Chilled Water Pump Guide: Types, Design & Operation
A chiller can be perfectly selected and a building still fail to cool, because the chiller only makes cold water available. The pumps decide whether it actually arrives. For a full picture of how the plant is configured before pump selection begins, see our guide to HVAC chilled water systems.They run more hours than almost anything else in the plant, they consume a serious share of its energy, and their design is one of the few places where a junior engineer's arithmetic is visible in the electricity bill for twenty years.
This guide covers the three pump types you will meet in a chiller plant, how the loop behaves in operation, what a decoupler actually does and why its flow direction is the best diagnostic in the plant room, the inputs you need before you can size anything, how to read a manufacturer curve, and how all of it lands on a pump schedule that a contractor can price.
A chilled water pump room with duty and standby sets on inertia bases. Note the suction guides, flexible connectors and gauge points: every one of those is a line on the pump schedule. [REPLACE with your own project photograph or a licensed image.]
- TL;DR
- Types of Chilled Water Pumps
- Pumping Configurations Compared
- Chilled Water System Operation
- Decoupler Operation
- Live Decoupler Flow Simulator
- Pump Design: System Overview
- Selecting Pump Design Inputs
- Pump Sizing & Schedule Generator
- System Volume & Minimum Loop Volume
- Expansion Tank & Fill Pressure
- Understanding Pump Performance Outputs
- The Chilled Water Pump Schedule
- FAQs
- Sources & Further Reading
TL;DR
Key takeaways
- Three pump roles: primary pumps move water through the chiller evaporator at constant flow and low head; secondary pumps move it out to the coils at variable flow and much higher head; condenser water pumps serve the tower loop.
- A chilled water loop is closed, so the pump overcomes friction only. Building static height cancels out and must not be added to pump head.
- A decoupler is a short unvalved pipe linking supply and return headers. It hydraulically separates the primary and secondary loops so neither pump set fights the other.
- Decoupler flow direction is the best diagnostic in the plant. Supply toward return means primary surplus, which is normal. Return toward supply means deficit flow, which is a fault.
- In deficit, warm return water mixes into the supply and raises the temperature reaching the coils above setpoint. Coils then cannot meet duty no matter how far their valves open.
- Pump head is built up along the index circuit only: coil, control valve, balancing valve, strainer, pipe friction and fittings, plus the evaporator for a primary pump.
- Select at the duty point on the manufacturer curve, keep it near best efficiency point, check NPSH, then size the motor above end of curve power.
- Everything lands on the pump schedule, the single document contractors price from and procurement orders against.
Types of Chilled Water Pumps
Three roles, distinguished not by the machine but by which circuit it serves. The same physical pump model can be any of them; what changes is the flow, the head and the control regime.
| Factor | Primary (evaporator) | Secondary (distribution) | Condenser water |
|---|---|---|---|
| Circuit served | Chiller evaporator and primary header | Building distribution to coils and back | Chiller condenser to cooling tower |
| Loop type | Closed | Closed | Open |
| Flow regime | Constant, one pump per chiller | Variable, tracks building load | Usually constant, one per chiller |
| Typical head | Low, often 10 to 15 m | High, driven by distribution length | Moderate, plus static lift to tower |
| Static lift in head? | No, closed loop | No, closed loop | Yes, open tower basin |
| Control | Fixed speed, staged with chiller | VFD on differential pressure | Fixed speed, staged with chiller |
| Water quality | Treated, closed, stable | Treated, closed, stable | Open, fouling and corrosion prone |
| Common type | End suction or inline | End suction or split case | End suction or split case |
The one difference that catches people out
Chilled water loops are closed: the water that rises up a riser also comes back down, so the elevation terms cancel and the pump only fights friction. A condenser water loop is open at the cooling tower basin, so the pump genuinely has to lift water to the tower and that static lift is real head that must be added. Treat the two the same way and you will either oversize the chilled water pump by the height of the building or undersize the condenser pump by the height of the tower. Both happen.
Which Pumps a Configuration Needs
The loop arrangement determines the pump count, not the other way round. A primary only system has one set of chilled water pumps. A primary secondary system has two sets, hydraulically separated. A variable primary flow system goes back to one set but adds control complexity. That choice is the subject of the next section.
