While turbo pumps cover the full range of specific speeds, the overwhelming majority run at low specific speeds and are comprised of an inlet pipe or duct, an impeller and an exit volute. Recent design advances have been few since some producing companies do not have a single hydrodynamicist on their staffs.

By contrast, industrial process pumps, including boiler feed pumps, have a complex inlet, an impeller of any specific speed (and may require special treatment to deal with higher velocities), a diffuser and either a return channel or a crossover to deliver the flow to the next stage, in the case of the multistage machines. (Industrial process pumps include irrigation pumps and deep well pumps for home or business water supply.)

Progress toward assisting pump designers to deliver superior, efficient pump designs for these complex constructions must be predicated on the basis of possible design improvements, the tools to assist such work and the availability of necessary staff. Improving pump design is a worthy goal; according to The Freedonia Group, Inc., a Cleveland-based industry market research firm, "Global demand for fluid handling pumps is forecast to increase at a 4.4 percent annual rate to $47 billion in 2012." To address this challenge, we will look at what is needed technically and then what is becoming available.

Possibilities for Improvement

Figure 1 is a well-known plot from Stepanoff (1948, 1957), showing roughly the breakout of losses in various pumps. To achieve superior designs, we must mitigate these losses. Similar diagrams exist for radial compressors and turbines because they all have the same core issues involved: friction dominates at low specific speeds, and kinetic energy effects dominate at high specific speeds. (For instance, items 5a and 5b in Figure 1 are actually kinetic energy effects also found in other machinery classes.)

Estimated distrubution of loss elements in typical centrifugal pumps

Figure 1 shows that at low specific speeds, the process is dominated by friction (particularly disk friction) and, in some cases, leakage losses. On a direct assault, we will make little headway against such a tough challenge. At high specific speeds, the game changes: one must be careful to deal with kinetic energy effects (disk friction may be a tiny contributor), so we must treat all viscous processes with care and achieve excellent diffusion and overall control of the flow field. This is a different design challenge, and one for which superb tools exist.

Do not forget the middle ground in Figure 1 because the highest efficiencies are found there. Efficiencies can be several points higher than at nearby higher and lower specific speeds. A change in speed is needed to reach this middle ground, at least for many instances where both head and flow may be frozen.

Speed can be adapted quite easily today and carries a small price for gaining efficiency, a price often compensated by improved part load operations and savings. Our traditional reliance on simple synchronous motors needs to be updated, and, in some cases, greater care will have to be taken to avoid cavitation (see below). The good news is that higher rotational speeds lead to smaller diameter machines, which afford greater design opportunity. One of these opportunities is the potential to include a proper diffuser to achieve a more efficient stage, and while these are now used worldwide for high performance engineered pumps, they are not common for mass-produced pumps.

Figure 2 gives an example of a pump stage reported in the past in Holland.

A conventional clear water pump with an added diffuser cascade.

Now, with better design speed, some reduction in diameter and a diffuser included, there is the possibility of a performance gain-one that is worthy of real design optimization. Other possibilities exist as well, including the usage of high performance crossovers for multistage pumps, such as shown in Figure 3.

Vertical pump stage with impeller and continuous crossover, orderly streamlines 

 

These crossovers have been shown to have excellent diffusion and low losses, a fact that can be carried over to many other applications.

Tools to Improve Performance

Clearly, possibilities exist to achieve improved performance. The modern tools to improve the designs include Computational Fluid Dynamics (CFD) and three dimensional (3-D) finite element stress and modal analysis (FEA). Interestingly, however, these tools will not achieve much if we stay locked in the design world of the past and try to tweak low speed designs where disk friction will dominate. An inexperienced designer might pick up a point here or there with such tools, but a better designer probably does not leave such glitches in his designs in the first place.

When modern applications are considered with modern motors and controls, it is tempting to start into new designs with some of the available modern tools available. However, much has been learned from experience that should not be lost at this stage of the process.

Meanline design must come first so that intelligent choices are made for the optimal inlet velocity triangles, both with respect to efficiency and cavitation avoidance. Likewise, the exit velocity triangles must be carefully selected and matched to the diffusers downstream, a process that has been handled better in the compressor industry than in the pump industry. Therefore, we would be well-advised to borrow a few pages from compressor design experience in this particular case. One should never miss the opportunity to sort out issues at the meanline level; otherwise, one pays for such oversights all the way through the process and never fixes the damage done.

For the past 50 years, designers have learned how to develop good stages without CFD, and better ones today aided by CFD. Using conventional blade shaping techniques with throughflow solvers permits clear exercise of the skills and results of the past five decades.

Optimization is key for competitive success today. The skilled practitioner of modern design can explore from a stress (FEA) and a performance (CFD) basis and has several alternative codes available in each area. At this point, one considers the full impact of open versus closed impellers on the details of leakage, thrust and optimum blade design. One looks at the process of design for minimum weight and stress to maximize life and lower cost but keep best performance. Material selection enters in, including even plastics (which are rather common today in many pump applications). In addition, the choice of metals, when required, is made with greater confidence.

CFD is based on a full 3-D solution of the Navier-Stokes equations, which are the full viscous flow equations governing the processes of interest. FEA is the full 3-D solution for the solid body forces and the body response to the operation of the hardware. Figure 4 illustrates some of the computational possibilities.

The use of CDF (left) and FEA (right) permits welcomed percision in the design

It is only recently that we could even dream of possible solutions for these equations, and only the last few years have given us fast, powerful computers for individual usage with these systems. We must, therefore, be careful of shortcomings that might accompany these powerful but new tools.

Conclusion

Even though we use billions of pumps today, there will be a large increase in the number of pumps in use worldwide, and we need all of the efficiency that we can get to minimize the power usage. This will be achieved by a combination of historical experience and powerful new tools such as CFD and FEA solvers.

Pumps & Systems, February 2010