By Kurnia · Published November 13, 2026 · Version 1 · 21 min read
Alarms, trends, and schedules in a BMS
alarm priorities and classes, hysteresis and time delay, alarm floods, trend logs and sampling intervals, schedules and manual overrides, and handling alarms
After a brief voltage dip and a transfer to the generator set, the BMS operator station of a hospital received 186 alarms in ten minutes, and the alarm for an isolation room exhaust fan was acknowledged along with the others without anyone going to the field. This chapter covers the three BMS functions that the duty technician uses most: alarms, trend logs, and schedules. It covers the difference between alarms, events, and status; alarm classes and priorities; hysteresis and time delay; alarm floods and standing alarms, with the alarm rate targets from the guidance of the UK HSE; trend logs, sampling intervals, and storage capacity; occupancy schedules and manual overrides and how to record them; and the sequence for handling an alarm from the sound to the record. Changing alarm limits, delays, hysteresis, priorities, and schedules in the controller program, and disabling alarms, are set out as the authority of the system integrator through change control; this chapter gives no way to switch off or hide alarms.
At 02:40 in the morning, the PLN supply to a hospital dips briefly, the generator set starts, and the transfer switch moves the load. In the next ten minutes the BMS operator station receives 186 alarms: communication loss from more than a dozen controllers that restarted, fans that stopped, pumps that failed to start, and room temperatures beginning to rise. The duty technician is alone; to stop the sound, they acknowledge the alarms page by page, twenty at a time. One alarm on the third page has high priority: the isolation room exhaust fan is commanded on, but its status is off. Its variable speed drive needs a local reset after the voltage dip, and no one goes to it. The isolation room nurse calls forty minutes later because the pressure indicator at the room door does not show the normal condition.
The next morning, when the log is reviewed, 31 of the 186 alarms turn out to be old alarms that had been standing for days, and the outpatient clinic supply fan ran all night because a manual override for a Saturday event was never released. No machine was seriously damaged that night. What failed was the alarm design, the habits for handling alarms, and the record of overrides.
Scope
This chapter covers alarms, trend logs, and schedules in a BMS from the building technician’s side: reading, acknowledging, checking, following up, recording, requesting trends, checking schedules, and using the manual overrides they are assigned. BMS architecture, the points list, and point naming are in Chapter 32 on BMS architecture. Communication faults between controllers are in Chapter 104 · Electrical & Control on building communication protocols. Sampling rate, aliasing, and clock synchronization are in Chapter 12 · Data & Measurement on time-series data. Trends for machine condition monitoring are in Chapter 14 · Mechanical & Mechatronics, probability and fault statistics in Chapter 10 · Data & Measurement, and how to present data in Chapter 16 · Data & Measurement. Energy dashboards are in Chapter 34 on building energy dashboards.
The fire alarm panel is a separate system with its own rules, covered in Chapter 19 to Chapter 23; the BMS does not replace it, and the fire status display on a BMS station must not be used as a substitute for the fire panel. Pressure and airflow in isolation rooms and operating rooms are covered in Chapter 30 on air conditioning for special hospital rooms.
Work outside the authority of the building technician: changing alarm limits, hysteresis, time delays, priorities, and alarm routing; disabling, delaying, or hiding alarms; changing controller programs and their sequences; changing accounts, passwords, access rights, and network configuration. All of that is the authority of the system integrator or IT administrator through change control approved by the building manager. The technician records the proposed change with its evidence and hands it over. This chapter gives no steps for switching off or hiding alarms, and gives no steps for bypassing safety devices.
Alarms, events, and status at the operator station
An alarm is an annunciation, audible or visible or both, that alerts the operator to an off-normal condition that may require action. That definition contains two conditions. The condition is abnormal, and someone has to do something. The alarm handling guidance from the UK HSE states that process status indicators must not be designated as alarms, and that every alarm must have a defined response.
