Application of Machine Learning Techniques in the Pulp and Paper Industry

Abstract

Bleaching in the pulp and paper industry is essential to produce high-quality paper, but it faces challenges due to the complexity and interdependence of processes, as well as limitations in traditional control methods. Machine learning techniques appear to be a promising solution to predict pulp brightness and adjust operating conditions in real time. This study evaluates four algorithms (Support Vector Regression (SVR), Random Forest (RF), Linear Regression (LR), and Decision Tree (DT)) to predict brightness at the D1 bleaching stage in a pulp and paper industry. Experiments were performed using all features, with normalized features, with Principal Component Analysis (PCA) as feature selection, and with feature selection based on process knowledge. The RF model obtained the best results, with R² of 0.6576, 0.6576, 0.7177, and 0.7291, respectively, in each experiment.