Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{204627,
        author = {Dr. Shwet Vashishtha},
        title = {Machine Learning–Driven Real-Time Prediction of Critical Paper Quality Properties: A Review of Soft Sensors, Process Analytics, and Industrial Deployment in Pulp and Paper Mills},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {3865-3887},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=204627},
        abstract = {The pulp and paper industry is rapidly transitioning from delayed, laboratory-based quality verification toward real-time, data-driven quality control systems built on machine learning, soft sensors, and Industry 4.0 architectures [1–3]. This transition is particularly important for critical paper properties such as grammage or basis weight (g/m²), Cobb value, tensile index, burst index, tear index, ring crush strength, and related strength parameters, as these variables directly influence convertibility, end-use performance, production cost, and customer claims, while several of them are still measured offline or at relatively low sampling frequency [4–6]. Machine-learning models can bridge this gap by learning the relationships among process variables, furnish characteristics, chemical additives, environmental conditions, and delayed laboratory test results, thereby enabling soft sensors for real-time prediction and closed-loop or advisory control [1–2,7–8].
This review critically examines the current state of data-driven quality prediction in papermaking, with particular emphasis on the online prediction of basis weight, Cobb value, and strength-related properties. Published industrial case studies and related research indicate that linear regression, support vector regression, random forests, gradient boosting methods, neural networks, and hybrid first-principles plus machine-learning models have been applied with varying degrees of success, depending on data quality, process stability, sensor availability, and target-property dynamics [7–11]. Recent evidence further suggests that hybrid and adaptive soft-sensor frameworks are particularly promising, as they combine physical consistency with the flexibility required to handle grade changes, regime shifts, and transient operating conditions [2,7,10–11].
The review organizes the field around six major themes: measurement challenges in paper quality, data infrastructure requirements, model families and feature engineering, interpretability and deployment, control-system integration, and research gaps for Indian mills. A practical framework is proposed for mills seeking to deploy machine-learning-based quality prediction models, beginning with data governance and target-property selection and extending to model maintenance, human oversight, and economic validation [1–2,5,12]. The paper concludes that machine learning has matured sufficiently to support real-time prediction of key paper properties in production environments; however, long-term value depends less on algorithmic novelty than on reliable data pipelines, robust recalibration strategies, and effective integration with process knowledge and mill operations [1–2,7,10].},
        keywords = {Machine learning; soft sensors; paper quality prediction; basis weight (GSM); Cobb value; strength indices; papermaking process control; Industry 4.0},
        month = {June},
        }