From those conditions come three kinds of messages at the operator station. An event is a record of a change of state that requires no action, for example a pump changing from the duty pump to the standby pump on its rotation schedule. Status is the state of a point at that moment, for example a fan on or off. An alarm is an abnormal condition that requires action. Messages without an action that are configured as alarms train technicians to ignore the sound, including the sound of important alarms.
Every BMS alarm comes from a point: an analog point that crosses a limit, a digital point that changes to an abnormal state, a command that does not match its feedback status, or a controller that stops communicating. A good alarm message names the point, the kind of alarm, the time detected, the value and its limit, and advisory information.
Alarm classes and priorities
An alarm class is a grouping of alarms by their source or kind of condition, and an alarm priority is the order of importance of alarms according to the consequence if they are not answered. Common classes in buildings are process and comfort alarms (temperature, humidity, room pressure), equipment alarms (trip, failure to start, command and status mismatch), system alarms (communication loss, sensor out of range), and maintenance notices (filters, run hours).
The guide specification for a BAS (building automation system, another name for a BMS) based on direct digital control, prepared by NIST in 1991 for the GSA, the agency that manages US government buildings, uses three alarm levels.
| Level | Meaning | Examples in that guide |
|---|---|---|
| 1 | needs attention without urgency | temperature of noncritical rooms, air filter condition |
| 2 | needs immediate attention | temperature of critical rooms outside limits, safety or security sensors not working |
| 3 | concerns safety and security | fire and smoke sensors activated, security sensors activated |
The same guide requires level 2 and 3 alarms to be reported within 10 seconds of the condition occurring, to display the related graphics automatically, and to sound until acknowledged by the operator. Level 3 there includes fire sensors because buildings designed with that guide display the fire system on the BAS. In a building whose fire system is separate, that display is information only; action still follows the fire panel and Chapter 19 on fire alarm panels.
The HSE guidance suggests about three priorities, priorities set from the consequence if the operator fails to respond, prioritization rules applied consistently to every alarm, and a proportionate distribution, for example 5 percent high, 15 percent medium, and 80 percent low. That distribution comes from the chemical industry and is quoted by HSE from the paid EEMUA 191 guide; for buildings, the figures are used as a comparison, not a requirement. The idea still holds: if half of the alarms have high priority, priority no longer distinguishes anything.
Hysteresis and time delay
A measured value near an alarm limit moves up and down because of sensor noise and controller action. Without protection, the alarm activates and clears repeatedly within an hour. Such an alarm is called a chattering alarm, and the HSE guidance lists adjusting deadbands on repeating alarms as one of the quick fixes.
Alarm hysteresis (alarm deadband) is the difference between the value that activates the alarm and the value that clears it. For a high-limit alarm:
is the measured value, the alarm limit, and the hysteresis, all in the units of the point (°C, Pa, percent humidity). For a low limit the signs are reversed: active if and clears if . Hysteresis filters out swings smaller than , but it makes the alarm clear only after the value has moved far enough back from the limit.
An alarm time delay is how long the out-of-limit condition must persist before the alarm is announced. A delay filters out short spikes, for example a cold room temperature that rises briefly when the door is opened or duct pressure that spikes when a damper moves. Its price is lateness. If the value rises at a constant rate, the value when the alarm is announced is
with the rate of rise (units per minute) and the delay (minutes). So every minute of delay is a minute lost from the response time. The HSE guidance asks for alarm levels to be set so that operators have enough time to carry out their defined response before the condition escalates.
Hysteresis, delays, and alarm limits are part of the configuration of the controller or the BMS server. A technician who finds a chattering alarm collects the evidence, namely the trend of the value, the number of activations per hour, and the times they occur, and then proposes a change to the integrator. The technician does not adjust them.
Alarm floods and standing alarms
An alarm flood is a condition in which alarms arrive faster than the operator can read and respond to them. The most frequently cited example comes from a refinery at Milford Haven, UK, in 1994: in the last 11 minutes before the explosion, two operators had to recognize, acknowledge, and act on 275 alarms. The HSE investigation found that there were too many alarms and that they were poorly prioritized.
