A new meshless approach for marine power component modelling using digital twin reduced-order technique

In the background of the significant data era, green power represents the future development direction of ships. It is critical to establish fast-response marine power models to evaluate various green power-driven power components. However, when evaluating a finite model, the calculating often requires re-meshing of the model once the model parameters change, resulting in huge amount computing costs. The reduced-order model can address this issue by rapid calculation of the model responses; however, real-time computing of the reduced-order model is still a challenging task. In this paper, we study the basic principle of the digital twin reduced-order model and propose a digital twin reduced-order model. Firstly, the geometric model of a marine centrifugal pump is imported into ICEM meshing software for calculation domain meshing; Secondly, import the mesh file generated by ICEM software into Fluent for CFD numerical simulation and numerical simulation of the reduced-order model; finally, the reduced-order model is verified: under the rated working condition, the calculated head of reduced-order model is 31.98 m, the maximum speed of impeller edge is 24.4 m/s, and the minimum speed at the impeller shaft center is 0.0475 m/s. The errors are less than 1% compared to the theoretically calculated values. This study provides a proof basis for the high-accuracy model required by the fault detection technique based on the reduced-order model, and also proves that the fault detection method based on the reduced-order model is feasible.