuro-fuzzy system

Abstract

In chemical industries, as paper pulp, quality control is a decisive task for competitiveness. Bleaching is a determinant operation in the quality of white pulp for paper. Quality prediction is decisive in quality control. However, the complexity of the bleaching process (and in general of industrial processes), its nonlinear and time-varying characteristics does not allow to develop reliable prediction models based on first principles. New tools issued from fuzzy systems and neural networks are being developed to overcome these difficulties. In this paper a neuro-fuzzy strategy is proposed to predict bleaching quality by predicting the outlet brightness. Firstly, a fuzzy subtractive clustering technique is applied to extract a set of fuzzy rules; secondly, the centers and widths of the membership functions are tuned by means of a fuzzy neural network trained with backpropagation. This technique seems promising since it permits good results with large nonlinear plants. Furthermore, it describes the plant using a set of linguistic rules, which can be a basis for interpretable models, more intuitive for operators.