Machine learning-based digital twin of a conveyor belt for predictive maintenance

The problem of achieving a good maintenance plan is well-known in the modern industry. One of the most promising approaches is predictive maintenance, which schedules interventions based on predictions made by collecting and analyzing data from the process. However, to the best of the authors' knowledge, this approach is still not widespread and known enough, and particularly, the real-case scenarios of its application appear not exhaustive. To contribute to fill this gap, this work proposes a digital twin (DT), which performs a predictive maintenance approach for a conveyor belt within a real-case scenario with the overall goal of predicting faults during normal belt operations. Specifically, the core of the implemented DT is a model that analyzes the data collected by various sensors distributed along the conveyor belt. In turn, this model exploits a machine learning-based algorithm that predicts the insurgence of faults. The tests of the developed solution, conducted within a real scenario, demonstrated good precision and accuracy in identifying the fault status and also in a time deemed acceptable for the involved stakeholders.