A 550 kW water-cooled chiller leaves the factory with a fixed heat-transfer area. Once the shell is welded and the tube bundle is installed, the area cannot change. What can change in service is the overall heat-transfer coefficient and the mean temperature difference. The coefficient is set by tube geometry, water velocity and fouling; the temperature difference is set by the chilled-water outlet and the evaporating temperature. When capacity falls after three years, the cause is more often fouling and a stretched temperature difference than compressor wear — and both can be calculated from pressure readings already shown on the controller.
Four field criteria follow: tube velocity between 1.5 and 2.5 m/s; evaporator approach of 1–2 K; condenser approach of 1–3 K; and a fouling allowance that is monitored rather than assumed. Each can be checked without dismantling the machine.
Only two of the three variables can move in service
Heat transfer is Q = U × A × ΔT. Area is fixed at design. Increasing area by 10% for the same duty relaxes both U and ΔT, but costs a larger shell, more refrigerant and more mechanical room space. The area in a catalog is therefore not “larger is always better”; it is sized to be just sufficient at the specified fouling factor, leaving margin for fouling growth.
The overall coefficient is the reciprocal of four series resistances: tube-wall conduction, water-side convection, water-side fouling and refrigerant-side convection. A clean shell-and-tube evaporator commonly has U of 1,800–2,600 W/(m²·K). A condenser benefits from condensation on the refrigerant side and commonly reaches 2,000–2,800 W/(m²·K). A plate heat exchanger, with narrow channels and higher velocity, can reach 3,500–5,000 W/(m²·K), but only with better water quality.
Check the numbers for consistency. For a 550 kW machine with 70 m² evaporator area, 12/7 °C chilled water and 5 °C evaporating temperature, the log-mean temperature difference is 3.99 K. The implied coefficient is 550,000 ÷ (70 × 3.99) ≈ 1,970 W/(m²·K), which sits inside the normal shell-and-tube range. If a new machine calculates to only 1,200 W/(m²·K), either the area is overstated or the nominal water temperature difference is not 5 K.
Why tube velocity is held between 1.5 and 2.5 m/s
Water-side convection rises approximately with velocity to the power 0.8. Raising velocity from 1.5 to 2.5 m/s increases the water-side coefficient by about 48%, but pressure drop rises with velocity to the power 1.8–2.0, so the same circuit needs more than 2.5 times the pressure. Chilled-water and condenser-water pumps typically account for 15–25% of total refrigeration-system electricity. Above 2.5 m/s, the pump consumes much of what the chiller saves.
The lower limit is set by fouling. Deposition of suspended solids and calcium/magnesium scale accelerates below 1.2 m/s, especially in horizontal tubes, where a sludge band can form along the bottom. The usual 1.5 m/s design lower limit is not a round number; it leaves roughly 20% margin for flow imbalance.
The upper limit is set by erosion. Copper tubes above 2.5–3.0 m/s can suffer impingement corrosion at the inlet and on the outside of bends, shown as local thinning and pinholes. With high suspended sand or open cooling water, the limit should be reduced to 2.0 m/s. 316L stainless steel and titanium allow higher velocities, but their heat-transfer behavior differs and copper-tube values should not be transferred directly.
Velocity can be checked without opening the bundle. For a Φ19 × 0.9 mm tube, the bore is 17.2 mm and the flow area is 2.32 × 10⁻⁴ m². If chilled-water flow is 73 m³/h, the evaporator has two passes and 44 tubes per pass, velocity is 73 ÷ 3,600 ÷ (44 × 2.32 × 10⁻⁴) ≈ 2.0 m/s. If the same machine has 60 tubes per pass, velocity falls to 1.47 m/s, already in the fouling-sensitive range. The owner cannot change tube count, but can verify flow: a meter reading more than 10% away from design invalidates the velocity check.
Log-mean temperature difference and approach temperature
With chilled water entering at 12 °C and leaving at 7 °C, and evaporating at 5 °C, the two end differences are 7 K and 2 K. The arithmetic mean is 4.5 K, but the log-mean is 5 ÷ ln(7/2) = 3.99 K. Substituting the arithmetic mean would undersize the required area by about 11%, and area cannot be changed after manufacture. This is one reason a chiller can be short of capacity in summer with no obvious fault.
