I'm not the engineer who designs the machines. I'm the person who buys the parts for the engineers—50-person automation company, about $1.8M in annual purchasing, 14 vendors, and quite a few lessons learned since I took over in 2020. When something breaks, I'm the one who hears 'order another one' before anyone has looked at why it broke.
After five years, I've learned to slow that process down. Forcing a little review before a replacement purchase has saved us more money than any supplier negotiation I've ever done.
Here's what I tell anyone who asks about motion components, whether it's a linear actuator, an AC servo motor, a NEMA 8 stepper motor, or a VFD: the part is rarely the real problem.
The part that looked right on paper
Let's start with the surface problem. A linear actuator fails at the same spot in every cycle. Carriage hesitates, then stalls, then trips an alarm. The maintenance team replaces it. For a few weeks, the machine runs fine. Then the same alarm comes back.
It looks like a bad part. To be fair, bad batches happen—I've seen them. But after the second identical failure, the issue isn't the part. It's the way the part was selected.
In 2022, we had a run of thomson linear actuators dying at the same point in the stroke. The initial reaction was 'get another one.' I asked to see the duty cycle calc before I wrote the PO. The actuator was rated for 30% duty at the applied load. The machine was running it near 60%. The part wasn't failing randomly; it was exceeding its rated duty cycle every day.
That's an uncomfortable conversation to bring to an engineer. But the data was right there on the thomson linear official website (source: thomson-linear.com, accessed January 2025). It took ten minutes to find.
Why the replacement didn't fix it
Here's the part I had to learn the hard way: if you use the same selection logic, you get the same result. The replacement isn't a fix. It's a repeat.
There are three traps I see constantly in our own purchase history and in the questions we get from customers.
Spec sheets are best-case scenarios
A spec sheet is kinda like a resume: it highlights the best case. It isn't a performance guarantee in your machine. Load, speed, duty cycle, mounting stiffness, ambient temperature—all of those shift the real limit.
I don't have hard data on how many of our component failures trace back to a misread spec sheet. But based on five years of order history, my sense is that it's over half. I wish I had tracked the root cause more carefully from the start. What I can say anecdotally is that almost every repeat failure shared the same pattern: someone chose a part based on the headline number, not the curve.
An AC servo motor is a classic example. The datasheet lists continuous torque and peak torque. It also lists duty cycle assumptions, thermal resistance, and ambient limits. If you size the motor at the continuous limit and ignore the duty cycle, it heats up and trips. It isn't a bad motor. It's an under-sized motor.
At the other end, a NEMA 8 stepper motor is one of the smallest standard stepper frames—about 0.8 inches square. It works well in compact lab equipment and light indexing. But stepper motors lose torque as speed increases, and a NEMA 8 loses it quickly. I've seen a NEMA 8 chosen for a small rotary table, then stall at a speed that looked fine on paper. The torque-speed curve was in the datasheet. Nobody looked at it.
Bigger isn't safer
When a part fails, the natural instinct is to oversize the replacement. I get why. Downtime is expensive, and nobody wants to be accused of undersizing. But oversizing creates its own problems.
A larger stepper motor produces more heat. A larger AC servo motor has more rotor inertia, which can make acceleration worse instead of better. A heavier actuator adds moving mass to the machine, which can shift resonance and excite other components.
Once, we 'upgraded' an actuator to a bigger frame because the smaller one kept tripping on overload. The bigger actuator was strong enough, but the extra mass made the gantry vibrate. We ended up spending more on the 'solution' than the original system cost.
The bigger-is-safer trap is understandable, but it's still a trap.
Motor-to-load mismatch
The third trap is ignoring what the motor is actually driving. A motor can be perfectly sized on paper and still fail if the load has inertia or torque characteristics the motor wasn't chosen for.
This is also where the 'what size VFD for 5hp motor?' question comes up in our office. If you just match nameplate horsepower, you can miss the full-load current. A 5HP motor on a particular voltage may draw more current than a given VFD is rated to supply. According to NEMA MG 1, the motor nameplate carries the information needed to apply the motor correctly. The full-load amp number on that plate is the one that matters for VFD sizing. The drive trips under load, the motor runs rough, or the drive fails. The answer isn't always '5HP VFD'—it depends on the motor's FLA, the input voltage, and whether the load is constant torque or variable torque.
What this actually costs
The replacement part is easy to quantify. The real cost is everything around it.
In 2023, an actuator failure shut down a packaging line for eight hours. The replacement part was $1,400. Overnight freight, a service call, lost production, and the expedited setup added up to something closer to $12,000. The CFO didn't want to hear 'bad part.' He wanted to know why nobody caught the duty cycle issue before the first install.
Every time I'm tempted to just reorder the same part, I run the risk trade. The upside is speed. The risk is a third failure. I keep asking: is two days of lead time worth another eight-hour shutdown?
I don't have an exact breakdown of every expedite fee from that year. I can tell you anecdotally that rush orders were the biggest frustration on our purchasing card, and very few of them were true emergencies. Most were the result of a downstream problem that had been visible for weeks.
That's the expensive pattern. It's not one big failure. It's a series of 'small' replacement orders that never solve the root cause.
What I do differently now
I'm not an engineer, and I don't pretend to be. I've just built a few checks into the purchasing process.
- Before any motion component order, I ask for the application conditions: load, speed, duty cycle, ambient temperature. If those aren't documented, I won't place the order.
- I compare the spec against the actual application, not against a similar machine across town.
- If a replacement fails twice, it's no longer a bad part. It's a selection problem, and we review the design before ordering a third one.
- I keep the thomson linear official website open in a tab when I'm comparing actuator part numbers. Their load/cycle calculators flag limits that I would never catch on my own.
For linear actuators specifically, the thomson-linear line has enough documentation to make a buyer look smart. Use the online selector, check the duty cycle curves, and if you're close to 70-80% of rated load, go up a frame or change the screw lead. It costs a little more upfront and saves a lot of downtime later.
For motors, I ask our controls team to talk me through the torque-speed curve before I approve a PO. If it's an AC servo motor, I ask about continuous vs. peak torque and the duty cycle. If it's a NEMA 8 stepper motor, I ask to see the torque-speed curve and the driver current setting. If someone asks 'what size VFD for 5hp motor?' I don't just say '5HP.' I ask for the full-load amps from the motor nameplate. The VFD's output current rating is the real limit.
So glad I started doing this in 2021. Almost ordered the same replacement actuator twice before I understood the duty cycle issue. Dodged a bullet when I asked for the motor datasheet on a 'replacement' servo—it was 20% smaller than the original spec required. One page of a datasheet saved us a second trip.
Bottom line
If your machine keeps eating the same part, the part isn't your problem. The selection process is.
Spend the time on the specs before you buy. Pull the duty cycle curve, read the motor datasheet, and ask a human being at the manufacturer for confirmation when the application is close to the limit. An informed buyer makes better decisions, and better decisions keep machines running.
I'd rather spend ten minutes explaining options to our engineers than process a $3,000 order that fails again in six months. That's the whole approach now.