An improved body force model based on a data-driven region-segmentation combinational loss model for a transonic fan rotor under uniform and non-uniform inflow

Boundary layer ingestion (BLI) fans are required to continuously operate under distorted inflow conditions, which severely reduces the fan's efficiency. Therefore, it is necessary to focus efforts on designing a high-performance distortion-tolerant fan. During the preliminary design stage of the fan, it is inevitable to repeatedly evaluate the aerodynamic performances of different design cases through low-fidelity computational approaches. However, the predictive accuracy of these low-fidelity computational methods generally depends on the precision of the loss models integrated into them. These loss models are built using simplified physics or human experience, making it difficult to depict the complex non-linear mapping relationships between losses and their influencing factors under BLI inflow distortion. In this work, a high-accuracy data-driven based region-segmentation combinational loss model is used for the first time to attempt to improve the overall prediction accuracy of the body force model. As a low-fidelity computational approach, the body force model has the ability to predict the spatial distribution of three-dimensional, non-uniform flow under inflow distortion conditions. Various operating conditions, including different rotational speeds, radial inflow, and BLI-distorted inflow, are selected to test the predictive accuracy of this improved body force model based on a region-segmentation combinational loss model. The results show that, compared with the one based on traditional loss prediction approach, the improved body force model in this work has shown higher prediction accuracy for the fan adiabatic efficiency characteristic curve. In addition, further evaluations of the flow loss spatial distributions under BLI inflow distortion conditions indicate that the body force model based on region-segmentation combinational loss model can more accurately capture the radial loss distributions at different circumferential locations. Specifically, for the loss distribution in the tip area, it can capture the sharp increase in variation trend near the end wall, and the average prediction errors can be reduced by more than 10% at different annulus locations under various BLI inflow distortions. Meanwhile, this improved body force model also enhances the accuracy of predicting the loss distribution in the circumferential direction at different span positions, including more accurately capturing the circumferential positions corresponding to the ma