Condition monitoring of an industrial book-cutting machine using a novel mixture of vibration and physics-based features

This study presents an enhanced approach to machinery diagnostics and prognostics, using artificial intelligence and machine learning to improve vibration monitoring systems for industrial equipment maintenance. Targeting a complex automatic book-cutting machine, the proposed diagnostic system integrates a mixture of diverse feature types, combining accelerometric data with physics-based features to provide a more comprehensive and accurate monitoring solution. A dataset, developed through a full factorial design of experiments, covers a range of operational states, while data processing generates an optimal feature set for maintenance decision-making. This hybrid approach, exploiting the novel combination of features, offers superior identification of machine states of health compared to traditional vibration monitoring, generating benefits such as reduced downtime and costs, as well as enhanced product quality. In the specific application analyzed, the proposed method improved classification accuracy by approximately 7% compared to traditional vibration monitoring techniques. By integrating these advancements, achieved through sensor fusion and multi-channel data, into a Supervisory Control And Data Acquisition (SCADA) system, this research aligns with the goals of Industry 4.0, supporting digital and cyber-physical manufacturing systems with an intelligent, data-driven condition-based maintenance strategy suitable for modern automated environments.