Comparison of Numerical Simulations to a Reduced-Order Model Extended with Splitter Blades

The present work illustrates the numerical validation of the reduced-order model recently developed by some of the authors for the performance prediction and geometry definition of mixed-flow turbopumps with splitter-bladed impellers. The computed performance of the reference turbopump with its unsplitted impeller closely matches the experimental results. The generalized Polynomial Chaos (gPC) method is used to reduce the computational cost of the comparison with the predictions of the model for the same turbopump equipped with a splitter-bladed impeller, whose geometry is parametrized in terms of the length and azimuthal position of the splitters. The two approaches consistently indicate that the introduction of the splitters reduces the static head of the machine as a consequence of viscous losses, flow distortions, and limited room for slip reduction in the original design. The two approaches also consistently indicate that optimum head performance is obtained at design flow from centrally mounted splitters, regardless of their length. The model predictions are in very satisfactory agreement with the computational results for both long and medium-length splitters. Alternate leading-edge stall of the full blades is manifest in the simulations, with the generation of relatively intense tip vortices. The observed moderate overestimation of the machine head by the model in the case of short splitters is thought to be the consequence of the dissipative interaction of such vortices with the leading edges of the shorter splitter blades, a phenomenon only imperfectly captured by the analytical model. The successful validation of the proposed model confirms that it represents a useful tool for the preliminary design and performance prediction of mixed-flow turbopumps with splitter-bladed impellers at a negligible fraction of the computational cost of full-fledged numerical simulations.