Project
Numerical Investigation of Gas Accumulation Mitigation by Flushing Holes in Radial Centrifugal Pumps
Hydraulic machines such as centrifugal pumps lose performance when gas accumulates in their decelerating, diverging flow passages, where the phases separate and coherent gas pockets form. This project investigates how such accumulations can be mitigated passively, using stationary surface structures, a solid step and cross streamwise bars, mounted on the upper wall of a horizontal, gradually diverging channel, following experiments at Otto von Guericke University Magdeburg. The channel is a controllable laboratory analogue of a pump passage, with lateral buoyancy emulating the Coriolis force driven phase separation of a rotating impeller.
While the experiments demonstrate the mitigation, computational fluid dynamics simulations can be used to investigate the responsible local mechanisms. To this end, two-dimensional simulations with a statistical turbulence model and three-dimensional large-eddy simulations are carried out. Large-eddy simulations resolve the large turbulent structures directly and therefore require a very fine computational grid. The multiphase flow is simulated with a hybrid Eulerian method, in which separate conservation equations for momentum and turbulence are solved for each phase. This is additionally coupled with a population balance model, which adds further conservation equations for the bubble sizes. Together with the long physical times required to reach statistical convergence, this makes the simulations computationally very expensive and only feasible on a High Performance Computer.
Project Details
Project term
July 8, 2025–July 7, 2026
Affiliations
Ruhr University Bochum
Institute
Chair of Hydraulic Fluid Machinery
Principal Investigator
Methods
The simulations were performed with a customized solver based on OpenFOAM version 9. The multiphase flow is described by a hybrid Eulerian method, which captures both dispersed bubbly flow and coherent gas structures within the same numerical framework, so that the transition between both flow regimes can be predicted. Separate conservation equations are solved for each phase, and the method is coupled with a population balance model that accounts for bubble coalescence and breakup.
For the two-dimensional parametric study, turbulence was modelled with the statistical k-omega shear stress transport model. The three-dimensional simulations were performed as large-eddy simulations, in which the large turbulent structures are resolved directly and only the small scales are modelled. A precursor duct simulation provided a fully developed, time resolved turbulent inflow.
Results
Two-dimensional study: The simulations reproduced the experimentally observed trends: the accumulation grows with increasing inlet air volume fraction and shrinks with increasing Reynolds number, and the mitigating effect of the solid step was captured well. The bars in their original position barely disperse the pocket, but when repositioned further upstream, their mitigation effect is recovered for all cases. Furthermore, the flow separation location differs between single-phase and multiphase flow conditions, with phase separation occurring upstream of flow separation.
Three-dimensional study: The three-dimensional simulations predicted the flow separation location in agreement with the experiments. They reproduced both the reference accumulation and its mitigation by the step and the bars, with wall pressure recovery within the experimental uncertainty band. The step raises both the turbulence kinetic energy and the dissipation rate, whereas the bars raise the turbulence kinetic energy but only marginally the dissipation rate, and bubble breakup is confined to the first few bars. In contrast to the step, the bars reduce accumulation yet lower pressure recovery. A persistent spanwise skewness of the accumulation was observed.
Discussion
The results show that air accumulation is mitigated only when turbulent mixing is intensified upstream of the phase separation location. A disturbance introduced beyond the onset of phase separation has no mitigating effect. This is the governing design principle of the whole project.
This principle also explains why the two-dimensional simulations initially did not capture the mitigation by the bars. The two-dimensional approximation predicts the single phase flow separation too far upstream, which displaces the phase separation point upstream as well. In their original position, the bars therefore lie downstream of the onset of phase separation. Only when the bars were shifted upstream accordingly was their mitigation effect recovered. In addition, the two-dimensional setup cannot represent the spanwise secondary flow, so the operating points are not directly comparable to the measurements. The two-dimensional study nevertheless served as an efficient screening tool.
The three-dimensional simulations identified bubble breakup as the controlling sub process. Because breakup is driven by the turbulence dissipation rate, raising the turbulence kinetic energy alone is insufficient, which explains the weaker effect of the bars compared to the step. Moreover, mitigation and pressure recovery are decoupled, because the bars introduce additional hydraulic losses. Air removal is thus necessary but not sufficient for a performance gain. The coalescence and breakup models were originally validated for pipe flow and bubble columns, not for diffuser flow, which limits the quantitative prediction of the bubble size distribution. Since inlet bias, grid asymmetry and temporal non convergence were excluded, the spanwise skewness of the accumulation is regarded as a genuine physical symmetry breaking.
In the project extension, this knowledge will be transferred to radial centrifugal pumps, where gas accumulates within the blade passages of the six-bladed impeller. There, flushing holes in the impeller blades will be investigated under rotating flow conditions.
Additional Project Information
DFG classification: 404-04 Hydraulic and Turbo Engines and Piston Engines
Software: Hybrid 2-Phase solver
Cluster: CLAIX