Real-Time Monitoring and Performance Prediction of a Centrifugal Pump System Using Simulation-Based Digital Twin with Reduced Order Model (ROM)

This study applies simulation-based digital twin technology to a centrifugal pump system for real-time monitoring and performance prediction. The analysis method is to analyze the flow field of the centrifugal pump system by flow analysis, and the structural analysis of the impeller is performed using the analyzed flow field. 100 DPs(Design Points) were selected using the Latin Hypercube Sampling(LHS) method, and Reduced Order Models(ROM) were created by learning 90% of the DPs. The RS ROM(Response Surface ROM) predicts flow and pressure difference(PD), and the Static ROM predicts pressure, speed, and stress on the impeller. We conducted a lift and efficiency comparison experiment, and the results showed an error of 0.3 [m] in lift and about 10 [%] in efficiency. Pressure difference and flow comparison experiments showed an average pressure difference error of 0.927 [kPa] and an average flow error of 20.00 [liter/min]. The virtual sensor predicted the speed, pressure, and stress on the impeller in the event of cavitation.