HSE quotes alarm rate targets from the EEMUA guide: a long-term average during normal operation of no more than one alarm every ten minutes, and no more than ten alarms displayed in the first ten minutes after a major upset. The alarm rate is calculated as
with the number of new alarms and the observation period, for example alarms per ten minutes. In buildings, the most common major upset is a power failure and the transfer to the generator set. One cause, the voltage dip, produces dozens of consequences: every controller that restarts sends a communication alarm, every fan and pump that stops sends a command and status mismatch alarm, and then room temperatures begin to cross their limits. Alarms that need a specific action, such as a fan that will not restart by itself, are buried among them.
A standing alarm is an alarm that stays active for a long time, often because the equipment is out of service, its sensor is faulty, or its limit does not suit the conditions. Standing alarms fill the alarm list, so new alarms are hard to see. HSE asks for the number of standing alarms to be counted as one measure of the problem, and lists suppressing alarms from out-of-service plant as one of the fixes, together with reviewing alarms that have no defined response and re-engineering repeating alarms. Suppression of that kind is a design and change-control decision, with a list and time limits, not an action of the duty technician.
On the technician’s side, three habits reduce the danger of an alarm flood. First, during a flood, alarms are filtered by priority and the high ones are read first; alarms are not acknowledged page by page without being read. Second, once things are calm, the list of active alarms is compared with the list from before the upset to find equipment that has not recovered. Third, the HSE measures are used as a monthly record: alarms per day, the most frequent alarms, and standing alarms. Those figures go to the building manager as input for an alarm review.
Trend logs and sampling intervals
A trend log is a record of the value of a point over time stored by a controller or the BMS server. There are two common ways of recording. Interval recording stores one value at each fixed interval. Change-of-value recording stores a new value only when it changes by more than a set threshold, so a flat line on the graph does not mean the value is still. Chapter 12 · Data & Measurement on time-series data explains sampling rate, aliasing, and logging deadbands; this chapter covers their use in a BMS.
The sampling interval is the time between two stored values. The interval is chosen from the question. To see whether an AHU starts on schedule, a 15-minute interval is enough. To see a fan that cycles on and off every few minutes, the interval must be much shorter than that cycle. The 1991 GSA guide specification requires trend reports in which the operator can select intervals from 30 seconds to 1 hour and display spans from 1 hour to 5 days.
The number of records and the retention time are calculated with
with the recording period, the sampling interval, the number of records, the record buffer capacity in the controller, and how long the data is held before the oldest record is overwritten. Field controllers usually hold few records, and the server then moves them to a database. The maintenance guide from FEMP, the US federal energy management program, requires trend data to be archived from field equipment to the database no less than once a day, storage capacity for at least two years of data for all points, and timestamps on all data. If the server is down for longer than , the data in between is lost.
Requesting trends is the most useful diagnostic task a technician does. A good request names the points, the interval, the recording period, and the question, for example “supply air temperature, chilled water valve position, and fan status of AHU-3, one-minute interval, for three days, to see whether the valve is hunting.” Creating or changing trend definitions in the controller follows the access rights set by the building manager; in many buildings it is the integrator’s task. The FEMP metering guide also asks for data collection systems to be alarmed to detect gaps and out-of-range values. A trend that suddenly stops is a finding, not a coincidence.
Schedules and manual overrides
An occupancy schedule is the list of times when a system runs in occupied mode and unoccupied mode. A weekly schedule is usually supplemented by exception schedules for holidays and events. The FEMP maintenance guide lists the ability of a BMS to schedule the HVAC system for night setback, holiday and weekend schedules with override control, optimal start and stop, and morning cool-down. The same guide asks operators to review schedules periodically so that they match the occupants’ schedules, and notes that in a humid climate the HVAC system during unoccupied hours is put into standby mode to maintain humidity limits, not simply switched off.
Checking schedules is a comparison, not a program change. The technician compares the BMS schedules with the hotel event schedule, clinic hours, or laundry operating hours, and reports the differences. Schedules for rooms used 24 hours a day, such as wards and isolation rooms, must not follow an office schedule; such rooms are recorded as exceptions that must not be changed without the approval of the person responsible for them. Controller clocks that are out of sync shift every schedule, as explained in Chapter 12 · Data & Measurement on time-series data.
