Digital Twin Based on Historical Data and Simulation Results: Fault Detection and Estimation of the Remaining Useful Life of a Cyclone Bag Filter

Abstract

This study deals with the development of a digital twin for monitoring the operating conditions of a cyclone bag filter installed on the suction system of a wheat mill. The model aims to be used for fault identification and real-time prediction of the remaining useful life (RUL). Computational fluid dynamics simulations were performed to characterize in detail the fluid-dynamic behavior of the airflow inside the system under different conditions of filter sleeve clogging. Furthermore, the simulation results were used to identify a location for the installation of a new velocity sensor that would allow, together with the pressure drop measured at the ends of the filter, monitoring of the systems’ conditions. A model able to assess the filter’s operating state, identify failure events or operating anomalies, and make a prediction of the RUL was then developed. A possible implementation of the developed model, based on the simulation results that aimed to optimize the management of the sleeve cleaning cycles was also proposed. The developed digital model was then tested on a working cycle lasting one year, in which a sleeve failure was simulated. It was shown how the simultaneous monitoring of the two identified quantities allows for the correct identification of the failure and the accurate prediction of the RUL.