Paper quality enhancement and model prediction using machine learning techniques

• Published in: Results in engineering Vol. 17; p. 100950
• Main Authors: Kalavathi Devi, T., Priyanka, E.B., Sakthivel, P.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2023; Elsevier
• Subjects: Caliper; Error; Machine learning; Moisture; Steam; Weight; Caliper; Error; Moisture; Weight; Steam; Machine learning
• ISSN: 2590-1230; 2590-1230
• DOI: 10.1016/j.rineng.2023.100950

A machine learning approach demonstrated in the proposed study predicts the parameters involved in paper quality enhancement in real time. To control the steam pressure during paper manufacture, machine learning algorithms have been used to model different parameters such as moisture, caliper, and weight (grammage). The training and testing data sets were obtained to develop several machine learning models through several data from the parameters of the paper-making process. The inputs considered were moisture, weight, and grammage. As a result, the developed model showed better results by showing less execution time, fewer error values such as root mean squared error, mean squared error, mean absolute error, and R squared score. In addition, modeling was carried out based on model interpretation and cross-validation results, showing that the developed model could be a more useful tool in predicting the performance of the steam pressure and input parameters in the paper-making process. A comparison of results shows that the k-Nearest Neighbor algorithm outperforms the other machine learning techniques. Machine learning is also used to predict the efficiency of steam pressure reduction. •Implementation of ML based model on parameter estimation and control on paper manufacturing.•Based on model interpretation and cross validation, performance prediction is done by taking steam pressure.•Statistical result reveals the k-Nearest Neighbor algorithm outperforms to predict the efficiency on steam pressure reduction.•ML Model takes inputs as are moisture, weight, and grammage to attain less computation time with minimum errors.