Latest / NOXTalks / Crying Wolf: How Alarm Fatigue Quietly Endangers Your Building

Transcript

Crying Wolf: How Alarm Fatigue Quietly Endangers Your Building · NOXTalks

• 0:07
  SPEAKER_00: Welcome to Knox Talks, the AI-generated podcast exploring the groundbreaking ways technology shapes our built environment.
• 0:14
  SPEAKER_00: Today we're diving into a story you already know: The Boy Who Cried Wolf.
• 0:19
  SPEAKER_00: Except in this version, the boy is your building automation system, and it's crying wolf a few thousand times a day.
• 0:26
  SPEAKER_00: Here's the uncomfortable truth about most BAS installations.
• 0:30
  SPEAKER_00: The alarm system was never designed.
• 0:33
  SPEAKER_00: It just happened.
• 0:35
  SPEAKER_00: Every point got a high limit and a low limit at commissioning, someone clicked accept all, and 20 years later, the alarm log scrolls past faster than anyone can read it.
• 0:44
  SPEAKER_00: Operators cope the only way humans can.
• 0:47
  SPEAKER_00: They stop looking, and that works fine, right up until the one alarm that actually matters arrives, looking exactly like the 400 that don't.
• 0:56
  SPEAKER_00: The process industries learn this lesson the hard way.
• 1:00
  SPEAKER_00: July 1994.
• 1:02
  SPEAKER_00: The Texaco Refinery at Milford Haven in Wales.
• 1:05
  SPEAKER_00: A lightning strike kicks off a plant upset, and for hours the control room is flooded with alarms.
• 1:11
  SPEAKER_00: In the final 11 minutes before the explosion, according to the Health and Safety Executives' investigation, two operators had to recognize, acknowledge, and act on 275 alarms.
• 1:22
  SPEAKER_00: That's one every two to three seconds.
• 1:25
  SPEAKER_00: 26 people were injured, and the damage ran to roughly 48 million pounds.
• 1:30
  SPEAKER_00: The investigators didn't just blame a valve, they found too many alarms, poorly prioritized, presented faster than any human could respond.
• 1:39
  SPEAKER_00: That incident directly shaped EEMUA 191, the guideline that still defines what a healthy alarm system looks like.
• 1:47
  SPEAKER_00: Now, a commercial building isn't a refinery, but the human factors are identical.
• 1:53
  SPEAKER_00: A tired brain tuning out noise at 2 in the morning doesn't care whether the noise comes from a cracking unit or an air handler.
• 1:59
  SPEAKER_00: So what does healthy actually look like?
• 2:01
  SPEAKER_00: EEMUA 191 and the American Standard ANSIISA 18.2 converge on some surprisingly concrete numbers.
• 2:12
  SPEAKER_00: In normal operation, an operator should see no more than about one alarm every 10 minutes.
• 2:17
  SPEAKER_00: Call it roughly 150 a day.
• 2:19
  SPEAKER_00: During an upset, no more than 10 alarms in the first 10 minutes, and priorities should follow a rough 5, 15, 80 split, 5% critical, 15% high, 80% low.
• 2:33
  SPEAKER_00: When everything's flagged urgent, nothing is.
• 2:36
  SPEAKER_00: Compare that to a typical campus BAS generating thousands of alarms a day, and the gap is humbling.
• 2:44
  SPEAKER_00: Here's the fix, and it borrows the process industry's playbook almost word for word.
• 2:49
  SPEAKER_00: It's called alarm rationalization.
• 2:52
  SPEAKER_00: You sit down, controls contractor, operators, whoever owns the response, and you put every single alarm through a simple test.
• 3:01
  SPEAKER_00: If this fires at three in the morning, what does someone actually do about it?
• 3:05
  SPEAKER_00: If there's a defined operator action, the alarm stays, and it gets a priority based on consequence in time to respond.
• 3:14
  SPEAKER_00: If there's no action, then it's not an alarm, it's data.
• 3:18
  SPEAKER_00: Demote it to an event log or delete it.
• 3:21
  SPEAKER_00: Teams that run this exercise often report cutting alarm volume by half or more because so much of the load turns out to be duplicates, chattering points, and set points nobody ever tuned after commissioning.
• 3:34
  SPEAKER_00: Second move, routing.
• 3:36
  SPEAKER_00: Only actionable, critical alarms should reach a human in real time.
• 3:40
  SPEAKER_00: And the ones that do should land somewhere with accountability, which usually means integration with your CMMS.
• 3:47
  SPEAKER_00: An alarm that generates a work order with an owner and an escalation path gets closed.
• 3:52
  SPEAKER_00: An alarm that flashes on a graphic nobody's watching gets ignored.
• 3:56
  SPEAKER_00: Again.
• 3:58
  SPEAKER_00: Third, and this is the culture shift, track your alarm rate as a KPI.
• 4:03
  SPEAKER_00: Process plants report alarms per operator per day the same way they report energy or downtime.
• 4:10
  SPEAKER_00: The moment that number goes on a dashboard, alarm health becomes something the team manages instead of something it endures.
• 4:16
  SPEAKER_00: Chattering points get fixed, priority creep gets caught, and when the rate suddenly spikes, that's a signal in itself.
• 4:24
  SPEAKER_00: Something in the building just changed.
• 4:26
  SPEAKER_00: One more thing worth saying plainly, alarm fatigue isn't an operator failure.
• 4:31
  SPEAKER_00: It's a design failure that operators absorb on behalf of everyone who never looked at the alarm summary.
• 4:36
  SPEAKER_00: The people ignoring alarms today are behaving rationally, given the system they've been handed.
• 4:42
  SPEAKER_00: Because here's the thing about the boy who cried wolf: the wolf was real.
• 4:46
  SPEAKER_00: It showed up at the end of the story exactly when nobody was listening anymore.
• 4:51
  SPEAKER_00: Every unmanaged alarm system is quietly writing that same ending.
• 4:55
  SPEAKER_00: The question isn't whether your building will ever face its one real wolf, it's whether on that day the alarm announcing it will sound like anything other than noise.