Computer-aided engineering enables the improvement of pump hydraulic performance in power plants.

Cooling water pumps for power plant condenser cooling systems are large flow-rate, low- to medium-head units that operate at low rotational speeds. For these pumps’ parameters, a typical hydraulic design stage configuration, optimal from the efficiency point of view, is either a mixed-flow or axial-flow single stage unit. 

These pumps are developed and delivered by several leading pump manufacturing companies and are operational worldwide in both fossil-fuel and nuclear power stations (in the latter case, in a secondary cooling loop).

Reference 1 and Reference 2 outline pump hydraulic design procedures which are applicable to develop hydraulic shapes of both the impeller and vane diffuser, which are typical components of these pumps, shown in Figure. 1A, while Reference 3 gives a good example of the development and construction of the cooling water pump for Cumberland power station in the U.S.

The use of computer-aided design (CAD) based geometry modellers in the industry (see Reference 4) and computational fluid dynamics (CFD) (see Reference 5 and Reference 6) allows for the definition and optimization of the shapes of the through-flow channels and blades. Designers arrive at design solutions that have improved performance features, when compared to machines designed using earlier and, to a large extent, test-based methods outlined in References 1 and 2. 

The pump performance and the design systems’ features, which become increasingly important are:

  • The desired position of the best efficiency point (BEP)—BEP parameters prediction
  • The shape of performance characteristics, such as H = f(Q), η = f(Q)—overall performance prediction
  • Slope of the H = f(Q) characteristics—CFD prediction of 
instability regions
  • Achievable efficiency at the BEP and shape of the efficiency = f(Q) characteristics
  • Net positive suction head (NPSH) performance—CFD prediction of cavitation
  • Hydrodynamic, including vibration performance—CFD studies of stage components’ interaction

This article addresses some technical capabilities of computerized, pump design and redesign systems, which are now in use in performance upgrade projects. It also discusses how these technologies’ methods allow for the analysis of existing components even under conditions with limited access to the original part’s geometrical design data, such as what is available from the existing pump hardware.

Figure 1. Case 1 pump's digital design model

Computer Aided Engineering

To satisfy the operational requirements of cooling water pumps and ensure that the high efficiency values of these pumps, their hydraulic layout results in the design solutions in the high specific speed mixed-flow and axial-flow ranges. Early generations of these pumps were designed using basic engineering approaches, which combined one-dimensional modeling of flow properties with the empirically derived design factors and coefficients. 

The inaccuracy of these methods and the limited validity range of the empirically derived design factors had to be compensated for by the cost- and time-intensive test-bed optimization effort at the manufacturers’ sites to achieve performance goals and meet project specifications. 

The development of advanced, computer-aided engineering (CAE) and, particularly, the numerical flow simulation technologies makes it possible for manufacturers to further develop the existing pump ranges and optimize them for each application, including the design solutions for large capacity water cooling pumps. 

Original equipment manufacturers (OEMs) now use   CAE technologies when designing new pumps. In addition, specialized pump engineering service companies, their consulting engineering groups and/or external partners are able to further enhance the performance of existing, already installed pumps that were, until recently, considered state-of-the-art solutions in terms of efficiency, cavitation and reliability. 

A prime example of such a redesign effort is a completed project aimed at upgrading the Edmonston pump station (Tehachapi) units, (see Reference 7). The key required capabilities of technologies used in the redesign and performance improvement project were:

  • The quality of the CFD (numerical flow prediction) used to support the redesign process
  • The capability of the geometric modellers to represent and interactively modify the pump components and pump stage geometries
  • The balance between the computer-based design/optimization effort and the model testing, which is the ultimate source of information on the actual achieved overall performance of the developed design solution

Figure 2. Case 1 pump's CFD solutions at BEP


Figure 3. Case 1 pump's CFD solutions at the partial-load operation

Pump Research Data and CFD Validation

Research on the pumps’ internal flow patterns at design and off-design conditions—such as those reported in Reference 8 and Reference 9—provides pump designers with knowledge of the flow phenomena, which can be correlated to the measured performance. For the design process, the flow simulation technology must employ tools that are capable of predicting measured performance curve and, flow conditions and phenomena during the design phase, before the pump is manufactured and tested as a physical prototype. 

A study was conducted, to validate a CFD program when applied to a real, mixed-flow pump stage, with the impeller and diffuser assembly typical for large capacity cooling water pumps with the specific speed of 6,350 (U.S. units). 

Pump Case 1 based on Reference 8 data and shown in Figure 1, was selected for this analysis. It represented a research pump designed by the National Engineering Laboratory (NEL), East Kilbride, Scotland and experimentally investigated at the Strathclyde University, Glasgow. 

This research resulted in the pump’s overall performance characteristics (for example, Q-H and Qη curves) as well as an assembly of data on the internal velocity distributions and flow patterns which, using the Laser Doppler Anemometry (LDA), were measured at several flow rates. A distinct performance feature of this stage is its Q-H curve instability at flows below Q = 0.74 QBEP, (see Figure 4). 

The cross-sectional drawing, showing the pump assembly, is shown in Figure 1A. The data given in Reference 7, providing a complete geometrical definition of all the components in terms of their surfaces points (x, y, z coordinates), were used to develop the digital design model, which is the integral part of a design package applicable for pump and turbomachinery design. 

Both the meridional flow path, Figure 1B, and the fully 3D digital model of the pump stage, Figure 1C, were created to enable the performance prediction and the flow analysis process. Figure 2A shows the grid of H-type which was created for this case using a CFD mesh generating code to facilitate the Navier-Stokes CFD solution on this mesh. 

Figures 2B and 2C demonstrate sample outputs of the CFD results in terms of the flow-field and the velocity distribution in the rotating and non-rotating stage components at BEP conditions.

