Enhancing sustainability in tyre manufacturing with machine learning and knowledge graphs: An energy-based maintenance solution

The manufacturing industry is facing increasing pressure to adopt sustainable strategies. Many focus on improving existing maintenance practices, prioritising energy efficiency and operational sustainability. However, conventional maintenance often relies on energy waste indicators (e.g., vibration, sound), neglecting the potential offered by primary energy sources (e.g., hydraulic, electrical). The study addresses this gap by introducing an Energy-Based Maintenance (EBM) solution that leverages Machine Learning (ML) and Deep Learning (DL) algorithms to monitor primary energy signals, enabling a more sustainable fault prediction. The selection of ML/DL algorithms is identified through a meta-analysis and Recursive Feature Elimination (RFE) performs feature selection. EBM applicability is demonstrated in a rubber mixing machine's tyre manufacturing process. Additionally, to address the challenges of understanding latent failure mechanisms, an Exploratory Network Analysis using the Gaussian Graphical Model (GGM) was introduced, offering novel insights into fault patterns. The results show that EBM can reduce energy consumption by 3.05 %–7.75 % compared to the existing maintenance practice, suggesting significant energy-efficient advancements. By operationalising primary energy source as a predictive variable, in combination with knowledge graphs, the work contributes to the advancement of sustainable prescriptive maintenance practices.