Pumping Configurations Compared
Select a tab to see how each arrangement is put together.
Figure 1: The three chilled water pumping configurations. Count the pump sets and note where the flow is allowed to vary in each.
Chilled Water System Operation
Follow one parcel of water around a primary secondary loop and the pump's job becomes obvious.
-
Primary pump draws from the return header
It takes warm return water at roughly 12°C and pushes it through the chiller evaporator. Its head is small because its circuit is short: the evaporator, a strainer, an isolating valve, a check valve and a few metres of header pipe.
-
The evaporator removes heat
Water leaves at design supply temperature, typically around 7°C, and enters the supply header. The primary pump's job is finished at this point.
-
Secondary pump takes over
It draws from the supply header and pushes water out into the building against the full resistance of the distribution: risers, branches, coils, control valves, balancing devices and every fitting on the index run. This is where most of the pumping energy is spent.
-
Coils absorb heat, valves modulate flow
Two way control valves throttle flow to each coil as its zone load falls. Total secondary flow therefore falls with building load, and the VFD reduces pump speed to hold the differential pressure setpoint. Because pump power varies with roughly the cube of speed, this is where the savings come from.
-
Water returns and the difference goes through the decoupler
Warmed water returns to the return header. Whatever difference exists between primary flow and secondary flow passes through the decoupler, in one direction or the other. That direction is the subject of the next section, and it is the most useful thing in the plant room.
Decoupler Operation
A decoupler, also called a common pipe or bypass, is a short length of pipe connecting the supply and return headers between the two loops. It has no valve, no pump and deliberately almost no resistance.
Its purpose is hydraulic separation. Without it, the primary and secondary pumps would be in series and each would see the other's resistance, so changing one loop's flow would disturb the other. With it, each pump set sees only its own circuit, and the two loops can run at completely different flow rates without fighting.
Three rules for a decoupler that actually works
Keep it short and fat. Its resistance must be negligible compared with any other path, otherwise it stops decoupling. Never put a valve in it. A balancing valve or check valve in the decoupler defeats its entire purpose and reintroduces the coupling you were trying to remove. Locate it correctly, close to the headers, so the primary and secondary connections are properly separated and flow can pass either way without restriction.
Reading the Flow Direction
Because the decoupler has no valve, flow finds its own direction, and that direction tells you the state of the whole plant.
- Primary flow greater than secondary flow. The surplus flows from the supply header toward the return header. Chilled water at supply temperature bypasses the building and mixes into the return, so chiller return temperature drops and chiller delta T falls. Coils still get water at full design supply temperature. This is the normal, intended condition.
- Flows equal. No flow in the decoupler. Momentary and unstable in practice.
- Secondary flow greater than primary flow. Deficit flow. Flow reverses, travelling from the return header into the supply header. Warm return water mixes into the chilled water supply, so the temperature reaching the coils rises above setpoint. This is a fault condition.
- Qp total primary flow (all running chillers)
- Qs total secondary flow to the building
- Tchws chiller leaving water temperature
- Tret building return water temperature
Live Decoupler Flow Simulator
Move the two sliders and watch the decoupler. This is the single most useful mental model in chiller plant operation.
The failure loop worth recognising on sight
Deficit flow is self reinforcing, which is what makes it dangerous. Coils receive water warmer than setpoint, so they cannot meet their duty. Their control valves open further trying to compensate. That increases secondary flow, which deepens the deficit, which raises the supply temperature further. Operators watching zone temperatures rise often respond by speeding up the secondary pumps, which is exactly the wrong move. The only correct response is to add primary flow by staging on another chiller. Set the sliders to a deficit above and watch the supply temperature climb as you increase secondary flow alone.
Pump Design: System Overview
The design sequence is short and strictly ordered. Each step depends on the previous one, which is why sizing a pump before the coil selections are settled produces a schedule you will have to reissue.
- Establish flow
From cooling load and design delta T, per pump role.
- Identify the index circuit
The hydraulic path with the greatest total resistance, which is not always the longest.
- Build up head
Sum every loss along that path, and only that path.
- Select from the curve
Duty point near best efficiency point, NPSH verified.
- Size the motor and issue the schedule
Above end of curve power, then document everything.
The Pump Room
Worth settling early, because pump rooms are routinely allocated by whatever space is left over and then have to accommodate machines nobody measured.