A manual override is an operator command that replaces the decision of the automatic program for a point, for example starting an AHU outside its schedule or holding a valve at a fixed position. There are three common forms: an override from the operator station, a hand switch on a field panel, and an occupant override button on a room sensor. The 1991 GSA guide requires occupant overrides and local lighting switches to revert to the automatic program after 8 hours.
The danger of an override is forgetting it. An override left in place keeps equipment running all night, keeps off equipment that should be running, or prevents command and status alarms from ever appearing. The FEMP maintenance guide asks for every override to be recorded: what was overridden, for what purpose, and when it can be released, and asks for a continuous override inspection program so that all overrides are released as soon as possible. The commissioning guidance in the same document asks for a check that overrides have been released.
Limits on overrides for technicians: only on points assigned by procedure, with the name, reason, and release time recorded in the override register, and only on comfort or operating functions. Overrides are not used on safety points, smoke control, fire dampers, isolation room pressure, or other points marked critical, and are not used to force equipment that has stopped because of a safety device.
Handling an alarm: from the sound to the record
Three words at the operator station are often confused. To acknowledge means that the operator states they have seen the alarm; the condition has not necessarily changed. Return to normal means the value is back within limits. Reset means returning equipment that has locked out after a trip so that it can run again. An alarm that has been acknowledged but has not returned to normal still requires action.
The sequence for handling one alarm:
- Read the whole message: point, kind, value and limit, time, and priority.
- Acknowledge the alarm after reading it, not to stop the sound.
- Check whether the alarm is genuine: compare it with other points and a short trend, and check in the field if needed.
- Carry out the defined response for that alarm, or contact the designated party.
- Report and escalate to a supervisor or another party if the response is outside your authority or the alarm does not return to normal within the set time.
- Record the time, action, name, and final condition in the logbook.
The HSE guidance lists the requirements for operator reliability: a very obvious display of the alarm, few false alarms, a low workload, a simple written response, training, and testing of the response. For building technicians, this means every high-priority alarm has a written response card, and alarms that are not understood are reported for review.
Resetting equipment that has tripped on a safety device, such as a chiller high-pressure switch or a motor overload relay, is done by a person authorized for that equipment after the cause is known, in line with Chapter 17 · Fluids & Thermal on chillers and Chapter 76 · Electrical & Control on safe work.
Worked example
The numbers in the examples below are practice numbers, unless their origin is stated.
Example 1: hysteresis stops a chattering alarm. The high-temperature alarm of a kitchen cold room is set at °C without hysteresis. The cold room controller makes the temperature swing between 5.7 and 6.2 °C with a period of 4 minutes, so the alarm activates and clears times per hour. With a hysteresis of K, the alarm activates at 6 °C and clears only at °C. The lowest point of the swing, 5.7 °C, does not reach 5 °C, so from its first activation the alarm stays active and appears only once. One alarm that stays active says that the temperature keeps touching the limit.
Example 2: the price of a time delay. The chilled water supply temperature rises at a constant rate of K/min after the chiller stops. The alarm limit is 9 °C with a delay of 5 minutes. When the alarm is announced, the temperature is already °C. A short spike lasting 3 minutes when the pumps change over does not trigger the alarm because it is shorter than the delay. If the defined response takes 20 minutes and the temperature must not exceed 14 °C, the time left after the alarm is minutes, exactly enough with nothing to spare, so a proposal for a shorter delay is recorded for the integrator.
Example 3: alarm rate during a flood and in normal operation. In the opening case, 186 alarms arrived in the first ten minutes after the transfer to the generator set, which is times the target of ten alarms. Over a week of normal operation, the log recorded 1,150 alarms. A week contains ten-minute intervals, so alarms per ten minutes, above the target of one alarm per ten minutes. With the 31 standing alarms removed from the flood list, there are still alarms arriving within ten minutes.
