2026-09-07 · Elena Markovic
Harmonic Drive Gearbox Repair: A Six-Step Checklist from a Technician Who Learned the Hard Way
-
Step 1: Let the Machine Tell Its Story Before You Touch It
-
Step 2: Measure Lost Motion at the Output, Not by “Feeling” It
-
Step 3: Check the Ball Bearings Before You Condemn the Gearbox or Motor
-
Step 4: Disassemble Like You Are Going to Be Audited
-
Step 5: Repair What the Evidence Supports — Not What the Manual Makes Easy
-
Step 6: Reassembly, Lubrication, and Efficiency Verification
-
Mistakes That Still Catch Us
Before you authorize another harmonic drive gearbox repair — or replace a costly servo motor on a hunch — I want you to read this. I’ve spent about eight years on repair intake at a precision motion shop. In that time, I’ve personally made and documented 17 expensive mistakes that totaled roughly $37,000 in scrapped parts, unpaid rework, and missed delivery dates. That number still hurts. It’s also why our bench now runs on a strict diagnostic checklist.
This checklist is not generic advice. It’s for people who support axes where a servo motor, a harmonic drive, and a set of bearings all work together. That includes machines with FANUC servo motors, OEM motion stages, indexing tables, and anything where “the motor just stopped holding position” turns out to be a $50 bearing pretending to be a $3,000 gearbox problem. There are six steps. Step 1 costs you an hour of reading. Step 3 is the one almost everyone skips. Step 5 is where the money is saved or burned.
Step 1: Let the Machine Tell Its Story Before You Touch It
Start with the alarm log, the maintenance history, and the operator’s own words. I know it sounds obvious. In my first year, I skipped it because the customer was in a hurry. The alarm read “position deviation excess” and pointed at the motor, so I sold them a replacement FANUC servo motor. The motor was fine. The real problem was a damaged harmonic drive that had been overloaded in a crash two weeks earlier. The new motor masked the symptoms for a few days, then burned out because the gearbox was still dragging. That mistake cost about $2,800 in parts plus a week of angry phone calls.
What do you actually look for? Write down when the problem first appeared, whether it happens after warm-up, and what changed in the machine right before the fault — a tool change, a jam, a recent lubrication service, a new operator. A harmonic drive that loses stiffness gradually is a wear story. A drive that fails suddenly is usually an impact story. Those two stories lead to completely different repairs.
Step 2: Measure Lost Motion at the Output, Not by “Feeling” It
Harmonic drives are advertised with zero backlash, but in the repair world “zero” is never the problem. The problem is lost motion that has grown beyond the machine’s acceptable error. You need to measure it the way your customer’s process will experience it.
Lock the input side, mount a dial indicator on the output flange, apply a small alternating torque, and record the indicator movement. Do this at several positions around the rotation of the output — not just one comfortable spot. A worn flexspline will often show different lost motion depending on where the wave generator stopped. I like to take five readings and use the worst one, because that is what the customer’s encoder will see when the machine is under load.
One mistake I made twice was measuring lost motion while the servo brake was still applied. The brake flexed enough to make a good gearbox look bad. We quoted a full overhaul, then discovered the brake was dragging. Not every “backlash” reading is actually backlash. Disconnect or release the brake, then repeat the test. The difference will surprise you.
Step 3: Check the Ball Bearings Before You Condemn the Gearbox or Motor
This is the step my own technicians resist most, because it feels like going backward. You have an expensive harmonic drive sitting on the bench and a diagnosis that says “replace the gearbox.” Before you do that, ask what happens if a ball bearing goes out. Most people imagine a sudden crash, with metal chunks and a stopped shaft. In reality, a failing ball bearing usually whispers first. It creates a small vibration, a periodic torque ripple, or a faint noise that gets blamed on the motor, the coupling, or the gearbox. By the time it screams, the damage has spread to every component on that axis.
I saw this clearly in 2024 on a machine where the load carriage rode on ball bearing wheels along a steel rail. One of the wheel bearings had developed a tiny flat spot. Every revolution of that wheel sent a small shock into the axis. The harmonic drive amplified the disturbance into a repeating position error. The customer was convinced the servo motor was failing. We traced it to a $45 bearing in a wheel assembly. Replace the wheel bearing, and the error disappeared. A new motor or a rebuilt harmonic drive would not have fixed it.
