Search

With the rapid growth of China's economy, applied research on hydropower engineering has received an increasing amount of attention. However, since hydraulic electromechanical devices often work in actual industrial manufacturing environments at high loads for a long time, their health status is hardly predicted. By introducing digital twins technique, this paper proposed a predictive maintenance model for electromechanical devices to solve the problems. Firstly, multiple sensors are implemented on critical parts of the hydraulic electromechanical devices to collect devices' physical and spatial signals. Secondly, constructing the digital twins model of electromechanical devices with the sensing data and the devices' structural characteristics. Finally, by transfer learning, a comprehensive and reliable fault diagnosis method is designed to predict the remaining life of the devices and make decisions for facilities maintenance. Experiments show that the proposed model performs the best accuracy rate compared with the other methods. [Display omitted] • Use edge devices to integrate heterogeneous data in smart water conservancy. • Build four kinds of models with digital twins for qualitative analysis. • Train neural networks with digital twins' real-time data to make fault diagnoses. • Make safety assessment criteria for smart water devices with diagnosis results. [ABSTRACT FROM AUTHOR]