Predictive Maintenance in the Production of Steel Bars: A Data-Driven Approach

The ever increasing demand for shorter production times and reduced production costs require manufacturing firms to bring down their production costs while preserving a smooth and flexible production process. To this aim, manufacturers could exploit data-driven techniques to monitor and assess equipmen’s operational state and anticipate some future failure. Sensor data acquisition, analysis, and correlation can create the equipment’s digital footprint and create awareness on it through the entire life cycle allowing the shift from time-based preventive maintenance to predictive maintenance, reducing both maintenance and production costs. In this work, a novel data analytics workflow is proposed combining the evaluation of an asset’s degradation over time with a self-assessment loop. The proposed workflow can support real-time analytics at edge devices, thus, addressing the needs of modern cyber-physical production systems for decision-making support at the edge with short response times. A prototype implementation has been evaluated in use cases related to the steel industry.