A Fault Detection Framework Based on Data-Driven Digital Shadows

The popularization of Industry 4.0 and its technological pillars has allowed prognostics and health management (PHM) strategies to be applied in complex systems to optimize their performance and extend their useful life by taking advantage of a digitalized, integrated environment. Due to this context, the use of digital twins and digital shadows, which are virtual representations of physical systems that provide real-time monitoring and analysis of the health and performance of the system, has been increasingly used in the application of fault detection, a key component of PHM. Taking that into consideration, this work proposes a framework for fault detection in engineering systems based on the construction and application of a digital shadow. This digital shadow is based on a digital model composed of a system of equations and a continuous, real-time communication process with a supervisory control and data acquisition (SCADA) system. The digital model is generated using monitoring data from the system under study. The proposed method was applied in two case studies, one based on synthetic data and another that uses a simulated database of an operational generating unit of a hydro-electric power plant. The method, in both case studies, was able to detect faults accurately and effectively. Besides, the method provides by-products that can be used in the future in other applications, helping with the PHM in other aspects.