Example 4: priority distribution. A hospital BMS server holds 400 alarm points: 60 high, 80 medium, and 260 low. The distribution is percent, percent, and percent. Compared with the reference of 5, 15, and 80 percent, there are three times as many high-priority alarms, 60 against . The list of 60 high alarms goes to the alarm review with the consequence of each if it is not answered, so that their priorities are reassigned by the integrator and the building manager.
Example 5: trend log retention. An AHU controller stores 2,000 records per trend. At a 5-minute interval, minutes, that is, days. At a 1-minute interval it is only days. If the server is shut down for maintenance for two days, the 1-minute trend loses data, while the 5-minute trend survives. For 300 points trended at 1 minute for 30 days, the server stores records.
Example 6: the cost of an override that was not released. The 15 kW supply fan of the outpatient clinic AHU was overridden on for a Saturday event and was found only nine nights later. Outside its schedule the fan ran 14 hours a night, so the energy wasted is kWh. At a tariff of Rp1,500 per kWh, the cost is rupiah, not counting the energy for cooling outdoor air. An override register with a “release no later than” column would have flagged it on Sunday morning.
Stop conditions
- A high-priority alarm cannot be checked or answered within the set time because of an alarm flood or because the technician is alone: escalate immediately to the duty supervisor and ask for help; do not acknowledge alarms without reading them.
- An alarm concerning an isolation room, an operating room, or another room marked critical: report it immediately to the user unit and to the person responsible for air conditioning according to hospital procedure; do not change settings or use an override on that point.
- A fire, smoke, or smoke-control status appears on the BMS station: action follows the fire alarm panel and the building fire officers; the BMS is not used to override, switch off, or delay fire functions.
- Equipment has tripped on a safety device and the BMS offers a remote reset button: do not reset before the cause is known; hand it over to the technician authorized for that equipment.
- An alarm seems to need disabling, delaying, hiding, or a change of limit: do not do it; record the evidence and hand the proposal to the system integrator through change control.
- An override is found that is not recorded or whose owner is unknown: do not simply release it if it involves equipment being worked on; ask the work supervisor and record it in the override register.
- Trends stop, controller clocks jump, or many controllers lose communication at once: report it to the integrator or IT administrator; do not change the network configuration, accounts, or passwords.
Common misinterpretations
Guess what’s wrong first, then open the explanation.
Common misconception
The alarm has been acknowledged, so the problem is solved.
Why is it wrong?
Acknowledging only states that the alarm has been seen. An alarm is finished when the condition has returned to normal and the action has been recorded.
Common misconception
The more alarms, the safer the building.
Why is it wrong?
Alarms that require no action and standing alarms bury the important alarms. The HSE targets actually limit the alarm rate.
Common misconception
Hysteresis and delays make the system calmer at no cost.
Why is it wrong?
Hysteresis makes an alarm clear later, and a delay postpones a genuine alarm. Both must be calculated against the response time required.
Common misconception
A flat trend means the value is constant.
Why is it wrong?
In change-of-value recording, a flat line only means the change has not crossed the threshold. In interval recording, events between two records can be missed.
Common misconception
A manual override is only temporary, so it does not need recording.
Why is it wrong?
An override that is not released is one of the most frequent causes of waste and unsafe conditions, and it can only be traced if it is recorded.
Links to other chapters
Chapter 32 on BMS architecture supplies the points and names that make up alarm messages, and Chapter 104 · Electrical & Control on building communication protocols explains where communication alarms come from. Chapter 12 · Data & Measurement on time-series data and Chapter 16 · Data & Measurement on presenting data help read and display trends. Chapter 34 on building energy dashboards uses trends and schedules to find waste. Chapter 19 on fire alarm panels, Chapter 30 on air conditioning for special hospital rooms, and Chapter 17 · Fluids & Thermal on chillers explain the equipment whose alarms are most demanding.
Exercises and worked solutions
The numbers in the exercises are for practice.
Try to answer first, then open the worked solution.
The high-humidity alarm of an archive room is set at 65 percent with a hysteresis of 3 percent. At what value does the alarm clear, and will a swing between 63 and 66 percent make the alarm chatter?
