Building automation
Points by AHU, VAV, fan-coil and chiller; schedule and override audit; simultaneous heating and cooling; set-point drift; trend-log quality. Back to Buildings and chillers · Data centres
2 units with a remaining-life estimate
Equipment tree
Click a check to see why an alarm would be raised, at your chosen level.
- Air handlers (AHU)
- Dampers, mixed air, heating and cooling valves, fans
- VAV boxes
- Damper, airflow reading, reheat
- Fan-coil units
- Valves, fan, filter, room sensor, schedule
- Chillers
- See the Buildings and chillers page and the Data centres page
- Cleanroom suites
- Differential pressure across doors
BMS points view
The points CAPMAC reads for each kind of equipment. All are read only.
AHU
| Point | BACnet object type | Why it is read |
|---|---|---|
| Supply air temperature | analogInput | Discharge temperature after the coils |
| Mixed air temperature | analogInput | After the dampers, before the coils |
| Return and outdoor air temperature | analogInput | For the free-cooling and damper reasoning |
| Outdoor-air damper position | analogOutput | Command and, if wired, feedback |
| Heating valve and cooling valve position | analogOutput | Both are needed to see simultaneous heating and cooling |
| Supply fan status and speed | binaryInput / analogOutput | On-off and speed |
| Static pressure | analogInput | Duct pressure the fan holds |
| Occupancy schedule state | binaryValue | Which mode the BMS thinks it is in |
VAV box
| Point | BACnet object type | Why it is read |
|---|---|---|
| Zone temperature | analogInput | Space reading |
| Zone set-point (cooling and heating) | analogValue | Watch for drift and overrides |
| Airflow reading and airflow set-point | analogInput / analogValue | Needed for the bias and stuck-damper checks |
| Damper position | analogOutput | Command |
| Reheat valve position | analogOutput | Reheat running while the zone is warm is a waste |
| Discharge air temperature | analogInput | After reheat |
Fan-coil unit
| Point | BACnet object type | Why it is read |
|---|---|---|
| Zone temperature and set-point | analogInput / analogValue | Space reading and target |
| Fan speed or state | multiStateOutput / binaryOutput | Left on out of hours |
| Heating and cooling valve position | analogOutput | Valve against command |
| Supply and return water temperature | analogInput | Water-side fouling and valve leak |
Chiller
| Point | BACnet object type | Why it is read |
|---|---|---|
| Leaving and entering chilled-water temperature | analogInput | Delivered cooling |
| Leaving and entering condenser-water temperature | analogInput | Heat rejection |
| Compressor power or current | analogInput | Efficiency |
| Chilled-water flow | analogInput | Duty |
| Status and alarm word | binaryInput | Trips and lockouts |
Audits and reports
These look across your whole BMS. Evidence level on each says how far we have tested it.
Lists every point where the BMS is not following its own schedule: a fan or valve forced on, a set-point manually held, an occupancy mode stuck. Needs the schedule state and the override flags read read-only over BACnet.
Evidence The closest tested piece is the fan-coil left-on-out-of-hours check, caught in two held-out simulated tests. The audit as a whole has no test yet.
Flags units whose heating and cooling valves are both open at the same time. Reads the two valve positions of each AHU, VAV and fan-coil.
Evidence Valve-leak checks for air handlers passed 8 of 8 and 6 of 6 on simulated records. A report of both valves open in your own data needs no test: it is a count of your own readings.
Shows where a zone set-point has moved away from the commissioned value over weeks. Reads the set-point point each night.
Evidence No test yet. A plain count of changes, not an alarm.
Shows gaps, flat-lines, stuck values and spikes in the logs the BMS already keeps.
Evidence It stays a report you read yourself; it does not raise alarms.
Working checks
| Check | Equipment | Evidence level | Result |
|---|---|---|---|
| VAV box | Real data: partly passed | Real building: evening 18 of 18 passed; night 15 of 18 passed; morning 10 of 18 stopped on an older version. | |
| VAV box | Real data: partly passed | Real building: evening 7 of 8 passed; night 0 of 8 stopped; morning 1 of 8 stopped on an older version. | |
| Fan-coil unit | Outside simulator: passed | Two LBNL laboratory-unit files (outside simulator, not real plant data): severe fouling alarmed on 81 of 89 days. Heating-side fouling is not proven. | |
| Fan-coil unit | Held-out test on simulated records: passed | Six checks caught faults in two held-out tests. | |
| Air handler | Simulated test: passed | Stuck damper 8 of 8, mixed-air bias 8 of 8, cooling-valve leak 8 of 8, heating-valve leak 6 of 6, 0 false in 17 records and 120 healthy days. | |
| Air handler | Real data: partly passed | Real Korean office air-handler data: injected bias found with no false call on a sham run; 9 of 19 stretches could be judged and natural faults could not be tested. | |
| Cleanroom suite | Simulated test: passed | Door ajar between rooms 1 and 2 caught in 6 of 6 records, 0 false in 10. |
Case studies
Real data first.
Stuck damper: evening 18 of 18 found, night 15 of 18 found. Airflow sensor bias: evening 7 of 8 found. 1 of 234 healthy boxes raised an alarm in the evening set.
Severe fouling alarmed on 81 of 89 days. Heating-side fouling is not proven.
Injected bias was found on a narrow test (9 of 19 stretches judged). The Iowa healthy air handler stayed quiet; fault detection could not be tested there.
A check built from the unit's own normal level found 1 of 2 damper days on a thin test. Small sample; trial on your own records recommended.
Valve leak, filter, sensor bias, stuck valve, fan loss and left-on-out-of-hours were caught in both fan-coil tests.
Chiller evaporator fouling 6 of 6; air-handler damper 8 of 8; cleanroom door 6 of 6; 0 false alarms in the healthy records. Power for false-alarm evidence was thin (2 to 4 look-alikes per family).
What is not claimed yet
- Schedule and override audit, and set-point drift, have no test yet and raise no alarms.
- Trend-log quality is a report to read, not an alarm.
- Heating-side fan-coil fouling is not proven; the real test covered the cooling circuit only.
- Airflow sensor bias at night is not confirmed on real data.
Connect
CAPMAC only reads. For a building automation system it can read BACnet/IP points through a read-only connector (optional extra, installed only when used). It never writes, overrides or releases a point. See the Connect page.