Why Our Fire Alarm Went Off at Random — And What It Revealed About Our Safety Specs

Why Our Fire Alarm Went Off at Random

Between February and April 2024, our shop floor fire alarm went off eleven times. Three of those were at 2 a.m. Two happened during lunch. Once, right at shift change, everyone assumed it was a scheduled drill.

After the third false alarm, the night manager blamed "the old system." After the sixth, our maintenance lead said it was probably a sensor calibration issue. By the tenth, a line worker looked up and said, "Might not be the system. Might be the welders."

He was right.

I'm the quality and compliance manager at a mid-sized fabrication company. I review every batch that goes out to customers—roughly 200+ SKUs a year. A bad batch means a customer complaint, and a customer complaint lands on my KPI. So when the alarm started acting up, I didn't treat it as a maintenance issue. I treated it as a non-conformity to investigate.

Now—I'm not a fire systems engineer. I can't speak to how smoke detectors differentiate between welding fume and actual smoke. That's outside my expertise. But what I can tell you from a quality-control perspective is how to trace a problem to its root cause and then build a fix that holds.

What I should have done from day one was track the alert log more methodically. I wish I'd done that—it would've saved us about a month of head-scratching. What I can say anecdotally is that once I finally pulled the log and cross-referenced it with our welding schedule, a pattern jumped out fast: 9 of the 11 alerts happened within 6 to 8 minutes of welding starting up.

So the issue wasn't the system. It was the weld smoke drifting out of the welding bay and into the corridor. There was no physical barrier between the welding area and the adjacent paint station, and fume was hitting a ceiling-mounted smoke detector. For anyone else who's ever found themselves Googling why did my fire alarm randomly go off—check the layout of your workspace before you call an electrician.

But before I went to engineering with a barrier solution, I did what any quality manager does when something feels off: I audited every piece of equipment in that welding bay.

If there's one thing I've learned from years of doing inspections, it's this: when one thing goes wrong, the stuff around it usually isn't as tight as you think. Not a curse. Just experience.

What I found was worse than I expected.

First, hand protection. Two of our four welders were wearing leather welding gloves with visible cracks at the knuckle and burn-through on the thumb. I checked the packaging. No manufacturer name, no batch number, no date. Those gloves had been bought through an intermediary last fall, and the contract called them "premium leather welding gloves." That was it. No thickness spec. No cut-resistance rating. Not even a reference to the ANSI/ISEA 105 standard for hand protection. The problem wasn't the gloves themselves—it was that our purchase spec allowed this. We never told the supplier what we actually needed, so they filled in the blanks with whatever was cheapest.

Then there was the eye protection. We'd been using edge eyewear for years. Consistent supplier, brand, no issues. But here's the thing—we'd never verified that the specific model matched the task. Our welding bay was using a model from the edge eyewear line, but it wasn't designed for the kind of peripheral reflection you get when you're working close to a bright arc. Two welders had complained about glare off the lens edges. One had quietly gone out and bought his own pair from a hardware store. When workers start bringing their own safety equipment to a job that already provides it, that's a signal you can't ignore.

That last part annoyed me more than anything. We were paying for edge-eyewear. If our people were wearing something else because the model didn't fit the job, that meant our procurement was effectively wasting money.

So I pulled up the edge eyewear catalog and picked a different model—the edge eyewear Robson safety glasses—and asked my welders to trial them for a few weeks.

The result actually surprised me. The same guys who'd been the loudest complainers about eye protection became the strongest advocates for the switch once they tried the Robson. One of our senior welders—a guy with 18 years under his belt—told me the Robson's nose bridge sat in exactly the right spot relative to his face shield, so he stopped having to push the glasses up mid-weld. That's the kind of thing you can't measure from a spec sheet, but it matters when someone's wearing the glasses for eight hours a day.

I ran a small blind test after that. I mixed up several pairs—the Robson, two other models, and a no-name pair—and asked seven welders to rate them on comfort and fit without knowing which was which. Five out of seven picked the Robson.

On the cost side, the Robson ran about 40% more per pair than the no-name glasses some of our welders had been buying themselves. But if that eliminates the incentive to buy their own replacements, the purchase price evens out. Not to mention the compliance benefits of having one standard across the whole crew. There's something satisfying about watching a spec sheet translate into something people actually choose to wear.

There's one more piece to this.

To fix the smoke drift, we needed to physically separate the welding bay from the adjacent stations. My first instinct was to use a batch of hog wire fence panels we already had in inventory. Cheap, available, right there. Done.

Then I measured the mesh. Two-inch by four-inch gaps. On paper it looked fine for keeping people out. But for keeping welding spatter in? Not even close. Molten metal would pass through those gaps like they weren't there and land on the paint station floor. Using those panels would have been worse than nothing—it would have given everyone a false sense of security. We ended up spec'ing a proper steel isolation barrier instead, not a repurposed fence panel.

Lesson logged: not every fence panel is a safety barrier. And "we already have it" is not a valid reason to skip the spec.

Looking back, the whole chain—random fire alarms → welding bay audit → cracked leather welding gloves with no spec → wrong edge eyewear model → hog wire fence panels that almost became a camouflage solution—the root cause was one thing, repeated in different forms: vague purchasing specifications.

Vague specs mean you're outsourcing the decision to the vendor's interpretation of "industry standard." And when a vendor is operating in a gray area, their interpretation isn't always going to favor you.

I changed two things after that. First, every PPE purchase order now requires the relevant standard number on the spec sheet—ANSI Z87.1 for eyewear, ANSI/ISEA 105 for gloves. No exceptions. Second, I added a verification step at receiving: random sampling, checked against the spec, documented. Basic stuff. But we'd fallen down on the basics.

As for the fire alarms—once the welding bay got a proper barrier and we sealed the gap to the corridor, we didn't have a single random alarm for the rest of the year. People finally stopped looking at the ceiling every time the shift changed. I'd call that a win for everyone's peace of mind.

There's one thought I keep coming back to. We almost installed those hog wire fence panels because they were in inventory and they were cheap. If nobody had measured that mesh gap on a Tuesday afternoon, welding spatter could have gone through that fence and hit the next station. Maybe nothing would have happened. But "probably fine" isn't a quality standard. "We know it's fine" is. The difference between those two is every spec parameter you verified before the audit.

Now when I sit down with a supplier, my first question isn't "How much?" It's "What's not included in this price?" Boundaries first. Numbers second. I wrote that on the first page of my work notebook and I've been asking it ever since.