Check my answer
See the worked solution
The alarm clears at percent. A swing from 63 to 66 percent activates the alarm when it touches 65 percent, but never falls to 62 percent, so the alarm stays active once and does not chatter; it is the humidity that keeps touching the limit that needs checking.
Duct static pressure falls at 20 Pa per minute after a filter suddenly clogs, the low limit is 250 Pa, and the delay is 3 minutes. What is the pressure when the alarm is announced?
Check my answer
See the worked solution
The pressure when the alarm is announced is Pa. The 3-minute delay lets the pressure fall another 60 Pa before anyone knows, so for a fast drop the delay needs to be reassessed by the integrator.
In 24 hours, the BMS station receives 216 alarms. What is the rate per ten minutes, and how does it compare with the normal-operation target?
Check my answer
See the worked solution
A day contains ten-minute intervals, so alarms per ten minutes. That is above the target of one alarm per ten minutes, so the most frequent alarms are listed and handed to the alarm review.
A technician proposes disabling a filter alarm that sounds every day. What should be done?
Check my answer
See the worked solution
An alarm that sounds every day is a sign that the filter, the limit, or the sensor needs checking, not a reason to switch it off. The technician records the trend of the pressure difference and the filter change schedule, replaces the filter if it is actually dirty according to Chapter 16 · Fluids & Thermal on air handling units, and hands a proposal to change the limit to the integrator through change control.
A controller stores 1,440 records per trend. What interval in minutes keeps the data for at least 7 days?
Check my answer
See the worked solution
Seven days contain minutes, and , so the interval is at least 7 minutes. A shorter interval fills the buffer faster and overwrites the data before seven days.
During the morning round, a technician finds the standby chilled water pump in an “off” override with no record. What are the steps?
Check my answer
See the worked solution
An override without a record may come from work still in progress, so the technician does not release it straight away. The technician asks the work supervisor and the supervisor on duty, records the finding in the override register, and the pump is returned to automatic by an authorized person after it is confirmed that no work is being done on that pump.
References
External sources
- Better alarm handlinghse.gov.uk
- Alarm managementhse.gov.uk
- Guide Specification for Direct Digital Control Based Building Automation Systemnvlpubs.nist.gov
- Operations & Maintenance Best Practices: A Guide to Achieving Operational Efficiency, Release 3.0energy.gov
- Metering Best Practices: A Guide to Achieving Utility Resource Efficiency, Release 3.0wbdg.org
Revision history
| Date | Version | Change |
|---|---|---|
| November 13, 2026 | 1 | First published |
How to cite
Kurnia (2026). Alarms, trends, and schedules in a BMS: alarm priorities and classes, hysteresis and time delay, alarm floods, trend logs and sampling intervals, schedules and manual overrides, and handling alarms. ARSTEMA Academy. Revised November 13, 2026. arstema.com/en/academy/konsep/alarm-tren-dan-jadwal-di-bms-prioritas-dan-kelas-alarm-histeresis-dan-tunda-waktu-banjir-alarm-log-tren-dan-selang-cuplik-jadwal-dan-penimpaan-manual-serta-menangani-alarm
Self-check · 3 questions
Check your understanding before moving on
Write your answer in your own words, then compare it with the worked solution from this chapter’s exercises. The system doesn’t judge open answers as right or wrong; you decide which parts you’ve understood.
The high-humidity alarm of an archive room is set at 65 percent with a hysteresis of 3 percent. At what value does the alarm clear, and will a swing between 63 and 66 percent make the alarm chatter?
Duct static pressure falls at 20 Pa per minute after a filter suddenly clogs, the low limit is 250 Pa, and the delay is 3 minutes. What is the pressure when the alarm is announced?
In 24 hours, the BMS station receives 216 alarms. What is the rate per ten minutes, and how does it compare with the normal-operation target?
Nearby chapters
Reading order by shelf: Buildings & Facilities · Chapter 34.
Reading order by shelf: Buildings & Facilities · Chapter 32.