Machine Health Diagnosis and Prognosis: A Predictive Maintenance Approach for Fiberboard (MDF) Production

Abstract

Predictive maintenance (PdM) leverages machine learning (ML) to enhance equipment reliability and production efficiency in Medium-Density Fiberboard (MDF) manufacturing. By analyzing operational data - such as temperature, vibration, and pressure - PdM forecasts equipment failures, enabling proactive maintenance that minimizes downtime and reduces costs. This study introduces an ML-driven maintenance framework utilizing historical failure logs and performance records to predict the Remaining Useful Life (RUL) of critical components in MDF production. Focusing on key processes like chip refining, fiber drying, resin mixing, and hot pressing, the framework employs supervised learning models, including Random Forest and Gradient Boosting, to analyze maintenance trends and operational parameters. The implementation of PdM in MDF manufacturing presents challenges, such as high initial investments, sensor calibration, and the necessity for skilled personnel to interpret predictive insights. Despite these hurdles, the proposed framework demonstrates the potential to transition from reactive to proactive maintenance strategies, thereby enhancing production efficiency and equipment longevity. The study's findings suggest that integrating PdM into MDF production can lead to significant operational improvements, though careful consideration of implementation challenges is essential for success.