- Space and access. Clearance to withdraw the rotating element or lift the casing, room to pull a motor, and a route to get a replacement pump in and a failed one out without demolishing a wall. Back pull out designs help but they still need the clearance behind.
- Duty and standby. Chilled water pumps are almost always installed as duty plus standby, with automatic changeover and alternating run hours so both machines wear evenly. The room must hold both, plus the headers and valve sets.
- Noise and vibration. Inertia bases, anti vibration mounts, flexible connectors on suction and discharge, and pipe supports that do not transmit vibration into the structure. Pump rooms adjacent to or below occupied space need acoustic treatment designed in, not added later.
- Suction conditions. Straight pipe or a suction guide ahead of the pump inlet. An elbow directly on the suction distorts the flow into the impeller and costs performance that will not appear anywhere in your calculation.
- Drainage and floor finish. Gland leakage, strainer cleaning and maintenance all put water on the floor. Provide falls and a drain.
Selecting Pump Design Inputs
Four inputs, in order of how much trouble they cause when wrong.
1. Flow Rate
Comes from the cooling load calculation done upstream, not from the pump. The pump designer does not choose it. Pipe bore selection that determines velocity and friction loss is a separate step — covered in our pipe sizing calculation guide.
For a primary pump, use the load of the single chiller it serves. For a secondary pump, use the total block load the loop must deliver, which may be less than the sum of zone peaks because of diversity. Getting this distinction wrong produces a secondary pump sized for the whole plant when it only needs to serve part of it, or worse, the reverse.
2. Total Head
Built up along the index circuit only — the full method for calculating pipe friction losses is in our pressure drop piping guide. The critical rule: in a closed loop, do not add the static height of the building. Water rising up a riser also comes back down, so the elevation terms cancel exactly. Static height matters for filling the system and setting expansion vessel pressure, not for pump duty. Only an open circuit, such as condenser water to a tower basin, has genuine static lift.
3. Fluid Properties
Water at 7°C is roughly 40% more viscous than at 20°C. A 30% glycol solution at the same temperature is denser and several times more viscous again, which raises friction loss and reduces the head and efficiency a pump delivers against its water based catalogue curve. Where glycol is used, apply the manufacturer's correction factors rather than ignoring them.
4. NPSH Available
Net positive suction head available must exceed what the pump requires at duty, with margin. In a closed chilled water system this is usually comfortable because the expansion vessel maintains positive pressure at the pump suction, but it must be checked rather than assumed, particularly on the topmost plant rooms where system pressure is lowest.
Pump Sizing & Schedule Generator
Enter the load and the index circuit components. The calculator builds the head, sizes the motor and produces the schedule entry.
| Component | Typical drop | Note |
|---|---|---|
| Chiller evaporator | 30 to 80 kPa | From the chiller selection. Never assume, it varies widely by machine and pass arrangement |
| AHU / FCU cooling coil | 20 to 60 kPa | From the coil selection at design flow. Deeper coils and higher delta T mean higher drop |
| Two way control valve | 25 to 50 kPa | Sized for valve authority, typically 25 to 50% of the variable branch resistance |
| Balancing valve | 15 to 30 kPa | Needs a minimum signal drop to be measurable and adjustable |
| Y strainer, clean | 5 to 15 kPa | Rises steeply as it fouls. Design on a partially fouled figure |
| Pipe friction | 200 to 400 Pa/m | Design basis for HVAC water. Lower for long runs, higher only where space forces it |
| Fittings allowance | 25 to 40% of pipe friction | First pass only. Use loss coefficients for the issued calculation |
| Plate heat exchanger | 30 to 70 kPa | Where the loop is pressure broken on tall buildings |
Order of magnitude values for orientation and early sizing only. Every one of these must be replaced with the selected equipment's actual figure before the schedule is issued, because a coil that turns out to be 55 kPa rather than 35 kPa moves the pump duty by 2 m of head.
System Volume and Minimum Loop Volume
The calculation most often skipped, and the cause of a specific and very recognisable fault: chillers that short cycle at low load. A chiller needs enough water in the loop that its own capacity change does not swing the return temperature faster than its controls can respond. Too little volume and it loads, overshoots, unloads, overshoots again, and hunts.
