Sizing, selecting pumps and circulators

By making correct pump and piping selections, a hydronic system can operate at or near peak efficiency, avoiding cavitation and vibration.
By Tim Chadwick, PE, LEED AP, Alfa Tech, San Jose, Calif. November 12, 2012

Figure 1: Facebook’s innovative Prineville (Ore.) Building 1 data center has been operational for almost 2 years, establishing a high level benchmark in data center energy efficiency (see The first building included using reverse osmThe vast majority of HVAC and plumbing hydronic systems use centrifugal pumps to circulate fluids. The centrifugal pump was invented in 1689 by Denis Papin. The general principle behind the pump operation is fluids are led to the impeller hub and are then “flung” to the perimeter of the pump casing (or volute) by centrifugal force. The first centrifugal pumps were constructed with straight vanes, and the current curved blade version was invented in 1851 by John Appold. 

The most common centrifugal pumping arrangements are long- and close-coupled pumps. These terms refer to how the motor and pump impeller are connected. Long-coupled pumps include a coupling (or a series of couplings) between the motor and impeller while close-coupled pumps are directly connected. In general, maintenance and operations are simplified in a long-coupled pump arrangement, although the coupling requires proper alignment and balancing of the system to avoid vibration.

With long-coupled pumps, motor replacement can be completed without affecting the pump, and pump bearing service can be completed without disturbing the motor. In addition, motor bearings on long-coupled pumps are better protected from pump seal leaks due to the separation of pump and motor. Close-coupled pumps are generally cheaper to install but more difficult to maintain as noted above. In most pumping arrangements, close-coupled pumps are used for small applications (generally flows up to 500 gpm) or where first costs are the primary decision driver; long-coupled pumps are employed for larger scale applications where operation and maintenance concerns drive the decision process. 

Pump curves 

Figure 2: A typical pump family of curves shows pump performance (flow versus pressure) for a family of impeller sizes within a pump casing and the associated pump horsepower requirements and net positive suction head required (NPSHr). Courtesy: Alfa TechEquipment suppliers publish a set of pump curves for each of their product offerings that describe the operating characteristics of their pumps. A sample pump curve is included in Figure 2. Flow, head, required power input (motor selection), efficiency, and net positive suction head required (NPSHr) are generally graphed on pump curves. A typical pump curve is created for a single pump casing size, and the performance of various impellers within that casing are graphed on the same chart. In addition, the piping system or “system curve” can be graphed by the engineer onto these pump curves to determine how the system will perform with the associated pump. The system curve can be established by confirming the system pressure drop related to increased flow. Each published pump curve is related to a particular pump speed (rpm) that should be published with the curve. In addition, the pump curve relates performance of a fluid at a specific temperature and viscosity. When selecting pumps for fluids other than water (such as glycol mixes), correction factors must be applied to confirm the actual pump capabilities. 

For a particular pump and impeller size combination to produce a specified flow rate, the system pressure (or back-pressure created by the piping system) typically indicated as total dynamic head (TDH) must match the pump curve. As an example, in Figure 2, with an 8.5-in. impeller size, to achieve 600 gpm flow, the piping system must include 69 ft of pressure loss. If the system has less loss then the flow would be higher, and if the system had more loss then the flow would be lower as it follows the 8.5-in. curve. At this 600 gpm flow condition, the efficiency is 77%, the required pump hp is 14 hp, and the NPSHr is 6. 

Impeller sizing 

Within each pump casing, a variety of impeller sizes can be fit to produce a range of flow or pressure characteristics. In Figure 2, pump curves for impellers of 7 to 10.5 in. are shown. These are nominal impeller sizes, although the final installed impeller can be factory or field trimmed to any specific dimension to meet performance conditions in between these published curve values. 

A variable frequency drive (VFD) can control motor rotational speeds by varying the motor input frequency and voltage to achieve a desired motor speed. Adding a VFD to a fixed impeller size acts like allowing for infinite trimming of the impeller without removing any of the impeller metal, hence the attraction of a VFD not only for energy savings, but also for improved balancing and tuning of a system’s performance. 

There are countless reasons why pump and piping systems get oversized, including where the design engineer is applying too much of a safety factor and other system conditions that result through no fault of the engineer. In general, pumping systems are usually sized for the “peak” flow rate condition required by a heating or domestic water system; however, systems rarely operate at the peak conditions due to the nature of building performance and human interface. Engineers should strive to take advantage of this system oversizing and gain the most energy efficiency they can by applying a VFD to each pump system. Some oversizing can result from design safety factors and nominal equipment sizing. To ensure the peak flow can be achieved, an engineer will typically pick a system that meets the flow conditions, with a small cushion. If the system is installed with the extra capacity, the system must be balanced to the project specific requirements or additional energy is wasted or performance will suffer. 

