Redes neurais aplicadas à determinação de testes de resistência física em papel reciclado

Abstract

The production of recycled paper and the paper packaging market have seen significant growth in recent years due to economic and environmental advantages. Understanding the physical properties of paper is of great interest to the industry, as they are crucial in conversion processes and product quality. Along with the increase in automation and the interest in improving factory process control, there has also been an increase in the amount of available process data. This study aims to use multivariate analyses and machine learning models to understand the behavior of physical paper properties, such as delamination and compression resistance, based on certain process variables. Multivariate Linear Regression, Decision Tree, and Artificial Neural Networks models were implemented using Python algorithms, allowing for model fitting to the training dataset and for checking dispersion, associated error, partial dependence, and the influence of process variables on the physical resistance results of the validation set. The results for the Neural Networks model for delamination and compression resistance showed R² = 77% and R² = 73%, respectively, with partial dependence behavior very similar to that described in other bibliographic.