Considering the off-design operation, Figure 3A shows regions of flow recirculation in the inlet part of the impeller passage—identified by LDA—located at the upper part of the through-flow channel, close to the outer shroud contour. The flow recirculation was captured at Q = 0.46 QBEP

The experimental findings of Figure 3A were compated with the CFD solutions (see Figure 3B) derived for the same geometry and flow conditions in the instability region (which are also shown as the superposition onto the experimental data). The CFD was able to correctly capture the complex internal flow recirculation pattern as far as its location, size and direction at these partial load conditions. 

For Pump Case 1 (NEL), the measured global pump characteristic, Q-H, Figure 4, showed a region of positive slope (partial-load instability) which was correctly predicted using CFD, including the point of stalled conditions. Figure 4 shows raw CFD results extracted from the Navier-Stokes solution that used mixing plane for the impeller and vane diffuser assembly CFD analysis. 

No corrections were applied, due to the inherent loss components while generating Figure 4, which explains the discrepancy between the head values measured (as tested) and those numerically obtained (CFD).

Figure 4. Case 1 pump's Q-H performance—CFD vs. tested

Geometrical Modeling

Apart from CFD simulation-based, hydraulic performance prediction capability, the design/redesign CAE approach also required the use of suitable tools, allowing the build-up of the digital design model of all relevant components. Pump performance upgrades and retrofitting projects can be effectively supported by the development of the flow-simulation solution  for the existing (baseline) geometry as the first step. 

It is not a must, but it is a good thing to have  to make the right redesign decisions during the pump’s or its components’ optimization process. A dedicated, one-dimensional modeller, coupled to the three-dimensional software package can facilitate the process of generating a useful digital design model from a limited subset of the available geometrical parameters that are derived from the existing machine geometry. 

Pump Case 2, selected for the validation of this process involved a mixed-flow pump impeller (specific speed 3,800, U.S. units) shown in Figure 5A, meridional cross-section. The data available were limited to the normally accessible inlet and outlet portions of each component, at the leading and trailing edge locations of the impeller, and were used as the input to build the 3D, digital design model that is shown in Figure 5B. 

The exact input data were: 

  • The sketch of the pump stage, with  the diameter values
  • Through-flow channel widths
  • The blade angles at the leading and trailing edge locations and at the three streamlines (tip, mean and hub) 

In addition, the vane thickness, number of blades and the blade lengths, were known. This subset of geometric parameters is normally obtainable from the pump hardware components geometry, once access to them is given during the inspection or repair process. These data were used to provide the input to a one-dimensional modeller with its output transferred to the 3D geometry modeling process, the result of which is given in Figure 5B. 

A three-dimensional CFD flow analysis using the solver, which was also used for Case 1 of this article and described above, followed and enabled the generation of the performance curve shown in Figure 6. 

A good agreement was reached between the performance curves of the tested and digitally-generated pump and impeller configurations. Particularly, the CFD prediction considered the impeller only, while the tested pump curve represents the impeller plus vane diffuser stage configuration. 

Figure 5. Case 2 pump's impeller digital design model

This analysis confirmed that such a generated digital design model, even of an isolated component of the pump stage such as the impeller, may provide a good geometry starting point. 

Also, the CFD initial solution, for the optimization process in the case that only a limited, one-dimensional level of the geometrical data was available from the machine that was to be improved was a good starting point. 

It is important to note that the Case 2 pump had experimentally derived Q-H characteristics that were a continuously negative slope curve, with no occurrence of instability over the entire range.

Conclusion

Two important features and technical capabilities of the CAE systems applied for the performance analysis and upgrade projects and were discussed in this article can be considered as critical to any pump performance upgrade project:

  • Pump digital design model generation capability, based on the obtainable reduced 1D subset of the data in case no fully 3D geometry definition is available
  • Ability to numerically predict experimentally verifiable flow patterns, and the overall performance curves over the entire operational range, including the performance curve instability

These features particularly apply to pump redesign tasks, when the gains in efficiency and/or the improvement or adjustment of the shape of the performance characteristics are the project’s objectives.

Figure 6. Case 2 pump's Q-H performance: CFD vs. tested

References:

1. Pfleiderer, C. Kreiselpumpen fuer Fluessigkeiten und Gaese”, 5th ed., Springer, Berlin, 1961.

2. Stepanoff, A. Axial and Radial Flow Pumps, 2nd ed., Wiley & Sons, New York, 1957.

3. Rosemann, P. “Special Technical Problems of a Large Pump Project,” (in German), KSB Technical Report no. 16, 1973.

4. Favre, J.N. “Development of a Tool to Reduce the Design Time and to Improve the Radial or Mixed-Flow Pump Impeller Performance,” ASME FED-Vol.222, 1995. 

5. Denus, C.K., et al. “A Study in Design and CFD Analysis of a Mixed-Flow Pump Impeller,” ASME FEDSM 1999, San Francisco, Calif., 1999.

6. Denus, C.K., et al. “Hydraulic Development of a Centrifugal Pump Impeller Using the AGILE Turbomachinery Design System,” TASK Quarterly 6, No.1, 2002.

7. Nagahara, T. et al. “Investigation of the Flow Field in a Multistage Pump by Using LES,” ASME FEDSM2005, Paper no. 77319, Houston, Texas, 2005.

8. Carey, C. et al. “Studies of the Flow of Air in a Model Mixed-Flow Pump by LDA,” NEL Rep. 698 and NEL Rep. 699, East Kilbride, 1985.

9. Goto, A. “Study of Internal Flows in a Mixed-Flow Pump Impeller at Various Tip Clearances Using 3D Viscous Flow Computations,” ASME Paper 90-GT-36, 1990.

Pumps & Systems, May 2012