Figure 3: A typical triple-duty valve is used to prevent backflow, balance the pump flow and pressure, and provide isolation. Courtesy: Bell & GossettTo control the oversizing of a pump system and regulate pressure delivery from the pump, use a pressure reducing valve (PRV). The PRV will “throttle” the pump and is typically located at the pump discharge. It is quite common to employ a “triple-duty valve” (see Figure 3) at the pump discharge that allows for isolation, balancing, and check valve functions in one valve assembly (hence the term “triple-duty”). In addition, most triple-duty valves include a calibrated pressure drop such that flow through the valve can be confirmed by measuring the pressure drop through the valve. To achieve the desired system flow rate, the balancing valve is partially closed, introducing an artificial loss in the system, and so the pump “rides” up the pump curve to the desired flow rate. This wastes energy due to excessive pressurization of the hydronic system and losses at the balancing valves. Specifying the design flow to be set by trimming the pump impeller can significantly reduce these losses. At a minimum, all projects should include requirements for proper trimming of impellers to meet the actual project conditions if VFDs are not included. 

To improve performance, a VFD can be employed to allow for “on the fly” impeller trimming to create the desired flow conditions for a given system pressure drop. If a VFD is employed, the impeller can be “trimmed” by simply reducing the pump motor speed with the same size impeller in place. According to the pump affinity laws, a reduction in pump speed also translates to a cubic reduction in motor horsepower, so when flow is reduced by half, the brake horsepower is reduced by 88%. It is clear why the initial cost of a VFD can create a rapid return on investment in energy savings. ASHRAE 90.1 and other energy standards generally require variable flow systems on larger pumping systems. With the reduced cost of VFDs, it makes sense to use the technology on almost all pumping systems and, in many cases, even if variable flow is not part of the system control strategy. 

It should be noted that with the use of VFDs, triple-duty valves can still be used (to gain the benefits of the check valve, isolation valve, and flow measurement), but project specifications should ensure that the triple-duty valve is left 100% open and the VFD is used for balancing to avoid any false loads on the pumping system and the associated waste of energy. 

These valves represent a major maintenance challenge, and so the “triple-duty” function can also be achieved by installing three separate valves. One of the main benefits of the integral triple-duty valve lies in its ability to measure flow through its flow calibration; this capability is lost if individual valves are used. Another disadvantage to splitting the triple-duty valve function is that the isolation valve typically becomes a butterfly valve in larger pipe size applications, and butterfly valves can make accurate balancing a difficult operation as their performance is far from linear. One final issue that can arise with installing multiple valves in lieu of the triple-duty valve is higher labor costs to install three valves instead of one. 

Motor sizing and speeds 

In general, pumps are available with 1,150, 1,750, and 3,500 rpm motors. You may have heard recommendations that you should select 1,750 rpm pumps whenever possible, but why is that? A pump’s critical speed is the rpm at which shaft vibration increases dramatically. Typical critical speeds greatly exceed 1,750 rpm for most pumps; however, critical speeds can be near the 3,500 rpm motor speeds. In addition, 3,500 rpm pumps are noisier than comparable 1,750 rpm pumps. Theoretically, 3,500 rpm pumps are more efficient that 1,750 rpm pumps, but the efficiency gains are generally very small (5% to 10%) and may not outweigh the potential for premature failures due to vibrations and the increased noise associated with higher rpm pumps. In addition, maintenance costs can increase with higher speed pumps as seal faces generally wear out eight times faster for every doubling of rpm due to the higher heat involved. 

Another consideration when selecting pump motors for a particular pump is pump run-out (overloading of the motor). For the example flow condition of 600 gpm and 69 ft of head shown in Figure 2, motor load was 14 hp. In general, this would indicate that a pump provided with a 15 hp motor would meet the project requirements. However, consider if that pump were installed in a system where the initial pressure drop was only 50 ft. As discussed previously, the balancer could close a balancing valve and create an artificial system pressure loss of 69 ft to match the pump curve, but this wastes energy. 

Reviewing Figure 2, if the system pressure loss was only 50 ft, the actual flow through the system would be 940 gpm and the required motor horsepower would be 16 hp, so the 15 hp motor would be overloaded and the motor safeties or electrical system would trip to protect the motor. To avoid this situation, pumps are often selected with “non-overloading” motors or, in this example, a 20 hp motor.

When designing your hydronic system, you must pay close attention to the possibility for motor overloads and determine under what scenarios the motor might become overloaded. The alternate of always selecting non-overloading motors can add significant cost to larger pumping systems when you consider the upsizing of the motor, disconnects, VFD, wiring, and the rest of the electrical distribution system that can result. 

Net positive suction head 

In open pumping systems such as open cooling towers, an engineer must consider the net positive suction head of the pumping system. When a pump is placed into a situation where its NPSHr is not met by the system, cavitation can result. Cavitation occurs when pumped liquid is pulled down below the liquid vapor pressure, and so the pumped liquid vaporizes or flashes within the pump. When these bubbles move through the impeller and reach a higher pressure, they will collapse and can cause severe damage to the pump including impeller erosion, broken seals, or bearing failures. 