So check the load path first. Spin every accessible bearing by hand. Listen for roughness. Look at the old grease for metal particles. If the axis uses ball bearing wheels or cam followers, inspect the rolling surfaces for flat spots. A small fault on the load side can easily fake a motor fault, and no amount of FANUC servo motor repair work will solve a wheel bearing that was dying before the motor ever complained.
Step 4: Disassemble Like You Are Going to Be Audited
Once you commit to opening the unit, slow down. I know the production manager is standing behind you, but the worst thing you can do is pull a wave generator out of a flexspline without first documenting how it was seated.
Take photos before every disassembly step. Mark the orientation of the components. Note where the flexspline teeth show wear and where the old grease is discolored. If the harmonic drive has an elliptical wave generator bearing, protect it from contamination the moment it is exposed. A single grain of debris in that bearing can create a false “worn gearbox” diagnosis later.
If a servo motor is also being sent out for repair, mark the housing relationship before you remove it. Modern servo motors, especially absolute encoder models, are sensitive to reassembly mistakes. When the motor comes back from a repair shop, verify that the commutation or encoder alignment was not disturbed during the bearing replacement. I once installed a returned motor and got a violent fault on the first run. The repair shop had replaced the motor bearings and reassembled the rotor without re-establishing the electrical zero. The motor was functionally new, but it was unusable until we sent it back. That was a two-week delay and a lesson in asking the right questions before you approve any servo motor repair.
Step 5: Repair What the Evidence Supports — Not What the Manual Makes Easy
Here is where I have saved customers the most money and wasted the most when I got it wrong. In 2019, I quoted complete harmonic drive replacements twice because it was easier than sourcing individual flexsplines. Both customers approved the work. Both machines ran fine. But both budgets were far larger than necessary, and one customer told me plainly: “We didn’t need a new car. We needed new tires.”
If the circular spline teeth are intact and the housing bore is undamaged, a worn flexspline and a degraded wave generator bearing can often be replaced without replacing the entire harmonic drive. That is not true in every case — a cracked flexspline or damaged circular spline means the whole unit should be replaced. But you don’t know until you inspect. The evidence, not the fear of a second failure, should drive the repair scope.
I also want to address the “cheapest quote” trap. In my experience, the lowest repair estimate is rarely the least expensive repair. One of the most expensive jobs I ever approved had the lowest price on paper. The shop skipped a critical bearing inspection, reassembled the unit with a slightly bent wave generator, and the axis failed again three weeks later. Counting the emergency downtime, the second shipment, and the lost production, the customer paid more than a proper repair would have cost in the first place. My rule now is simple: ask what the repair includes, not just what it costs.
Step 6: Reassembly, Lubrication, and Efficiency Verification
After reassembly, the next mistake is trusting that everything is fine because it rotates smoothly by hand. A harmonic drive can feel perfect on the bench and still fail under load if it was lubricated incorrectly or assembled with the wrong preload.
Use the grease specified by the manufacturer. This is not the place for “any grease from the cabinet.” Harmonic drives depend on a thin, consistent film inside the flexspline; over-packing the grease creates high drag and heat, and under-packing causes premature tooth wear. The right amount usually looks surprisingly small to a technician who is used to greasing standard gearboxes.
Now check the efficiency. The harmonic drive efficiency listed on most public spec sheets sits around 80% at rated torque, though it varies with reduction ratio, input speed, and temperature — look at the actual datasheet for your unit. If the rebuilt drive draws noticeably higher current than it did before the failure, or if the housing gets hot during a no-load run, something is wrong. I once delivered a rebuilt unit that ran perfectly at low speed but drew 30% more current at rated speed. The cause was a misaligned wave generator. On paper, the backlash was fine. In practice, the unit was wasting energy and would have failed prematurely.
Mistakes That Still Catch Us
After years of this work, I keep a short list of things that still catch us when we get rushed:
- Trusting the absence of an alarm as proof that the motor is healthy. Servo drives do not always report a failing bearing until the fault has propagated.
- Greasing until grease appears at the seal. More is not better in a harmonic drive. Too much grease raises the temperature and lowers efficiency.
- Skipping a loaded run-in test. A unit that seems fine unloaded can show vibration, noise, or current spikes after a few hours under real load. Run it before you send it back.
The checklist I use has caught 23 avoidable replacements in the last 18 months. That is not because my team is careless; it is because repair decisions are hard to make correctly under production pressure. If you take one thing from this article, make it this: the most expensive repair is the one based on a guess, and the cheapest one is rarely the one with the lowest estimate. Investigate first. Measure second. Replace only what the evidence supports. Then verify the efficiency, the temperature, and the load before you call the job done.