- V minimum loop volume, litres
- Q chiller design flow, litres per second
- t minimum loop time, minutes
Comfort cooling with reasonably stable load is commonly designed at 3 minutes of design flow. Variable primary flow, tight temperature control or highly variable process loads need more, often 5 to 6 minutes. Note that 3 minutes works out at almost exactly 30 litres per TR, or 8 US gallons per ton, which is why the old 6 to 10 gallons per ton rule of thumb and the minutes based rule give the same answer. Always confirm against the chiller manufacturer's own stated minimum, which governs.
Where the volume actually is
Run the calculator and note which term dominates. On a compact plant with short pipe runs, the pipework contributes surprisingly little and the shortfall is real. On a sprawling campus with hundreds of metres of large bore distribution, the pipe alone often satisfies the requirement several times over. This is why small plants short cycle and large ones rarely do, and why the buffer tank question is really a question about how distributed the building is. A 250 TR plant serving one compact block is the classic candidate.
Expansion Tank and Fill Pressure
Here is where the static height of the building comes back, and it matters that a designer knows both halves of this. Static height does not go into pump head, because the loop is closed and the elevation terms cancel. But it absolutely does set the fill pressure and it drives the expansion tank size. Confusing the two produces either an oversized pump or a system that will not vent.
Where Hstatic is the height from the tank to the highest point in the system and Hmargin is typically 3 to 5 m. The margin ensures the highest point stays above atmospheric pressure at all times, so air vents can actually expel air rather than draw it in, and so the pump suction never falls to vapour pressure. A system that will not vent at the top floor is almost always underfilled.
E is the fractional expansion of water between the coldest and hottest states the system will see. P1 and P2 are the minimum (fill) and maximum (relief setting) pressures, both absolute, so add 1.013 bar to each gauge value. The denominator is the tank acceptance factor: the fraction of the tank shell that is actually usable.
Static height: in one calculation, out of the other
This is the distinction worth being precise about, because it is where two common errors live. Pump head: static height is excluded. The closed loop returns the water it lifts, so gravity is not something the pump fights. Include it and you oversize the pump by the height of the building. Fill pressure and expansion tank: static height is central. It sets the minimum pressure the system must be charged to, which in turn sets the tank acceptance factor and the NPSH available at the pump suction. Omit it and the top floor will not vent and the pump may cavitate. Same number, opposite treatment, and knowing which is which is a reliable marker of whether someone has actually designed a system or only read about one.
Understanding Pump Performance Outputs
Reading a Manufacturer Pump Curve
A pump curve sheet carries several curves on one set of axes, with flow on the horizontal and head on the vertical.
- Head vs flow. The main curve, falling from left to right. Your duty point must sit on it, not near it.
- Multiple impeller curves. A family of parallel curves for different impeller diameters within the same casing. This is how a manufacturer fine tunes a selection to your exact duty.
- Efficiency islands. Contour lines of constant efficiency, closing around the best efficiency point. Where your duty falls relative to these islands is the whole judgement.
- Power curves. Shaft power against flow, usually rising with flow. Read the value at your duty, and also at the end of the curve, because that is what the motor may have to deliver if system resistance turns out lower than calculated.
- NPSH required. A separate curve rising with flow. Compare against NPSH available with margin.
Why the Duty Point Should Sit Near BEP
Operating away from the best efficiency point costs energy continuously, and it also costs bearing and seal life through unbalanced radial loads on the impeller. Hydraulic Institute guidance defines a preferred operating region around BEP, and API 610 requires the rated point to fall within a tighter window still. We cover the operating regions, the efficiency formula and the affinity laws in detail in our companion guide to pump efficiency and BEP.
The habit that guarantees an off BEP selection
Padding the head calculation. Ten percent on the pipe friction, another ten on the coil, a round up on the valve, and the pump arrives oversized. The commissioning engineer then throttles the balancing valve to shed the excess head, so the pump operates left of its design point, burning energy across a valve to correct an arithmetic decision made months earlier. An honest head calculation produces a pump that lands on BEP. A padded one produces a pump that never can. If uncertainty genuinely exists, a variable speed drive absorbs it far more cheaply than a throttled valve.