ΔT is the driving force. For the same duty, a larger ΔT means a smaller area or a lower coefficient. Conversely, with fixed area and coefficient, the machine must depress the evaporating temperature to hold duty, and every 1 K reduction in evaporating temperature costs about 2.5–3% in COP. The chain from tube cleanliness to the electricity meter runs through this point.
Approach temperature makes the chain measurable. Evaporator approach is chilled-water outlet minus evaporating temperature; design is 1–2 K. Condenser approach is condensing temperature minus cooling-water outlet; 1–3 K is common for water-cooled machines. Both can be derived from pressure: convert the evaporating and condensing pressures to saturated temperatures and subtract the outlet water temperatures. No additional instrument is needed.
Three cautions apply. Use the outlet water temperature, not the average of supply and return, otherwise a change in water temperature difference is misread as fouling. Use a calibrated pressure reading; if the controller display comes from the same sensor, it is valid for trend comparison but not for absolute diagnosis. At loads below 40%, approach naturally rises by 0.3–0.5 K, so do not make a cleaning decision in that range.
What a fouling resistance of 0.0001–0.0002 m²·K/W means
Fouling resistance is the margin built into the nominal rating. A nominal condition is commonly based on 0.0001 m²·K/W on the evaporator water side and 0.0002 m²·K/W on the condenser water side; the catalog capacity is measured in that “half-fouled” state. After cleaning, U is above nominal; after two or three years it returns to nominal, and after that it falls below.
The series calculation is 1/U = 1/U₀ + Rf. With a clean U₀ of 2,000 W/(m²·K), adding 0.0001 reduces U to 1,667, a fall of 17%; 0.0002 reduces it to 1,429, a fall of 29%; 0.0003 reduces it to 1,250, a fall of 37%. To hold 550 kW, ΔT must rise proportionally, which means the evaporating temperature is depressed.
| Fouling resistance (m²·K/W) | U (W/(m²·K)) | LMTD for 550 kW | Evaporating temperature | Evaporator approach | COP change |
|---|---|---|---|---|---|
| 0 (cleaned) | 2,000 | 3.9 K | 5.0 °C | 2.0 K | Base |
| 0.0001 | 1,667 | 4.8 K | 4.3 °C | 2.7 K | About −2% |
| 0.0002 | 1,429 | 5.5 K | 3.6 °C | 3.4 K | About −3.5% |
| 0.0003 | 1,250 | 6.3 K | 2.9 °C | 4.1 K | About −5% |
Translate the last row into money. A 550 kW machine at COP 5.6 draws about 98 kW; a 5% loss adds 5.2 kW. At 2,500 hours per year and 0.75 CNY/kWh, that is about 9,750 CNY per year. A mechanical cleaning of the evaporator and condenser is typically a few thousand CNY, an order of magnitude less. That is the economic case for cleaning when approach rises by 1 K.
When to clean: three field criteria
First, evaporator approach 1 K above the commissioning baseline. At acceptance, record the evaporating pressure, condensing pressure and supply/return temperatures after 30 minutes at stable full load. Thereafter read them quarterly. Clean when approach exceeds baseline by 1 K, or exceeds 2.5 K absolute.
Second, condenser approach above 3 K, or the saturated condensing temperature minus cooling-water outlet rising continuously for two weeks. Condenser fouling affects efficiency faster than evaporator fouling because cooling water is an open circuit; poor cycles-of-concentration control can build 0.2 mm of scale in one summer.
Third, current rising at the same duty and the same operating conditions. This only means something if cooling-water inlet, flow and chilled-water outlet are identical between the two records. Comparing July with April will always show lower efficiency, because the cooling water is warmer.
Cleaning method follows the deposit. Calcium carbonate scale responds to low-concentration acid; biological slime and suspended solids need high-pressure water or brushing; iron oxide needs a dispersant first. Copper tubes should not be acid-cleaned often, because each clean removes some base metal. The more fundamental fix on the cooling-water side is to control cycles of concentration and dose scale/corrosion inhibitors, holding annual fouling growth below 0.00005 m²·K/W — roughly three to four years between mechanical cleanings.
Two boundaries matter. A plate exchanger has no measurable tube velocity and cannot be retubed, so monitor its water-side pressure drop as well as approach temperature; a 50% channel blockage raises pressure drop before approach alarms. An inverter chiller at part load can fall below 1.0 m/s, so set minimum-flow protection above the exchanger limit. Below 2 °C evaporating temperature, read approach together with glycol concentration, because glycol specific heat and density differ from water and water-based formulas overestimate capacity by 5–8%.