An engineer must calculate the net positive suction head available (NPSHa) by determining the ambient air pressure, the water level above the pump, losses in piping, strainers, and other devices between the open air surface and the pump inlet. Published pump curves include a NPSHr curve (Figure 2). As part of a pump selection process, ensure your system design (including any alternate construction methods that may be proposed during construction) includes sufficient NPSHa to avoid a cavitation problem. Pump selection software and catalogs include detailed procedures for determining NPSHa, but the general formula is:

NPSHa = hp + hz – hvpa – hf 


hp = atmospheric head for elevation of installation

hz = static elevation of liquid above center line of pump (negative if pump is above the water level)

hvpa = absolute vapor pressure at pumping temperature

hf = friction losses in pump suction piping 

At each flow condition (gpm) on the pump curve there is a corresponding required amount of NPSHr to prevent cavitation. In Figure 2 the NPSHr is 6, so the pump installation (location, elevations, water level, etc.) must be such that a minimum NPSHa of 6 is provided.

In our experience, one of the most common causes of cavitation is a clogged strainer placed at a pump’s suction inlet. In our engineering practice, unless the water being pumped has the potential to contain large, solid debris, we prefer to remove strainers or place strainers downstream of the pump. Pump manufacturers’ literature generally states that most pumps can pass relatively large diameter debris without damaging the pump. By relocating the strainer (or removing it entirely), we have drastically reduced the potential for cavitation at the pump, which can cause significant damage to a pump system. In cooling tower applications, most cooling towers can be equipped with basin strainers that remove large debris and provide a further measure of protection for the pump and piping system.

In open systems, it is generally advisable to install pumps below the minimum water level of the system to aid in NPSHa and to ensure the pump volute is flooded during pump start-up, preventing possible damage to the pump seals if the pump were to operate dry. In addition, ensure sufficient water is available at the pump inlet to ensure air is not entrained into the pump inlet, which can occur when strong pump suction creates a vortex allowing air into the pump inlet.

Performance of pumps in parallel 

Figure 4: In this flow diagram, two pumps are connected in parallel with triple-duty valves at pump discharge. Courtesy: Alfa TechMany hydronic system designs require connecting pumps in parallel. In a parallel pumping installation, the system pressure remains constant through the parallel pumps, while the system flow (or gpm) is additive. Domestic booster pump systems, heating hot water, chilled water, and condenser water systems are typical examples where parallel pumping is employed. Pumps connected in parallel often are used when the required flow is higher than one single pump can supply or when the system has variable flow requirements, which are met by switching parallel connected pumps on and off. 

The advantages of parallel pumping arrangements include redundancy and in many cases higher overall system efficiency. Typically, pumps connected in parallel are of the same type and size, although nonmatching pumps can be connected in parallel with varying speed control strategies to provide a measure of variable flow control. When operating two pumps in parallel with speed controls, it should be noted that many flow conditions could be met by running one pump at full speed or by running multiple pumps at a reduced speed. However, the total efficiency is generally higher when two pumps run at reduced speeds.  

Performance of pumps in series

When your system arrangement dictates connecting pumps in series, the flow of the system remains constant through the pumps, while the system pressure (or head) is additive. Typical booster pump arrangements employ pumps in series to “boost” the system pressure as needed to overcome some pressure resistance in the system or to bring water services to the upper floors of a high-rise building. A multi-stage pump is essentially a group of pumps in series, contained within one pump assembly, allowing for a relatively low flow and high head system performance.

Pump mounting

Just as there is a wide variety of pump types and applications, each style can be mounted in a variety of ways, depending on your project requirements. In-line pumps are typically installed directly ‘in line” with the piping and can be supported from the pipe or with separate supports for larger systems. Long- and close-coupled pumps are generally mounted directly to the slab, although in some cases where vibration is a major concern, pumps can be installed on spring isolators. In our experience, however, a properly balanced pump does not cause a significant vibration.

Large pump systems may be installed on inertia bases, which typically consist of a concrete-filled frame mounted on spring isolators. Inertia bases provide vibration and start-up torque dampening by adding mass and lowering the center of gravity of the installed pump, although inertia bases add significant weight and cost to an installation, and therefore generally are not required for most applications. 

Figure 5: This shows a chilled water pump (fully insulated piping and pump casing). Note the flexible connections for the electrical wiring to the motor and for the copper drain piping in the foreground. It is a long-coupled pump with inertia base mountinFlexible connectors are provided on vibration-isolated equipment to prevent vibrations from being conveyed through the piping system. Consult with the local authority having jurisdiction (AHJ) about these options as well as other piping products. Expansion joints also can reduce mechanical strain in connection with any shock waves that may be conveyed through the piping system, such as water hammer. Expansion joints can help absorb thermal expansion and contraction related to fluid temperature changes. It is highly recommended to install a pressure gauge across the inlet and outlet taps of a pump. By connecting a single gauge to the inlet and outlet (with isolation valves), operations and maintenance staff can measure the inlet and outlet pressure of the operating system and calculate the pump differential pressure. By utilizing a single gauge, calibration concerns are minimized as any errors in calibration affect both measurements and thus are offset. By confirming the differential pressure and the impeller size (or VFD speed), system flow can be determined from the pump curve.

Tim Chadwick is president of Alfa Tech. His expertise is energy-efficient design of HVAC and plumbing systems for commercial and industrial projects. He sits on the Data Center Environmental Performance Criteria (EPC) Working Group Panel and was past EA Committee Chair.