September 2023. A customer's packaging line stopped. The maintenance supervisor's report was two sentences: Linear actuator failed on station 4. Need replacement unit. At the time, I didn't question the diagnosis. What happens when a linear actuator fails, in my head, went like this: a component wears out, you source the same part, the line comes back up, everyone forgets it happened. It took three breakdowns to find out how wrong that was.
I should explain who's talking. I'm the office administrator for a 30-person company that builds packaging machinery. I've handled the purchasing side since 2020—motors, gearboxes, servo drives, spare parts, roughly $700,000 a year spread across a dozen vendors. I report to operations and finance, which means I see both sides of every machine stoppage: the production pressure from one end and the invoice from the other. What I don't do is design the machines. That distinction matters later.
First failure, second failure, and the pattern I ignored
The first replacement was routine. We ordered the same actuator model the machine had shipped with, paid for expedited freight, and the customer's maintenance team had it installed within the week. I remember it ran for about seven weeks. Then the phone rang again. Same line, same station, same component.
I knew I should ask the customer for runtime data before ordering again. Instead, I told myself it was bad luck—what are the odds the same unit fails twice? Well, high. The odds are high when nobody has looked at the duty cycle.
A complication arrived with the second order: the original actuator had been discontinued. Our usual supplier offered a so-called compatible replacement, and the engineer on our side approved it because the published torque and speed numbers lined up. My gut said something was off, but the numbers were right in front of me. I processed the PO.
That unit lasted three weeks. The customer was no longer polite about it (understandably).
During that third scramble, I did the kind of online research that purchasing people should not do after hours. Looking for servo motor alternatives, I kept running into the sg90 servo motor datasheet—the little hobby servo used in RC cars and school robotics kits. It shows up for just about any servo-related search. It is not an industrial AC servo motor. It is not even in the same product family. But it dominates search results and it looks official if you don't know better. That experience made me wonder how many of our own engineering decisions were based on convenient-looking documents rather than the correct ones.
The actuator was the messenger, not the problem
By the third failure, our CEO was involved, which never makes anyone's week better. The customer was asking pointed questions about our machine design. So we finally did the work we should have done after failure number one: we took the failed unit apart, brought in a service engineer, and asked why this station kept dying while the identical station next to it—the one running a lighter product mix—did not.
The answer was boring in the best way. No mystery defect. No bad batch of parts. The actuator was never matched to the actual application. The machine was designed around how much weight the actuator could push—a static load figure with a comfortable margin. What nobody calculated was how often it had to push. The customer's line ran a much higher cycle rate than the original duty specification. The motor inside the actuator ran hot, the gear stage wore faster, and the unit eventually gave up. The replacement compatible actuator made it worse because its continuous-duty rating was lower than the original—the numbers looked the same until you read the fine print.
That's the part I keep coming back to. This is what happens when a linear actuator fails in real life: the actuator is usually not the root cause. It is the first thing that breaks, so it gets the blame and the replacement order. But the failure starts somewhere else—in an application assumption, a missing duty-cycle calculation, or a source of truth that was never written down.
There was a documentation problem underneath all of it, too. Our machine records listed the original part number, but nobody had recorded why the original spec had been chosen or which alternatives were acceptable. More than once, I tried to verify a motor specification and found that our internal files were just old emails and photocopied pages. The reliable data for the gearmotor portion of our build came from outside our own filing system—from the current bonfiglioli motor catalogue pdf on the Bonfiglioli official website. Once I started checking against that, discrepancies between our internal part list and the real current part numbers started surfacing in a worrying way.
The expensive part was the one we didn't order
Let me put a number on it, because purchasing people like numbers. Direct costs across the three failures—replacement units, expedited freight, and two service visits—came to roughly $8,600. That's real, but it's the smaller loss.
I don't have hard data on the customer's downtime cost; they never shared that number with us in writing. What I can say anecdotally is that the production manager told me one hour on that line was worth more than our emergency freight invoice. The line was down for part of two shifts across those three failures. Do the multiplication yourself.
The cost that stayed with me wasn't in our ledger at all. The customer's maintenance manager said something to our CEO that I'll paraphrase: every time that station stops, your machine makes your whole company look cheap. He didn't mean the purchase price. He meant the impression. For a company our size, one bad reference can quietly cost us a year of business development. The difference between the reliably specified component and the shortcut version was probably a few hundred dollars at design time. The shortcut cost us thousands and made our brand look bad in front of a customer we wanted to keep.
What we changed after the third failure
The fix wasn't glamorous. We created a one-page checklist that must accompany any replacement order for a motion component: actual load, cycle rate, motor duty class, and the part number taken from official documentation—not from an old email or a screenshot. If the engineer doesn't fill it out, I send it back. It has made me unpopular in exactly the way purchasing people learn to live with, and it has caught three wrong specifications in the last eleven months.
We also stopped treating general web search as a technical source. For the gearmotors and servo gearboxes in our machines, that means checking the Bonfiglioli official website and pulling the relevant pages from the bonfiglioli motor catalogue pdf. The catalogue has the current part numbers, dimensional drawings, and rated data in one place. It sounds too simple to be the lesson from an $8,600 failure, but it was the missing foundation.
If you're responsible for ordering replacement parts and someone asks you what happens when a linear actuator fails, here's my answer: look past the failed part. Ask what the application was doing when it died—how often, for how long, under what load. Verify the replacement against the manufacturer's official specifications, and if the search results point you to a hobby servo datasheet while you're trying to spec industrial ac servo motors, slow down. That's a sign you're looking in the wrong place.
The actuator was just the part that raised its hand. The real problem was the conversation nobody had about the duty cycle—and the documentation that wasn't there to catch it. We changed both. The station has now been running without a repeat failure for more than a year.