Hydraulic Institute Pump Intake CFD Modeling Methodology Evaluation

Introduction
It's not presently possible to analytically develop an optimum pump station design due to interaction of turbulent flows with combinations of facility geometry and pump options. Researchers and engineers have identified hydraulic conditions that can reduce pump performance including air entrainment, vortices, swirl, uneven velocity, and excessive turbulence in the approach flow to the pumps. These conditions can lead to fluctuating pump impeller loads, vibration, cavitation, loss of pump capacity, and reduced efficiency (Sweeney and Rockwell 1982). To facilitate design of wet wells and pumps that deliver required flow and head reliably, in 1998 the Hydraulic Institute (HI) established the American National Standard for Rotodynamic Pumps for Pump Intake Design in Standard 9.8 (ANSI/HI 9.8-1998), which is regularly updated (2018) and a 2025 update coming. The standard is normative, long relying on scale physical modeling to verify adequate pump intake hydraulics. ANSI/HI 9.8 includes detailed procedures and acceptance criteria for scale physical modeling to identify and eliminate problematic hydraulic conditions within intakes. Physical modeling standards were developed over many years and review cycles, and are well established after thousands of model studies.

Currently, computational fluid dynamics (CFD) modeling cannot be used to demonstrate compliance with this standard, which has been questioned during recent review cycles as this newer technology continues to develop. During update of the 2018 standard, the ANSI/HI 9.8 committee recognized the need for a more rigorous workplan to help determine if the role of CFD as it pertains to the intake design standard should be expanded.

Hypothesis
With sufficient guidance, different CFD modelers could produce similar results when modeling a specific problem.

Approach
The committee developed and conducted a round robin modeling experiment for testing the hypothesis.

The committee developed a workplan consisting of:
1. A literature review of past CFD work related pump intake design
2. Develop best practice guidelines (BPG)
3. Develop CFD model evaluation methodologies and reporting structure
4. Execution of blind round robin simulations to evaluate the methodologies and consistency of results.

Results
The literature review found many examples comparing scale physical and CFD models, and many solely focused on CFD. Constantinescu and Patel (1998) conducted the earliest work showing CFD models could replicate general flow fields observed in physical models.

The ERCOFTC/IAHR workshop 'TURBINE 99' (Gebart, et al. 1999) demonstrated that targeting a specific key parameter, pressure recovery in a study case, different modelers produced widely varied results, often using the same software and/or turbulence models. Following TURBINE 99, several BPGs were developed to reduce errors and uncertainties in CFD simulations all based on the ERCOFTAC (2000) BPG.

Model evaluation approaches were developed from the literature and committee experience, with 4 each for swirl calculation and velocity calculation. The methods were explained in a CFD Evaluation Methodology Document (CEMD).

The study focused on a closed conduit split case pump geometry test case from a past physical model study shown in Figure 1, using test conditions in Table 1. The modelers were provided with the geometry in 3D format, BPG, CEMD, and reporting template.

Eight HI member organizations simulated the test case. Seven organizations each submitted 1 set of results, and one organization submitted 10 sets of results from 3 different modelers using different platforms and turbulence models, leading to 17 sets of results. Overall, 5 different CFD solvers were used, including Fluent, OpenFOAM, CCM+, CFX, and CFD++. Turbulence was simulated using k-e variants, SST, and DES.

Figure 2 shows streamline results images colored by velocity magnitude from two of the modelers, which indicate vortices are unlikely.

Swirl analysis resulted in a wide range of values for all proposed measurement methods, with one method shown in Figure 3. Baseline results ranged from 0.1 degrees from axial to over 18 degrees and modified results ranged from 0.1 to 10.5 degrees. In general, most modelers found the modified geometry had lower swirl than the baseline geometry, although not all.

Results of one velocity calculation method plotted in Figure 4 show values ranging up to +/- 5% of average for baseline and modified conditions. Not all modelers provided results for this method. Most modelers reported a slightly more uniform velocity distribution with the modified geometry, but a few reported a slightly larger range.

Conclusions
Leading pump CFD modelers using identical best practices and reporting evaluated pump intake hydraulics to compare approaches and tools. The study found considerable variability between modelers results, with no clear trends relative to software, turbulence model, mesh resolution, steady and unsteady modeling. Based on the considerable variability, the ANSI/HI 9.8 Intake Committee determined that it will not expand the acceptable uses of CFD as it pertains to the standard for the next revision. However, the study provided valuable information and efforts will continue to determine if a CFD modeling approach and acceptance criteria can be developed such that CFD can be used to show compliance with the standard in the future.