Bundle and plate types, and the boundary where the field criteria fail
The four field criteria above assume a shell-and-tube machine with a measurable tube velocity and a water side that can be brushed. The bundle type and the heat-exchanger family change what can be measured, and they decide when the criteria stop being reliable.
| Type | Clean U range | Refrigerant charge | Water-side cleaning | Failure boundary |
|---|---|---|---|---|
| Flooded shell-and-tube | 1,800–2,600 W/(m²·K) | Baseline | Brush, pull tubes, on-line ball cleaning | Liquid-level or oil-return fault causes liquid carry-over |
| Falling-film shell-and-tube | 2,200–3,000 W/(m²·K) | 50–70% of flooded | Brush, pull tubes | Below 25% load, poor distribution leaves tubes dry |
| Gasketed plate | 3,500–5,000 W/(m²·K) | 20–30% of shell-and-tube | Open plates, chemical cleaning | Channel blockage or gasket leakage |
| Brazed plate | 4,000–6,000 W/(m²·K) | 15–25% of shell-and-tube | Chemical cleaning only | Blockage cannot be restored; replace the block |
A gasketed plate exchanger earns its higher U with a channel gap of only 2–5 mm. When a channel blocks, the water-side pressure drop rises before the approach temperature moves, so log both readings. A brazed plate unit cannot be opened; a blocked channel usually means replacement, not cleaning. On these units, the tube-velocity rule is meaningless and pressure drop becomes the early warning.
Variable-speed machines reduce flow at part load. If flow falls below 1.0 m/s, fouling accelerates even when the controller reports a normal leaving-water temperature; set the minimum flow above the exchanger’s lower limit. Below 2 °C evaporating temperature, glycol changes both density and specific heat, and a water-based capacity calculation can overstate duty by 5–8%. For any plate unit, read pressure drop and approach together, and do not transfer shell-and-tube numbers to a plate block. Longcom’s water chiller page lists the bundle and plate options available for different water qualities.
Common questions
What is a normal evaporator approach?
Design is 1–2 K; a new machine at full load commonly reads 1.2–1.8 K. Above 2.5 K, or 1 K above the commissioning baseline, check water flow and tube fouling. Cleaning usually restores 0.5–0.8 K. At loads below 40%, the reading naturally rises by 0.3–0.5 K, so do not judge in that range.
Why does efficiency not recover after condenser cleaning?
If approach has returned to baseline but electricity has not fallen, the cause is not the water side. Check refrigerant charge, non-condensable gas and chilled-water flow. Cleaning restores U; if evaporating temperature is still depressed by low flow, COP will not improve.
Shell-and-tube or plate?
Use plate where water quality is stable, suspended solids are low and space is tight. Use shell-and-tube where water quality varies, sand is present and mechanical cleaning is required. A plate with 2–5 mm channels can block to a pressure alarm within a year on an open cooling circuit with poor concentration control.
Can velocity above 3 m/s improve heat transfer?
Not worth it. Water-side convection rises with velocity to the power 0.8, while pressure drop rises with the power 1.8–2.0. From 2.0 to 3.0 m/s, heat transfer improves by less than 20% but pump power roughly doubles. Copper tubes above 2.5–3.0 m/s also suffer inlet impingement; with sandy groundwater, stay below 2.0 m/s.
Flooded or falling-film?
Falling-film has 10–20% higher U at full load and can run 0.5–1 K lower condensing temperature. But below 30% load, poor distribution can cancel the advantage. For a load that swings widely, compare part-load curves before paying the premium.
How much does raising chilled-water outlet from 7 to 9 °C save?
Evaporating temperature rises by about 2 K, and COP improves by 5–6%. The condition is that the terminals still meet dehumidification at 9 °C supply. If fan coils must control humidity, raising supply can cause insufficient dehumidification, so check terminal capacity first.
How does approach show fouling?
A rising approach at the same load means fouling. Compare with the design and the last clean reading. Approach is the cheapest fouling indicator.
What is the design delta-T?
Usually 5 °C on chilled water and 5 °C on condenser water, but check the design. A lower delta-T means more flow and pump power.
Related reading
- What Is a Water Chiller? Working Principle and Selection
- Descaling a Chiller Condenser
- COP vs IPLV: What the Two Figures Measure
- Water Chiller Maintenance Schedule
Related reading
Evaporator cleaning · Chiller capacity loss · Chiller plant sensors