Fouling rarely announces itself. A chiller with scaled condenser tubes still makes cold water; a cooling tower with clogged fill still returns cool-ish water. What changes first is a handful of temperature differences, and most building automation systems already log the points needed to see them.
Four temperature differences worth trending
- Tower approach = water leaving the tower − outdoor wet-bulb. The wet-bulb is the lowest temperature evaporative cooling can reach, so approach says how close the tower gets. For a given load and wet-bulb it should be steady. If it creeps up, or the fans have to run harder to hold it, look at fill, airflow, water distribution and fouling.
- Condenser range = water to the tower − water from the tower. It follows load and condenser flow.
- Lift = condenser water leaving the chiller − chilled water leaving the chiller. Lift is a close proxy for the pressure difference the compressor works against. More lift means more kW per ton.
- Condenser approach = refrigerant condensing temperature − condenser water leaving the chiller. This is the most direct view of tube condition: scale and biofilm insulate the tubes, and the refrigerant has to run hotter to push the same heat through. It needs the chiller's refrigerant temperature or pressure, which many plants log only in the chiller's own controller.
None of these has a universal "good" number. Each depends on the equipment and the operating point. What matters is whether it drifts for the same load and weather.
What fouling looks like in data
Sample data: LBNL's public simulated chiller plant. LBNL publishes it as a fault-detection dataset: three chillers and three cooling towers serving a 12-storey office, modeled in Chicago weather, with one year of one-minute data for each fault. One case reduces cooling tower 1's heat-transfer coefficient to 65% of normal. We ran July from the fault-free case and the fouled case through Plant Check's calculations, with our reference mapping for both files and the same definitions (in Plant Check itself, the AI proposes a mapping and you confirm it):
| July, sample plant | No faults | Tower 1 fouled (65%) |
|---|---|---|
| Share of running hours with tower 1's fan at ≥ 95% speed | 9.3% | 85.9% |
| Tower fan energy per ton-hour (all three towers) | 67.1 Wh | 96.9 Wh |
| Tower approach, worst 10% of running intervals | 8.8 °F | 10.7 °F |
| Tower approach, median | 8.0 °F | 8.7 °F |
| Lift, median | 39.9 °F | 40.9 °F |
| Chiller kW/ton, energy-weighted | 1.681 | 1.710 |
Healthy vs fouled, same plant
Tower 1 fan at ≥95% speed
share of running 15-min intervals · CT_FAN_SPD_1
Fault-free run: 9.3%, fouled run: 85.9%
+76.6 ptsTower fan energy per ton-hour
sum of CT_POW_1–3
Fault-free run: 67.1 Wh/ton-h, fouled run: 96.9 Wh/ton-h
+44%Tower approach, worst 10%
p90 while running
Fault-free run: 8.8 °F, fouled run: 10.7 °F
+1.9 °FChiller kW/ton, energy-weighted
the chillers barely changed
Fault-free run: 1.681 kW/ton, fouled run: 1.710 kW/ton
+1.7%- Chillers off
- Standby (below 5% of peak load)
- Fan speed while running:
- <50%
- 50–80%
- 80–90%
- 90–95%
- ≥95%
Tap, hover or use ↑/↓ to read a day.
Three things stand out.
- The controls hid the fault from the chiller. The plant's control sequence holds tower leaving water at wet-bulb + 8 °F. With a fouled tower it did so by running tower 1's fan flat out, so median approach barely moved.
- The fans told the story. A fan at full speed in most running hours, and about 44% more tower fan energy per ton-hour, is the clearest signal in the data.
- The chiller paid a little, the tower a lot. Chiller kW/ton rose about 1.7%. The fans absorbed most of the penalty, but the worst-10% approach still opened up, and not in the hottest weather. The five days with the highest peak approach all peaked at a lower wet-bulb than any of the five hottest days, with tower 1's fan already at full speed for most of their running intervals.
Tower approach over running intervals
Tower approach, °F: water leaving the tower minus outdoor wet-bulb
View as table
| Tower approach, °F | No faults: intervals | No faults: share | Tower 1 fouled · UA × 0.65: intervals | Tower 1 fouled · UA × 0.65: share |
|---|---|---|---|---|
| 4.0 to <4.5 | 2 | 0.1% | 0 | 0.0% |
| 4.5 to <5.0 | 4 | 0.2% | 0 | 0.0% |
| 5.0 to <5.5 | 11 | 0.6% | 2 | 0.1% |
| 5.5 to <6.0 | 20 | 1.1% | 10 | 0.6% |
| 6.0 to <6.5 | 34 | 1.9% | 9 | 0.5% |
| 6.5 to <7.0 | 81 | 4.5% | 20 | 1.1% |
| 7.0 to <7.5 | 218 | 12.1% | 79 | 4.4% |
| 7.5 to <8.0 | 568 | 31.6% | 156 | 8.7% |
| 8.0 to <8.5 | 574 | 31.9% | 393 | 21.8% |
| 8.5 to <9.0 | 174 | 9.7% | 548 | 30.5% |
| 9.0 to <9.5 | 67 | 3.7% | 226 | 12.6% |
| 9.5 to <10.0 | 25 | 1.4% | 98 | 5.4% |
| 10.0 to <10.5 | 9 | 0.5% | 60 | 3.3% |
| 10.5 to <11.0 | 4 | 0.2% | 53 | 2.9% |
| 11.0 to <11.5 | 4 | 0.2% | 31 | 1.7% |
| 11.5 to <12.0 | 2 | 0.1% | 25 | 1.4% |
| 12.0 to <12.5 | 1 | 0.1% | 19 | 1.1% |
| 12.5 to <13.0 | 0 | 0.0% | 21 | 1.2% |
| 13.0 to <13.5 | 1 | 0.1% | 15 | 0.8% |
| 13.5 to <14.0 | 0 | 0.0% | 22 | 1.2% |
| 14.0 to <14.5 | 0 | 0.0% | 8 | 0.4% |
| 14.5 to <15.0 | 0 | 0.0% | 4 | 0.2% |
The pattern is what to look for in your own data: when a tower loses capacity, fan speed and the high-percentile approach move before the averages do.
Reading it from your own plant
- Trend tower approach against wet-bulb, and fan speed against load. A tower that needs more fan for the same job is losing capacity.
- Trend lift and kW/ton against load and wet-bulb, not as raw averages. Averages mix a hot week with a cool one.
- If your chillers log refrigerant condensing temperature, trend condenser approach. It is the direct measure of tube fouling and the best trigger for tube cleaning.
- Put the water-treatment log on the same timeline: cycles, conductivity, biocide residuals, cleanings. Fouling on the energy side usually has a cause on the water side.
LBNL's broader campaign found that buildings using fault detection and analytics saved a median of about 9% of energy. The analysis is not the hard part. What makes the savings real is acting on the findings consistently. Plant Check is built to make the first look take minutes: upload a BMS export and see these numbers for your plant.