The Chilled Water Pump Schedule
The schedule is where design work becomes a procurable deliverable. It is the document the contractor prices from, the vendor quotes against, and the site team checks deliveries with. A vague schedule produces a cheap pump that meets the letter of it.
| Field | Why it is there |
|---|---|
| Tag number | Unique identifier tying drawing, schedule, BMS point list and O&M manual together |
| Service | Which system and which part of it, so nobody installs a secondary pump in the primary position |
| Quantity and standby | Duty plus standby arrangement and changeover requirement |
| Design flow and head | The duty point. The two numbers everything else follows from |
| Fluid and temperature | Water or glycol, concentration, operating temperature, all of which change the selection |
| Minimum efficiency at duty | Stops a vendor meeting flow and head with a cheap, inefficient machine |
| Motor rating, speed, supply | Electrical load for the panel schedule and cable sizing |
| Suction and discharge sizes | Pipework connection sizes for the installation drawings |
| Pump type | End suction, split case, inline, vertical multistage |
| Control method | Fixed speed or VFD, and the control variable such as differential pressure |
| Accessories | Inertia base, AV mounts, flexible connectors, suction guide, triple duty valve, gauges |
How MEP Teams Actually Use It
- Procurement. The vendor quotes against the schedule. Every field left vague is a field the vendor gets to choose, usually in the direction of lower cost.
- Electrical coordination. Motor ratings feed straight into the panel schedule, cable sizing and load estimate. A pump schedule reissued late means an electrical redesign.
- Structural coordination. Pump and base weights feed the plant room floor loading, which the structural engineer needs before the slab is designed.
- Controls and BMS. Tags and control methods feed the points list and the sequence of operations.
- Commissioning and TAB. The balancing contractor measures against the scheduled flow and head, and the schedule becomes the reference the as built performance is judged against.
Where This Skill Leads
Pump selection sits at the junction of hydraulics, equipment knowledge, energy and coordination, which is why it separates people who can draw a plant room from people who can design one. An engineer who can build an honest head calculation, defend a duty point on a curve, and issue a schedule a contractor can price without a single query is doing recognisably professional work.
For the hydraulics underneath, see our guide to flow through pipes, and for efficiency, BEP and the affinity laws, see pump efficiency.
Frequently Asked Questions
Sources & Further Reading
- ASHRAE Handbook: HVAC Systems and Equipment, ASHRAE. Chapters on Hydronic Heating and Cooling and on Centrifugal Pumps, covering primary secondary and variable primary flow arrangements, decoupler design and pump selection.
- ASHRAE Handbook: Fundamentals, Pipe Sizing chapter, for the friction and fitting loss data behind any head build-up.
- ANSI/HI 9.6.3, Rotodynamic Pumps: Guideline for Allowable Operating Region, Hydraulic Institute. Defines the preferred and allowable operating regions referenced when judging a duty point against BEP.
- Manufacturer performance curves and selection software, for example from Grundfos, KSB, Wilo, Armstrong or Kirloskar. The certified curve for the specific pump always governs over any generic figure.
- ISO 9906, Rotodynamic pumps: Hydraulic performance acceptance tests, for the acceptance grade and tolerance that should be named in a specification where efficiency is contractually important.
Calculation basis for this article's tools
The decoupler simulator applies a steady state mass and energy balance at the header junction: mixed temperature is the flow weighted average of the two streams entering the header. It assumes ideal mixing, a decoupler of negligible resistance, and neglects pipe heat gain and thermal transport delay, so real plants respond more slowly and less cleanly than the model. The pump sizing tool builds head from the component pressure drops entered, converts kPa to metres by dividing by 9.81, computes hydraulic power as flow times head times gravity, and divides by the assumed efficiencies. Motor selection takes the next standard IEC rating at least 15% above shaft power at duty; a real selection should be checked against end of curve power from the manufacturer. Component pressure drops shown as defaults are typical order of magnitude values for orientation, not design data. These are teaching and first pass tools. Issued designs should use certified manufacturer curves and validated hydraulic calculations.
This article was last reviewed on 1 August 2026.
Ready to start your engineering career journey?
Investing in structured training is one of the best ways to set yourself apart in MEP engineering. Our Courses cover HVAC, electrical and plumbing design in depth, plus tools like HAP software — the exact skillset that sets you apart in a competitive field.