A Predictive Maintenance Framework for Soda Ring Mitigation in Rotary Lime Kilns
Open Access DepositedA Machine Learning Approach for the Pulp and Paper Industry
Abstract of Praxis Soda ring formation in rotary lime kilns remains a persistent challenge for the pulp and paperindustry, contributing to unplanned downtime, reduced energy efficiency, and elevated maintenance costs. Traditional mitigation strategies are largely reactive, relying on manual interventions that disrupt operations and increase production expenses. This praxis develops a predictive maintenance framework that leverages machine learning models to forecast soda ring-related downtime and improve kiln reliability. Using four years of historical process and maintenance data, the study evaluates lagged operational variables at 7-, 14-, and 30-day intervals to determine their statistical association with downtime events. Logistic regression identified key predictors, including cold-end temperature, kiln feed solids, and airflow, with the 30-day lag producing the most consistent predictive power. Building on these findings, five regression-based machine learning models Linear Regression, Ridge Regression, Lasso Regression, Random Forest, and XGBoost were applied to forecast kiln zone temperatures. Model performance was assessed using R2, CV(RMSE), MAPE, and classification metrics. Results show that Random Forest and XGBoost outperformed regression-based models, achieving R2 values exceeding 0.80 in most zones while maintaining low error rates. The findings confirm that machine learning can provide accurate, proactive forecasts of kiln instability, support earlier interventions and reduce reliance on reactive maintenance. By identifying the 30-day lag as the optimal predictive horizon and demonstrating the comparative strengths of ensemble models, this research contributes a practical, data-driven framework for improving kiln performance. This methodology offers broader applicability to other rotary kiln-based industries where downtime reduction and process stability is key.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.